ASTM D6620-19

Standard Practice for Asbestos Detection Limit Based on Counts

SIGNIFICANCE AND USE
4.1 The DL concept addresses potential measurement interpretation errors. It is used to control the likelihood of reporting a positive finding of asbestos when the measured asbestos level cannot clearly be differentiated from the background contamination level. Specifically, a measurement is reported as being “below the DL” if the measured level is not statistically different than the background level.  
4.2 The DL, along with other measurement characteristics such as bias and precision, is used when selecting a measurement method for a particular application. The DL should be established either at the method development stage or prior to a specific application of the method. The method developer subsequently would advertise the method as having a certain DL. An analyst planning to collect and analyze samples would, if alternative measurement methods were available, want to select a measurement method with a DL that was appropriate for the intended application.5 The most important use of the DL, therefore, takes place at the planning stage of a study, before samples are collected and analyzed.
SCOPE
1.1 This practice presents the procedure for determining the detection limit (DL)2 for measurements of fibers or structures3 using microscopy methods.  
1.2 This practice applies to samples of air that are analyzed either by phase contrast microscopy (PCM) or transmission electron microscopy (TEM), and samples of dust that are analyzed by TEM.  
1.3 The microscopy methods entail counting asbestos structures and reporting the results as structures per cubic centimeter of air (str/cc) or fibers per cubic centimeter of air (f/cc) for air samples and structures per square centimeter of surface area (str/cm2) for dust samples.  
1.4 The values stated in SI units are to be regarded as standard. No other units of measurement are included in this standard.  
1.5 This standard does not purport to address all of the safety concerns, if any, associated with its use. It is the responsibility of the user of this standard to establish appropriate safety, health, and environmental practices and determine the applicability of regulatory limitations prior to use.  
1.6 This international standard was developed in accordance with internationally recognized principles on standardization established in the Decision on Principles for the Development of International Standards, Guides and Recommendations issued by the World Trade Organization Technical Barriers to Trade (TBT) Committee.

General Information

• Status: Published
• Publication Date: 31-Dec-2018
• ICS: 13.040.99 - Other standards related to air quality
• Technical Committee: D22 - Air Quality
• Drafting Committee: D22.07 - Sampling, Analysis, Management of Asbestos, and Other Microscopic Particles

Relations

• Replaces: ASTM D6620-06(2010) - Standard Practice for Asbestos Detection Limit Based on Counts
• Effective Date: 01-Jan-2019
• Refers: ASTM D1356-20a - Standard Terminology Relating to Sampling and Analysis of Atmospheres
• Effective Date: 01-Sep-2020
• Refers: ASTM D1356-20 - Standard Terminology Relating to Sampling and Analysis of Atmospheres
• Effective Date: 15-Mar-2020
• Refers: ASTM D6480-19 - Standard Test Method for Wipe Sampling of Surfaces, Indirect Preparation, and Analysis for Asbestos Structure Number Surface Loading by Transmission Electron Microscopy
• Effective Date: 01-Jun-2019
• Refers: ASTM D7712-18 - Standard Terminology for Sampling and Analysis of Asbestos
• Effective Date: 15-Apr-2018
• Refers: ASTM D1356-15a - Standard Terminology Relating to Sampling and Analysis of Atmospheres
• Effective Date: 15-Oct-2015
• Refers: ASTM D1356-15 - Standard Terminology Relating to Sampling and Analysis of Atmospheres
• Effective Date: 01-Jul-2015
• Refers: ASTM D1356-14b - Standard Terminology Relating to Sampling and Analysis of Atmospheres
• Effective Date: 01-Dec-2014
• Refers: ASTM D1356-14a - Standard Terminology Relating to Sampling and Analysis of Atmospheres
• Effective Date: 01-May-2014
• Refers: ASTM D1356-14 - Standard Terminology Relating to Sampling and Analysis of Atmospheres
• Effective Date: 15-Jan-2014
• Refers: ASTM D7712-11e1 - Standard Terminology for Sampling and Analysis of Asbestos
• Effective Date: 01-Sep-2011
• Refers: ASTM D7712-11 - Standard Terminology for Sampling and Analysis of Asbestos
• Effective Date: 01-Sep-2011
• Refers: ASTM D6480-05(2010) - Standard Test Method for Wipe Sampling of Surfaces, Indirect Preparation, and Analysis for Asbestos Structure Number Concentration by Transmission Electron Microscopy
• Effective Date: 01-Oct-2010
• Refers: ASTM D1356-05(2010) - Standard Terminology Relating to Sampling and Analysis of Atmospheres
• Effective Date: 01-Apr-2010
• Refers: ASTM D6281-09 - Standard Test Method for Airborne Asbestos Concentration in Ambient and Indoor Atmospheres as Determined by Transmission Electron Microscopy Direct Transfer (TEM)
• Effective Date: 01-Dec-2009

Overview

ASTM D6620-19: Standard Practice for Asbestos Detection Limit Based on Counts is an international standard developed by ASTM International. This practice outlines the procedure for determining the detection limit (DL) for asbestos measurements conducted using microscopy methods. It is essential for laboratories, analysts, and environmental professionals involved in the sampling and analysis of asbestos in air and dust samples. ASTM D6620-19 helps control the likelihood of reporting false positives by establishing clear guidelines for distinguishing real asbestos presence from background contamination.

Key Topics

• Detection Limit (DL)
  The DL is a statistical threshold above which an asbestos measurement is considered distinguishable from background noise. If a measurement is not statistically higher than the background, it is reported as "below the detection limit."
• Measurement Methods
  • Phase Contrast Microscopy (PCM): Used primarily for airborne asbestos (fibers per cubic centimeter (f/cc)).
  • Transmission Electron Microscopy (TEM): Used for both air (structures per cubic centimeter (str/cc)) and dust samples (structures per square centimeter (str/cm²)).
• Background Determination
  The background level includes analyst counting errors and contamination from filters or laboratory procedures. Blank samples are analyzed to assess and minimize background influences.
• Statistical Principles
  The standard employs hypothesis testing and the Poisson distribution to set decision values, control error rates, and calculate detection limits with specified assurance probabilities.
• Measurement Reporting
  Results are flagged as "below detection limit" if they cannot be differentiated from background. This affords protection against false positive asbestos findings.

Applications

• Airborne Asbestos Monitoring
  Used in occupational and environmental hygiene to monitor workplace and ambient asbestos exposure levels, ensuring compliance with asbestos regulations.
• Dust Sampling and Surface Analysis
  Assists in determining asbestos contamination on surfaces using wipe sampling and microvacuum techniques followed by TEM analysis.
• Method Validation and Laboratory QA/QC
  Establishes detection limits for new analysis methods or prior to specific applications, helping laboratories validate their methodologies and participate confidently in interlaboratory comparisons.
• Study Planning and Method Selection
  Provides critical information for choosing a measurement method with an appropriate detection limit for project-specific needs. Correct selection at the planning stage ensures the method’s sensitivity matches the expected asbestos concentrations.
• Regulatory Compliance
  Facilitates adherence to government and industrial safety standards by providing clear, reproducible detection limit criteria for asbestos measurements.

Related Standards

• ASTM D1356: Terminology Relating to Sampling and Analysis of Atmospheres.
• ASTM D5755: Microvacuum Sampling and Indirect Analysis of Dust by TEM for Asbestos Structure Number Surface Loading.
• ASTM D6281: Airborne Asbestos Concentration in Ambient and Indoor Atmospheres as Determined by TEM Direct Transfer.
• ASTM D6480: Wipe Sampling of Surfaces and Analysis for Asbestos Structure Number Surface Loading by TEM.
• ASTM E456 & D7712: Terminologies Relating to Quality, Statistics, and Sampling and Analysis of Asbestos.

Practical Value

Adhering to ASTM D6620-19 enhances asbestos detection accuracy, minimizes interpretation errors, and supports defensible decision-making in risk assessment and regulatory reporting. By standardizing detection limits, the practice ensures data reliability across laboratories, increases confidence in negative results, and enables cost-effective asbestos risk management.

---

Keywords: ASTM D6620, asbestos detection limit, asbestos analysis, microscopy, phase contrast microscopy, transmission electron microscopy, air sampling, dust sampling, background contamination, statistical hypothesis testing, laboratory QA/QC, regulatory compliance.

Standards Content (Sample)

This international standard was developed in accordance with internationally recognized principles on standardization established in the Decision on Principles for the
Development of International Standards, Guides and Recommendations issued by the World Trade Organization Technical Barriers to Trade (TBT) Committee.
Designation: D6620 − 19
Standard Practice for
Asbestos Detection Limit Based on Counts
This standard is issued under the fixed designation D6620; the number immediately following the designation indicates the year of
original adoption or, in the case of revision, the year of last revision.Anumber in parentheses indicates the year of last reapproval.A
superscript epsilon (´) indicates an editorial change since the last revision or reapproval.
1. Scope 2. Referenced Documents
1.1 This practice presents the procedure for determining the 2.1 ASTM Standards:
2 3
detection limit (DL) for measurements of fibers or structures D1356Terminology Relating to Sampling and Analysis of
using microscopy methods. Atmospheres
D5755TestMethodforMicrovacuumSamplingandIndirect
1.2 This practice applies to samples of air that are analyzed
Analysis of Dust by Transmission Electron Microscopy
either by phase contrast microscopy (PCM) or transmission
for Asbestos Structure Number Surface Loading
electron microscopy (TEM), and samples of dust that are
D6281Test Method forAirborneAsbestos Concentration in
analyzed by TEM.
Ambient and Indoor Atmospheres as Determined by
1.3 The microscopy methods entail counting asbestos struc-
TransmissionElectronMicroscopyDirectTransfer(TEM)
tures and reporting the results as structures per cubic centime-
D6480TestMethodforWipeSamplingofSurfaces,Indirect
ter of air (str/cc) or fibers per cubic centimeter of air (f/cc) for
Preparation, and Analysis for Asbestos Structure Number
airsamplesandstructurespersquarecentimeterofsurfacearea
Surface Loading by Transmission Electron Microscopy
(str/cm ) for dust samples.
D7712Terminology for Sampling andAnalysis ofAsbestos
1.4 The values stated in SI units are to be regarded as
3. Terminology
standard. No other units of measurement are included in this
standard. 3.1 Definitions:
3.1.1 For terms not defined in this practice, see Terminolo-
1.5 This standard does not purport to address all of the
gies D1356 and D7712.
safety concerns, if any, associated with its use. It is the
responsibility of the user of this standard to establish appro-
3.2 Definitions of Terms Specific to This Standard:
priate safety, health, and environmental practices and deter-
3.2.1 average,n—thesumofasetofmeasurements(counts)
mine the applicability of regulatory limitations prior to use. divided by the number of measurements in the set.
1.6 This international standard was developed in accor-
3.2.1.1 Discussion—The average is distinguished from the
dance with internationally recognized principles on standard- mean. The average is calculated from data and serves as an
ization established in the Decision on Principles for the
estimate of the mean. The mean (also referred to as the
Development of International Standards, Guides and Recom- population mean, expected value,or first moment) is a param-
mendations issued by the World Trade Organization Technical
eter of the underlying statistical distribution of counts.
Barriers to Trade (TBT) Committee.
3.2.2 background, n—a statistical distribution of structures
introducedby(i)analystcountingerrorsand(ii)contamination
on an unused filter or contamination as a consequence of the
sample collection and sample preparation steps.
ThispracticeisunderthejurisdictionofASTMCommitteeD22onAirQuality
and is the direct responsibility of Subcommittee D22.07 on Sampling, Analysis,
3.2.2.1 Discussion—This definition of background is spe-
Management of Asbestos, and Other Microscopic Particles.
cific to this practice. The only counting errors considered in
Current edition approved Jan. 1, 2019. Published January 2019. Originally
this definition of background are errors that result in an
approvedin2000.Lastpreviouseditionapproved2010asD6620–06(2010).DOI:
10.1520/D6620-19. over-count(thatis,falsepositives).Analystcountingerrorsare
TheDLalsoisreferredtointhescientificliteratureaslimitofdetection(LOD),
method detection limit (MDL), and other similar descriptive names.
3 4
For purposes of general exposition, the term “structures” will be used in place For referenced ASTM standards, visit the ASTM website, www.astm.org, or
of “fibers or structures.” In the examples in Section 8, the specific term, “fiber” or contact ASTM Customer Service at [e-mail]. For Annual Book of ASTM
“structure,”isusedwhereappropriate.ThesetermsaredefinedseparatelyinSection Standards volume information, refer to the standard’s Document Summary page on
3. the ASTM website.
Copyright © ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959. United States
D6620 − 19
errors such as, determining the length of structures or fibers The DL may be used to make decisions about methods,
andwhether,basedonlength,theyshouldbecounted;counting laboratories,ormeasurementsystems.Forexample,toidentify
artifacts as fibers; determining the number of structures pro- trace levels of asbestos a measurement method with a suitably
trudingfromamatrix;andinterpretingaclusterasone,two,or small DL would be required.
more structures that should be counted only as zero or one
3.2.7 fiber, n—any of various discrete entities with essen-
structure. For purposes of developing the DL, assume that
tially parallel sides counted by a particular method that
background contamination sources have been reduced to their
specifies length, width, and aspect ratio. (This definition of
lowest achievable levels.
fiber, although adequate for purposes of this standard, Practice
3.2.3 blank, n—a filter that has not been used to collect D6620,isnottobeconfusedwiththemineralogicaltermfiber.)
3.2.7.1 Discussion—The definitions of “fiber” and “struc-
asbestos from a target population.
ture” are similar because the measurement method employed
3.2.3.1 Discussion—Blanks are used in this practice to
specifies the shape, length, width, and aspect ratio.
determine background levels (also referred to as contamina-
tion) for asbestos measurement methods. Blanks are analyzed
3.2.8 mean, n—the mean value of the number of structures
by the same method employed to analyze filters used to collect
in the population of air or dust sampled.
particulate from the target population.
3.2.8.1 Discussion—The mean in this definition is intended
tobethepopulationmean(equivalentlyexpectedvalue,orfirst
3.2.4 count, n—the number of fibers or structures identified
moment) of a statistical distribution. It is a theoretical param-
in a sample.
eter of the distribution that may be estimated by forming an
3.2.5 decision value, n—a numerical value used as a bound-
average of measurements.
ary in a statistical test to decide between the null hypothesis
3.2.9 power, n—the probability that a count exceeds the
and the alternative hypothesis.
decisionvaluegiventhatthecountisfromadistributionwhose
3.2.5.1 Discussion—In the present context, the decision
mean exceeds the background mean by a specified amount.
value is a structure count that defines the boundary between
3.2.9.1 Discussion—Power is the probability of selecting,
“below detection” (the null hypothesis) and “detection” (the
based on a statistical test, the alternative hypothesis when it is
alternativehypothesis).Ifastructurecountwerelargerthanthe
true. In the present context, this means the probability of
decision value, then one would conclude that detection has
making the correct decision to report a structure concentration
been achieved (that is, the sample is from a distribution other
forasamplethatwascollectedfromapopulationotherthanthe
thanthebackgrounddistribution).Ifthecountwerelessthanor
background population.The power of the statistical test equals
equal to the decision value, the result would be reported as
1 minus the type II error rate.
“below detection,” which means that the sample cannot be
differentiated from a sample that would have been collected 3.2.10 replicate, n—a second measurement is a replicate of
from the background distribution. the initial measurement if the second measurement is obtained
from an identical sample and under identical conditions as the
3.2.5.2 Discussion—The decision value is a threshold for
initial measurement.
measured values (or averages of measured values). If the
3.2.10.1 Discussion—“Identical,” as applied to sample, can
measured value (or average) exceeds this threshold, one can
conclude that asbestos has been detected in the sampled mean “same subsample preparation,” “separate preparation of
a distinct subsample,” or a distinct sample obtained from the
material. Measured values (or averages) below the decision
value may be reported but flagged to indicate that they fall same population as the initial sample. For this practice,
“identical” means distinct sample obtained from the same
below the decision value.Alternatively, they may be censored
(that is, actual measured values not reported) and simply population as the initial sample.
reported as non-detects. Then users of the data will be able to
3.2.11 sample,n—thetotalparticulatecollectedonafilteror
specifyinadvancewhethervaluesbelowthedecisionvalueare
dust collected on a wipe from a population of structures; also,
to be censored.
the portion a filter or wipe that is analyzed by microscopy
(also, see Terminology D1356).
3.2.6 detection limit—themeanofthestatisticaldistribution
of counts for a structure population that is sufficiently large so
3.2.12 sensitivity, n—the structure concentration corre-
ameasurementfromasamplefromthispopulationwouldhave
sponding to a count of one structure in the sample.
a high probability (for example, 0.95 or larger) of exceeding
3.2.13 statistical distribution, n—the set of probabilities of
the decision value that determines detection.
structure count outcomes (0, 1, 2, 3, …).
3.2.6.1 Discussion—The detection limit (DL) is determined
3.2.14 structure, n—any of various discrete entities counted
through the statistical hypothesis test described in 3.2.5.1. The
by a particular method that specifies shape, length, width, and
DL is the value of a parameter, the mean of a statistical
aspect ratio.
distribution of counts for a structure population, that is statis-
3.2.15 type I error, n—choosing, based on a statistical test,
ticallylargerthanthemeanofthestatisticaldistributionforthe
background population of structures.The DLis the smallest of torejectthenullhypothesiswhenthenullhypothesisis,infact,
true; a false positive outcome of a statistical test.
the means for which the power of the statistical test is
sufficiently large (for example, 0.95 or larger). The DL is not 3.2.15.1 Discussion—When testing a material that truly has
used to make decisions about individual measurements (for zero or background levels of asbestos, a type I error occurs
example, to decide whether they are detects or non-detects). when a measured value falls above the decision value. The
D6620 − 19
analyst, based on the data, would be led by the statistical test for the intended application. The most important use of the
to report a structure concentration where the result should be DL, therefore, takes place at the planning stage of a study,
reported as “below detection.” before samples are collected and analyzed.
3.2.16 type II error, n—failing to reject the null hypothesis
5. Descriptive Terms and Procedures
whenitis,infact,false;afalsenegativeoutcomeofastatistical
test. 5.1 Introduction:
3.2.16.1 Discussion—A type II error would occur if the 5.1.1 The DL is one of a number of characteristics used to
count for a sample does not exceed the decision value, but the describe the expected performance of a measurement method.
sample was, in fact, obtained from a population other than the The DL concept addresses certain potential measurement
interpretationerrors.Specifically,ameasurementisreportedas
background population. The analyst, based on the data, would
beledbythestatisticaltesttoreporta“belowdetection”result being “below the DL” if the measured level cannot be
distinguished from zero or from the randomly varying back-
where the result should be reported as a structure concentra-
tion. groundcontaminationlevel.Stateddifferently,theDLprovides
protection against a false positive finding. When a measured
3.2.17 type I error rate, n—the probability of a type I error
value is less than an appropriately specified decision value, the
(also referred to as the significance level,α-level,or p-value of
analystisinstructedtodisregardthemeasuredvalueandreport
the statistical test).
the result only as “below the DL.”
3.2.18 type II error rate, n—theprobabilityofatypeIIerror
5.1.2 TheDLconceptforasbestosmeasurements,whichare
(also referred to as the β-level of the statistical test).
based on microscopy, is simpler than the DL concept for
3.2.19 λ—lambda, the Greek letter used to represent the
measurement methods that depend, for example, on spectros-
population mean of a Poisson distribution.
copy. For asbestos, the measurement is derived from a direct
count of discrete structures using a microscope. For spectros-
3.2.20 λ —the population mean of the Poisson distribution
of background counts. copy methods, the measurement is indirect requiring a calibra-
tion curve, and is subject to interferences and unspecified
3.2.20.1 Discussion—λ is the population mean of the Pois-
son distribution under the null hypothesis in the statistical background signals that could be responsible for measurement
values that are false positives.
hypothesis testing problem that defines the DL.
5.1.3 The sources of false positives for asbestos counts are
3.2.21 λ —the population mean of the Poisson distribution
(i) analyst errors (for example, determining the length of
under the alternative hypothesis in the statistical hypothesis
structures or fibers and whether, based on length, they should
testing problem that defines the DL (DL = λ ).
be counted; counting artifacts as fibers; determining the num-
3.2.22 x —decision value for determining detection. If the
ber of structures protruding from a matrix; interpreting a
countinameasurementisnotgreaterthan x ,themeasurement
cluster as one, two, or more structures that should be counted
is reported as “below detection.”
only as zero or one), and (ii) contamination (for example,
3.2.23 X—Poisson distributed random variable used to de-
virgin filter contamination or contamination introduced during
note the number of structures (fibers) counted in a sample.
sample collection or sample preparation). Collectively, these
sources are referred to subsequently as “background.” For
3.2.24 A—the area of the filter inspected to obtain a struc-
purposes of developing the DL, assume that each background
ture count.
source has been reduced to its lowest achievable level.
3.2.25 P~X . x λ, A!—the Poisson probability of a struc-
?
ture count exceeding x structures (fibers) when the population 5.2 DL—General Discussion:
mean is equal to λ and an area, A, of the filter is inspected. 5.2.1 DLs often have been misspecified and misinterpreted
because the DL concept typically has not been defined with
4. Significance and Use
sufficient clarity for translation into operational terms;
4.1 The DLconcept addresses potential measurement inter-
however, the DLconcept and operational implementation have
pretationerrors.Itisusedtocontrolthelikelihoodofreporting
beenpresentedcorrectlyinthescientificliteraturebyanumber
apositivefindingofasbestoswhenthemeasuredasbestoslevel
of authors (1-4). These authors describe the DL as a theoreti-
cannot clearly be differentiated from the background contami-
calvalue,usuallythetruemeanconcentrationofasubstancein
nation level. Specifically, a measurement is reported as being
a sampled population. This true mean, the DL, must be large
“below the DL” if the measured level is not statistically
enough to ensure a large probability (for example, 0.95 or
different than the background level.
larger) that results for one or more samples of the substance
4.2 The DL, along with other measurement characteristics
such as bias and precision, is used when selecting a measure-
Forexample,thepurposeofthemeasurementsmightbetoassessdifferencesin
ment method for a particular application. The DL should be
the levels of a substance between two sources. If it were anticipated that the levels
established either at the method development stage or prior to
associated with each source are likely to be less than the DL of a particular
a specific application of the method. The method developer
measurement method, that method would not be appropriate for the intended
application.
subsequently would advertise the method as having a certain
Other characteristics are precision, bias, and for asbestos measurements,
DL.Ananalystplanningtocollectandanalyzesampleswould,
sensitivity.
if alternative measurement methods were available, want to
The boldface numbers in parentheses refer to a list of references at the end of
select a measurement method with a DL that was appropriate this standard.
D6620 − 19
from the population will be larger than the decision value, 5.2.4 Based on the structure outlined in 5.2.3.3 reporting
which leads to the conclusion that the samples were not from measurements subject to DL considerations would be imple-
the background population (that is the results were not below mented as follows:
detection). The DLshould be the smallest value that ensures a 5.2.4.1 Determine the decision value in the statistical test
high probability of detection. for determining if a measurement is large enough to conclude
that λ>λ is correct and determine the value of λ, say λ,to
5.2.2 Determining whether the mean concentration of a
0 1
achieve sufficient power. λ is the DL.
substance in a sample is consistent with the background
5.2.4.2 If the measured value exceeds the decision value,
concentration or is greater than the background concentration
reportthemeasuredvalue.Ifthemeasuredvalueislessthanor
is a statistical decision problem. Due to statistical variation,
equal to the decision value, report that the measurement is
replicate measurements of a sample or measurements from
“below the DL.”
replicate samples do not yield identical results; thus, a mea-
surement may exceed the true background mean level even if
6. Application to Air Samples
the sample were collected from the background distribution.
Differences in replicate results are characterized as statistical
6.1 The statistical hypothesis testing formulation described
variation. Values of replicate measurements are described by a
above and the Poisson distribution are employed to define and
probability distribution. The decision concerning whether or
calculate DLs for measurements of airborne structure concen-
not a measurement is consistent with the background concen-
trations.
tration fits the standard hypothesis testing framework in
6.2 For the DL concept to have meaning there must be a
statistics. The statistical testing problem, therefore, provides
background distribution of structure measurements. The back-
the necessary structure for determining a numerical value for
grounddistributionconsistsofsourcesofstructuresthatarenot
theDL,aswellasaruleforreportingmeasurementsas“below
themeasurementtargetsofinterestbutcannotbeeliminatedor
the DL.”
further reduced.
5.2.3 The DL is determined by formulating the statistical
6.2.1 The background distribution for airborne structure
testing problem as follows.
measurements is a combination of (i) analyst error and (ii)
5.2.3.1 Consider a statistical test, based on one
contamination (filter or laboratory).
measurement, of the null hypothesis that the true mean
6.2.1.1 Analyst errors are errors such as: determining the
concentration, λ, of substance in a sample is equal to the
lengthofstructuresorfibersandwhether,basedonlength,they
backgroundmean,λ ,versusthealternativehypothesisthatλis
shouldbecounted;countingartifactsasfibers;determiningthe
greater than λ . The typical decision rule leads to a choice of
number of structures protruding from a matrix; interpreting a
λ>λ overλ=λ if a standardized measurement is larger than a
0 0
cluster as one, two, or more structures that should be counted
specified decision value for the statistical test. The decision 10
only as zero or one.
value is chosen to control the type I error rate (also referred to
6.2.1.2 Filters may become contaminated from impurities
here as the false positive rate) of the statistical test. The false
that are inherent in their production or in the laboratory during
positive rate is the probability that a measurement will exceed
filter preparation for analysis. Filter contamination should be
thechosendecisionvalue,leadingtoacceptanceofλ>λ ,when 11
minimized by laboratory QA/QC procedures.
the true mean concentration is, in fact, λ .
6.2.2 All background sources should be reduced to their
5.2.3.2 The DL concept, although providing protection
lowest achievable levels. From an empirical perspective, it is
against false positives in measurement systems, also requires
neither practical nor necessary to quantify the background
consideration of probabilities associated with true positives.A
sources separately. The background level may be determined
high degree of confidence (that is, a high probability) is
by analyzing blanks without attempting to differentiate among
required that a decision in favor of λ>λ over λ=λ is correct.
0 0 sources.
In statistical hypothesis testing terminology, this probability is
6.3 Characterization of Sampling and Analysis to Measure
referred to as the “power of the statistical test.”
Airborne Asbestos—As an aid in the subsequent discussion, a
5.2.3.3 The power of a statistical test is the probability that
simplified characterization of air sampling and analysis for
a measurement exceeds the decision value (that is, the prob-
measuring airborne asbestos concentrations is used. Although
ability that the measurement leads to the choice, λ>λ ) when
this characterization of the measurement process may lack
the true mean concentration is a value larger than λ . The
important details from a microscopist’s perspective, it is
power of the test is an increasing function of the true mean, λ.
adequate for describing how to calculate a DL (refer to Test
The DL is the value of λ that makes the power sufficiently
Method D6281 and NIOSH 7400 (5) for additional details).
large. EPA definitions of the DL indicate that power, the
probability of a true positive result, should be 0.95 or greater.
Misclassification of a nonasbestos structure as an asbestos structure is not
treated as a false positive in the present discussion of DLs. For purposes of defining
In this statistical context, a standardized measurement is calculated as the aDL,consideronlythebackgroundsourcesdescribedaboveascontributingtofalse
measurement minus the background mean divided by the standard deviation of the positives.
background distribution. QA/QC procedures include: testing a sample of filters from a new supply
Thisprobabilityalsoisreferredtoasthesignificancelevelor p-valueofthetest before the new supply is used in the field; and diligently eliminating sources of
and typically is selected to be 0.05, but could be larger or smaller to reflect the asbestos contamination from the laboratory.
gravity of the consequences of a false positive. Background estimation methods are described in 6.4.2.
D6620 − 19
TABLE 1 Detection Limits for Different Background Means
6.3.1 Air sampling is accomplished by drawing air through
Measurement Unit Equals Number of Structures
a filter at a specified rate for a specified period of time.
(Nominal α = 0.05; Power = 0.95)
Airborne particles consisting of asbestos and other matter are
NOTE 1—“Structures” applies both to structures and fibers depending
deposited on the filter.After air sampling has been completed,
on the measurement protocol.
a section of the filter is prepared for inspection by microscopy.
Background Decision Actual Type I Detection Limit
A specified number of fields of view of known size (that is,
Mean (λ ) Value x Error Rate (α) (λ )
0 0 1
graticule fields for PCM and grid openings for TEM), are
0.00–0.05 0 0.000–0.048 3.00
randomlyselectedandinspectedmicroscopically.Theparticles
0.05–0.35 1 0.002–0.049 4.74
found in each field of view are classified as fibers for PCM or
0.35–0.81 2 0.006–0.049 6.30
0.81–1.36 3 0.010–0.049 7.75
asbestosstructuresforTEMandacountisrecorded.Thecount
1.36–1.97 4 0.013–0.050 9.15
obtained from the fields that were inspected is increased by an
1.97–2.61 5 0.016–0.050 10.51
appropriate factor to produce an estimated count for the total
filter. This estimate is divided by the volume of air collected
during sampling. The resulting measurement is interpreted as
an estimate of the asbestos concentration in the air, and is
6.4.1 Background Mean Known —Let X represent the
reported in units of fibers/cc of air (f/cc) for PCM or
number of structures counted in a sample based on inspection
structures/cc of air (str/cc) for TEM.
of a filter area equal to A (A= ∑FOV ·FOVA where FOV is
1 i i
6.3.2 The information described in 6.3.1 that is needed to
the number of grid openings with area FOVA). Let λ be the
i
address DLs can be summarized as a single number-
true average structure count. To establish the DL, set up a
measurement “sensitivity.” Sensitivity is a characteristic that
statistical test of the hypothesis H :λ=λ versus the alternative
0 0
applies to individual measurements. Sensitivity is defined as
H : λ>λ as described in 5.2.3. λ is the true mean count of
1 0 0
the structure concentration corresponding to a count of one
structures for the background distribution when an area, A,of
structure in the sample. Sensitivity, therefore, depends on air
thefilterisinspected. Thedecisionvalue,x ,isdefinedasthe
volume and the fraction (a proportion) of the filter that is
solution to P(X>x |λ=λ , A)=α (α is the significance level or
0 0
inspected. The fraction depends on the size of the effective
Type I error rate of the statistical test). The power of the
filter collection area, the size of the fields of view, and the
statistical test is calculated as P(X>x |λ=λ , A)=1–β. β is the
0 1
number of fields of view that are inspected.
Type II error rate of the test and 1–β is the value specified as
thepowerofthetest.TheDListhevalueofλ thatsatisfiesthe
Sensitivity ~S! 5 @~EFA!/~FOV*FOVA!#/V (1) 1
equation for the power of the test.
where:
6.4.1.1 The equations for calculating the DLare as follows:
EFA = the effective filter collection area in square milli-
Solve:
meters (mm );
P~X.x ?λ 5 λ , A! 5 α (3)
0 0
FOV = the number of fields of view;
FOVA = the average field of view area in mm ; and,
to determine the decision value, x .
V = air volume in cubic centimeters (cc).
Then solve:
6.3.3 Given any value as a requirement for sensitivity, the P~X.x ?λ 5 λ , A! 5 1 2 β (4)
1 1
air volume, field of view size, and number of fields of view
for λ , which is the DL.
may be varied to achieve the required value.
6.4.1.2 Calculate the probabilities indicated in Eq 4 and 5
using the following:
NOTE 1—Typical EFAs are 385 mm for a filter with a 25-mm diameter
and 855-mm for a filter with a 37-mm diameter.
t 2λ
P X.x ? λ, A 5 1 2 λ ·e /t! (5)
~ !
(
6.3.4 From the definition of sensitivity, it follows that the
where the index t in the sum takes the values 0, 1, 2, ., x.
structure concentration measurement for a sample is the
6.4.1.3 NumericalExamplesofDLsforAirborneAsbestos—
number of structures counted multiplied by sensitivity:
Based on assumptions about the true value of the underlying
str/cc 5 ~# structures!*S (2)
background mean, decision values and DLs have been deter-
mined and are recorded in Tables 1-4. The examples in Tables
6.4 Based on the usual assumption that the structure count
1and2havebeendevelopedforastatisticaltestofλ=λ versus
from an air sample is described by the Poisson probability
λ>λ with the nominal significance level of α = 0.05 and
distribution, equations were developed for calculating DLs.
nominal powers equal to 0.95 and 0.99, respectively. Because
TheDLisstatedasameanstructurecount.Themeanstructure
ofthediscretenatureofstructure
...

This document is not an ASTM standard and is intended only to provide the user of an ASTM standard an indication of what changes have been made to the previous version. Because
it may not be technically possible to adequately depict all changes accurately, ASTM recommends that users consult prior editions as appropriate. In all cases only the current version
of the standard as published by ASTM is to be considered the official document.
Designation: D6620 − 06 (Reapproved 2010) D6620 − 19
Standard Practice for
Asbestos Detection Limit Based on Counts
This standard is issued under the fixed designation D6620; the number immediately following the designation indicates the year of
original adoption or, in the case of revision, the year of last revision. A number in parentheses indicates the year of last reapproval. A
superscript epsilon (´) indicates an editorial change since the last revision or reapproval.
1. Scope
2 3
1.1 This practice presents the procedure for determining the detection limit (DL) for measurements of fibers or structures
using microscopy methods.
1.2 This practice applies to samples of air that are analyzed either by phase contrast microscopy (PCM) or transmission electron
microscopy (TEM), and samples of dust that are analyzed by TEM.
1.3 The microscopy methods entail counting asbestos structures and reporting the results as structures per cubic centimeter of
air (str/cc) or fibers per cubic centimeter of air (f/cc) for air samples and structures per square centimeter of surface area (str/cm )
for dust samples.
1.4 The values stated in SI units are to be regarded as standard. No other units of measurement are included in this standard.
1.5 This standard does not purport to address all of the safety concerns, if any, associated with its use. It is the responsibility
of the user of this standard to establish appropriate safety safety, health, and healthenvironmental practices and determine the
applicability of regulatory limitations prior to use.
1.6 This international standard was developed in accordance with internationally recognized principles on standardization
established in the Decision on Principles for the Development of International Standards, Guides and Recommendations issued
by the World Trade Organization Technical Barriers to Trade (TBT) Committee.
2. Referenced Documents
2.1 ASTM Standards:
D1356 Terminology Relating to Sampling and Analysis of Atmospheres
D5755 Test Method for Microvacuum Sampling and Indirect Analysis of Dust by Transmission Electron Microscopy for
Asbestos Structure Number Surface Loading
D6281 Test Method for Airborne Asbestos Concentration in Ambient and Indoor Atmospheres as Determined by Transmission
Electron Microscopy Direct Transfer (TEM)
D6480 Test Method for Wipe Sampling of Surfaces, Indirect Preparation, and Analysis for Asbestos Structure Number Surface
Loading by Transmission Electron Microscopy
E456D7712 Terminology Relating to Quality and Statisticsfor Sampling and Analysis of Asbestos
3. Terminology
3.1 Definitions:
3.1.1 For terms not defined in this practice, see Terminologies D1356 and D7712.
3.2 Definitions of Terms Specific to This Standard:
3.2.1 average, n—the sum of a set of measurements (counts) divided by the number of measurements in the set.
This practice is under the jurisdiction of ASTM Committee D22 on Air Quality and is the direct responsibility of Subcommittee D22.07 on Sampling and Analysis of
AsbestosSampling, Analysis, Management of Asbestos, and Other Microscopic Particles.
Current edition approved Oct. 1, 2010Jan. 1, 2019. Published November 2010January 2019. Originally approved in 2000. Last previous edition approved 20062010 as
D6620 – 06.D6620 – 06 (2010). DOI: 10.1520/D6620-06R10.10.1520/D6620-19.
The DL also is referred to in the scientific literature as Limitlimit of Detection (LOD), Method Detection Limitdetection (LOD), method detection limit (MDL), and other
similar descriptive names.
For purposes of general exposition, the term “structures” will be used in place of “fibers or structures.” In the examples in Section 8, the specific term, “fiber” or
“structure,” is used where appropriate. These terms are defined separately in Section 3.
For referenced ASTM standards, visit the ASTM website, www.astm.org, or contact ASTM Customer Service at [e-mail]. For Annual Book of ASTM Standards
volume information, refer to the standard’s Document Summary page on the ASTM website.
Copyright © ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959. United States
D6620 − 19
3.2.1.1 Discussion—
The average is distinguished from the mean. The average is calculated from data and serves as an estimate of the mean. The mean
(also referred to as the population mean, expected value, or first moment) is a parameter of the underlying statistical distribution
of counts.
3.2.2 background, n—a statistical distribution of structures introduced by (i) analyst counting errors and (ii) contamination on
an unused filter or contamination as a consequence of the sample collection and sample preparation steps.
3.2.2.1 Discussion—
This definition of background is specific to this practice. The only counting errors considered in this definition of background are
errors that result in an over-count (that is, false positives). Analyst counting errors are errors such as, determining the length of
structures or fibers and whether, based on length, they should be counted; counting artifacts as fibers; determining the number of
structures protruding from a matrix; and interpreting a cluster as one, two, or more structures that should be counted only as zero
or one structure. For purposes of developing the DL, assume that background contamination sources have been reduced to their
lowest achievable levels.
3.2.3 blank, n—a filter that has not been used to collect asbestos from thea target environment.population.
3.2.3.1 Discussion—
Blanks are used in this practice to determine the degree of asbestos contamination that is reflected in asbestos measurements.
Contamination may be on the virgin filter or introduced in handling the filter in the field or when preparing it for inspection with
a microscope. The data required to determine the degree of contamination consists, therefore, of measurements of field blanks that
have experienced the full preparation process.background levels (also referred to as contamination) for asbestos measurement
methods. Blanks are analyzed by the same method employed to analyze filters used to collect particulate from the target population.
3.2.4 count, n—the number of fibers or structures identified in a sample.
3.2.5 decision value, n—a numerical value used as a boundary in a statistical test to decide between the null hypothesis and the
alternative hypothesis.
3.2.5.1 Discussion—
In the present context, the decision value is a structure count that defines the boundary between “below detection” (the null
hypothesis) and “detection” (the alternative hypothesis). If a structure count were larger than the decision value, then one would
conclude that detection has been achieved (that is, the sample is from a distribution other than the background distribution). If the
count were less than or equal to the decision value, the result would be reported as “below detection,” which means that the sample
cannot be differentiated from a sample that would have been collected from the background distribution.
3.2.5.2 Discussion—
The decision value is a threshold for measured values (or averages of measured values). If the measured value (or average) exceeds
this threshold, one can conclude that asbestos has been detected in the sampled material. Measured values (or averages) below the
decision value may be reported but flagged to indicate that they fall below the decision value. Alternatively, they may be censored
(that is, actual measured values not reported) and simply reported as non-detects. Then users of the data will be able to specify
in advance whether values below the decision value are to be censored.
3.2.6 detection limit—the mean of the statistical distribution of counts for a structure count population that is sufficiently large
so a measurement from a sample from this population would have a high probability (for example, 0.95 or larger) of exceeding
the decision value that determines detection.
3.2.6.1 Discussion—
The detection limit (DL) is determined through the statistical hypothesis test described in 3.2.5.1. The DL is the value of a
parameter, the true mean of a structure count population in the statistical hypothesis testing problem, that underlies the DL concept.
Specifically, it is the true statistical distribution of counts for a structure population, that is statistically larger than the mean of the
alternative hypothesis that ensures a sufficiently high power for the statistical test that determines detection.statistical distribution
for the background population of structures. The DL is the smallest of the means for which the power of the statistical test is
D6620 − 19
sufficiently large (for example, 0.95 or larger). The DL is not used to make decisions about individual measurements (for example,
to decide whether they are detects or non-detects). The DL may be used to make decisions about methods, laboratories, or
measurement systems. For example, to identify trace levels of asbestos a measurement method with a suitably small DL would
be required.
3.2.7 fiber, n—any of various discrete entities with essentially parallel sides counted by a particular method that specifies length,
width, and aspect ratio. (This definition of fiber, although adequate for purposes of this standard, Practice D6620, is not to be
confused with the mineralogical term fiber.)
3.2.7.1 Discussion—
The definitions of “fiber” and “structure” are similar because the measurement method employed specifies the shape, length, width,
and aspect ratio.
3.2.8 mean, n—the mean value of the number of structures in the population of air or dust sampled.
3.2.8.1 Discussion—
The mean in this definition is intended to be the population mean, mean (equivalently expected value, or first momentmoment)
of a statistical distribution. It is a theoretical parameter of the distribution that may be estimated by forming an average of
measurements (refer to Terminology measurements. E456 for definition of population).
3.2.9 power, n—the probability that a count exceeds the decision value for a sample that was obtained from a population other
than the background population.given that the count is from a distribution whose mean exceeds the background mean by a specified
amount.
3.2.9.1 Discussion—
Power is the probability of selecting, based on a statistical test, the alternative hypothesis when it is true. In the present context,
this means the probability of making the correct decision to report a structure concentration for a sample that was collected from
a population other than the background population. The power of the statistical test equals 1 minus the type II error rate.
3.2.10 replicate, n—a second measurement is a replicate of the initial measurement if the second measurement is obtained from
an identical sample and under identical conditions as the initial measurement.
3.2.10.1 Discussion—
“Identical,” as applied to sample, can mean“ samemean “same subsample preparation,” “separate preparation of a distinct
subsample,” or a distinct sample obtained from the same population as the initial sample. For this practice, “identical” means
distinct sample obtained from the same population as the initial sample.
3.2.11 sample, n—the segment of the filter that is inspected, and thereby, embodies the air or dust that was collected and the
subset of structures that were captured on the portion of the filter subjected to microscopic inspection total particulate collected
on a filter or dust collected on a wipe from a population of structures; also, the portion a filter or wipe that is analyzed by
microscopy (also, see Terminology D1356).
3.2.12 sensitivity, n—the structure concentration corresponding to a count of one structure in the sample.
3.2.13 statistical distribution, n—the set of probabilities of structure count outcomes (0, 1, 2, 3, …).
3.2.14 structure, n—any of various discrete entities counted by a particular method that specifies shape, length, width, and
aspect ratio.
3.2.15 type I error, n—choosing, based on a statistical test, the alternative hypothesis over the to reject the null hypothesis when
the null hypothesis is, in fact, true; a false positive outcome of a statistical test.
3.2.15.1 Discussion—
A When testing a material that truly has zero or background levels of asbestos, a type I error would occur if the count for a sample
exceeded the decision value, but the sample was, in fact, obtained from the background population. The analyst erroneously occurs
when a measured value falls above the decision value. The analyst, based on the data, would be led by the statistical test to report
a structure concentration (that is, choose the alternative hypothesis of the statistical test), where the result should be reported as
“below the detection limit” (that is, the null hypothesis of the statistical test is true).detection.”
D6620 − 19
3.2.16 type II error, n—choosing, based on a statistical test, failing to reject the null hypothesis over the alternative hypothesis
when the alternative hypothesis when it is, in fact, true;false; a false negative outcome of a statistical test.
3.2.16.1 Discussion—
A type II error would occur if the count for a sample does not exceed the decision value, but the sample was, in fact, obtained from
a population other than the background population. The analyst would erroneously analyst, based on the data, would be led by the
statistical test to report a “below the detection limit” result (that is, choose the null hypothesis of the statistical test), detection”
result where the result should be reported as a structure concentration (that is, the alternative hypothesis of the statistical test is
true).concentration.
3.2.17 type I error rate, n—the probability of a type I error (also referred to as the significance level, α-level, or p-value of the
statistical test).
3.2.18 type II error rate, n—the probability of a type II error (also referred to as the β-level of the statistical test).
3.2.19 λ—lambda, the Greek letter used to represent the population mean of a Poisson distribution.
3.2.20 λ —the population mean of the Poisson distribution of background counts.
3.2.20.1 Discussion—
λ is the population mean of the Poisson distribution under the null hypothesis in the statistical hypothesis testing problem that
defines the DL.
3.2.21 λ —the population mean of the Poisson distribution under the alternative hypothesis in the statistical hypothesis testing
problem that defines the DL (DL = λ ).
3.2.22 x —decision value for determining detection. If the count in a measurement is not greater than x , the measurement is
0 0
reported as “below detection.”
3.2.23 X—Poisson distributed random variable used to denote the number of structures (fibers) counted in a sample.
3.2.24 A—the area of the filter inspected to obtain a structure count.
3.2.25 P X . x λ, A P(X>x/λ, A)——the Poisson probability of a structure count exceeding x structures (fibers) when the
~ !
?
population mean is equal to λ and an area, A, of the filter is inspected.
4. Significance and Use
4.1 The DL concept addresses potential measurement interpretation errors. It is used to control the likelihood of reporting a
positive finding of asbestos when the measured asbestos level cannot clearly be differentiated from the background contamination
level. Specifically, a measurement is reported as being “below the DL” if the measured level is not statistically different than the
background level.
4.2 The DL, along with other measurement characteristics such as bias and precision, is used when selecting a measurement
method for a particular application. The DL should be established either at the method development stage or prior to a specific
application of the method. The method developer subsequently would advertise the method as having a certain DL. An analyst
planning to collect and analyze samples would, if alternative measurement methods were available, want to select a measurement
method with a DL that was appropriate for the intended application. The most important use of the DL, therefore, takes place at
the planning stage of a study, before samples are collected and analyzed.
5. Descriptive Terms and Procedures
5.1 Introduction:
5.1.1 The DL is one of a number of characteristics used to describe the expected performance of a measurement method. The
DL concept addresses certain potential measurement interpretation errors. Specifically, a measurement is reported as being “below
the DL” if the measured level cannot be distinguished from zero or from the randomly varying background contamination level.
Stated differently, the DL provides protection against a false positive finding. When a measured value is less than an appropriately
specified decision value, the analyst is instructed to disregard the measured value and report the result only as “below the DL.”
5.1.2 The DL concept for asbestos measurements, which are based on microscopy, is simpler than the DL concept for
measurement methods that depend, for example, on spectroscopy. For asbestos, the measurement is derived from a direct count
For example, the purpose of the measurements might be to assess differences in the levels of a substance between two sources. If it were anticipated that the levels
associated with each source are likely to be less than the DL of a particular measurement method, that method would not be appropriate for the intended application.
Other characteristics are precision, bias, and for asbestos measurements, sensitivity.
D6620 − 19
of discrete structures using a microscope. For spectroscopy methods, the measurement is indirect requiring a calibration curve, and
is subject to interferences and unspecified background signals that could be responsible for measurement values that are false
positives.
5.1.3 The sources of false positives for asbestos counts are (i) analyst errors (for example, determining the length of structures
or fibers and whether, based on length, they should be counted; counting artifacts as fibers; determining the number of structures
protruding from a matrix; interpreting a cluster as one, two, or more structures that should be counted only as zero or one), and
(ii) contamination (for example, virgin filter contamination or contamination introduced during sample collection or sample
preparation). Collectively, these sources are referred to subsequently as “background.” For purposes of developing the DL, assume
that each background source has been reduced to its lowest achievable level.
5.2 DL—General Discussion:
5.2.1 DLs often have been misspecified and misinterpreted because the DL concept typically has not been defined with sufficient
clarity for translation into operational terms; however, the DL concept and operational implementation have been presented
correctly in the scientific literature by a number of authors.authors (1-4). These authors describe the DL as a theoretical value,
specificallyusually the true mean concentration of a substance in a sampled medium.population. This true mean, the DL, must be
large enough to ensure a highlarge probability (for example, 0.95 or larger) of concluding based on that results for one or more
measurements from a sample of the medium samples of the substance from the population will be larger than the decision
value,that the true concentration in the medium is, in fact, greater than zero or greater than an appropriately defined background
level. The DL, therefore, is a parameter in the statistical decision that determines whether the concentration of a substance in a
sample is consistent with the background level, which may be zero, or is greater than the background level. which leads to the
conclusion that the samples were not from the background population (that is the results were not below detection). The DL should
be the smallest value that ensures a high probability of detection.
5.2.2 Determining whether the mean concentration of a substance in a sample is consistent with the background concentration
or is greater than the background concentration is a statistical decision problem. Due to statistical variation, replicate measurements
of a sample or measurements from replicate samples do not yield identical results; thus, a measurement may exceed the true
background mean level even if the sample were collected from the background distribution. Differences in replicate results are
characterized as statistical variation. Values of replicate measurements are described by a probability distribution. The decision
concerning whether or not a measurement is consistent with the background concentration fits the standard hypothesis testing
framework in statistics. The statistical testing problem, therefore, provides the necessary structure for determining a numerical
value for the DL, as well as a rule for reporting measurements as “below the DL.”
5.2.3 The DL is determined by formulating the statistical testing problem as follows.
5.2.3.1 Consider a statistical test, based on one measurement, of the null hypothesis that the true mean concentration, λ, of
substance in a sample is equal to the background mean, λ , versus the alternative hypothesis that λ is greater than λ . The typical
0 0
decision rule leads to a choice of λ>λ over λ=λ if a standardized measurement is larger than a specified decision value for the
0 0
statistical test. The decision value is chosen to control the Typetype I error rate (also referred to here as the false positive rate) of
the statistical test. The false positive rate is the probability that a measurement will exceed the chosen decision value, leading to
acceptance of λ>λ , when the true mean concentration is, in fact, λ .
0 0
5.2.3.2 The DL concept, although providing protection against false positives in measurement systems, also requires
consideration of probabilities associated with true positives. A high degree of confidence (that is, a high probability) is required
that a decision in favor of λ>λ over λ=λ is correct. In statistical hypothesis testing terminology, this probability is referred to as
0 0
the “power of the statistical test.”
5.2.3.3 The power of a statistical test is the probability that a measurement exceeds the decision value (that is, the probability
that the measurement leads to the choice, λ>λ ) when the true mean concentration is a value larger than λ . The power of the test
0 0
is an increasing function of the true mean, λ. The DL is the value of λ that makes the power sufficiently large. EPA definitions of
the DL indicate that power, the probability of a true positive result, should be 0.95 or greater.
5.2.4 Based on the structure outlined in 5.2.3.3 reporting measurements subject to DL considerations would be implemented as
follows:
5.2.4.1 Determine the decision value in the statistical test for determining if a measurement is large enough to conclude that
λ>λ is correct and determine the value of λ, say λ , to achieve sufficient power. λ is the DL.
0 1 1
5.2.4.2 If the measured value exceeds the decision value, report the measured value. If the measured value is less than or equal
to the decision value, report that the measurement is “below the DL.”
Clayton, C. A., Hines, J. W., and Elkins, P. D., “Detection Limits with Specified Assurance Probabilities,” Analytical Chem. 59, 1987, 2506–2514; Currie, L. A., “Limits
of Qualitative Detection and Quantitative Determination: Application to Radiochemistry,” Analytical Chem., Vol 40, 1968, 586–593; Currie, L. A., “Lower Limit of Detection:
Definition and Elaboration of a Proposed Position for Radiological Effluent and Environmental Measurements,” National Bureau of Standards Report, 1984 ; Fowler, D. P.,
“Definition of Lower Limits for Airborne Particle Analyses Based on Counts and Recommended Reporting Conventions,” Ann. Occup Hyg., Vol 41 Supplement 1, 1997,
203–209. The boldface numbers in parentheses refer to a list of references at the end of this standard.
In this statistical context, a standardized measurement is calculated as the measurement minus the background mean divided by the standard deviation of the background
distribution.
This probability also is referred to as the significance level or p-value of the test and typically is selected to be 0.05, but could be larger or smaller to reflect the gravity
of the consequences of a false positive.
D6620 − 19
6. Application to Air Samples
6.1 The statistical hypothesis testing formulation described above and the Poisson distribution are employed to define and
calculate DLs for measurements of airborne structure concentrations.
6.2 For the DL concept to have meaning there must be a background distribution of structure measurements. The background
distribution consists of sources of structures that are not the measurement targets of interest but cannot be eliminated or further
reduced.
6.2.1 The background distribution for airborne structure measurements is a combination of (i) analyst error and (ii)
contamination (filter or laboratory).
6.2.1.1 Analyst errors are errors such as: determining the length of structures or fibers and whether, based on length, they should
be counted; counting artifacts as fibers; determining the number of structures protruding from a matrix; interpreting a cluster as
one, two, or more structures that should be counted only as zero or one.
6.2.1.2 Filters may become contaminated from impurities that are inherent in their production or in the laboratory during filter
preparation for analysis in the laboratory. analysis. Filter contamination should be minimized by laboratory QA/QC procedures.
6.2.2 All background sources should be reduced to their lowest achievable levels. From an empirical perspective, it is neither
practical nor necessary to quantify the background sources separately. The background level may be determined by analyzing
blanks without attempting to differentiate among sources.
6.3 Characterization of Sampling and Analysis to Measure Airborne Asbestos—As an aid in the subsequent discussion, a
simplified characterization of air sampling and analysis for measuring airborne asbestos concentrations is used. Although this
characterization of the measurement process may lack important details from a microscopist’s perspective, it is adequate for
describing how to calculate a DL (refer to Test Method D6281 and NIOSH 7400 (5) for additional details).
6.3.1 Air sampling is accomplished by drawing air through a filter at a specified rate for a specified period of time. Airborne
particles consisting of asbestos and other matter are deposited on the filter. After air sampling has been completed, a section of the
filter is prepared for inspection by microscopy. A specified number of fields of view of known size (that is, graticule fields for PCM
and grid openings for TEM), are randomly selected and inspected microscopically. The particles found in each field of view are
classified as fibers for PCM or asbestos structures for TEM and a count is recorded. The count obtained from the fields that were
inspected is increased by an appropriate factor to produce an estimated count for the total filter. This estimate is divided by the
volume of air collected during sampling. The resulting measurement is interpreted as an estimate of the asbestos concentration in
the air, and is reported in units of fibers/cc of air (f/cc) for PCM or structures/cc of air (str/cc) for TEM.
6.3.2 The information described in 6.3.1 that is needed to address DLs can be summarized as a single number-measurement
“sensitivity.” Sensitivity is a characteristic that applies to individual measurements. Sensitivity is defined as the structure
concentration corresponding to a count of one structure in the sample. Sensitivity, therefore, depends on air volume and the fraction
(a proportion) of the filter that is inspected. The fraction depends on the size of the effective filter collection area, the size of the
fields of view, and the number of fields of view that are inspected.
Sensitivity ~S! 5 @~EFA!/~FOV*FOVA!#/V (1)
where:
EFA = the effective filter collection area in square millimeters (mm );
FOV = the number of fields of view;
FOVA = the average field of view area in mm ; and,
V = air volume in cubic centimeters (cc).
Sensitivity ~S! 5 @~EFA!/~FOV*FOVA!#/V (1)
where:
EFA = the effective filter collection area in square millimeters (mm );
FOV = the number of fields of view;
FOVA = the average field of view area in mm ; and,
V = air volume in cubic centimeters (cc).
6.3.3 Given any value as a requirement for sensitivity, the air volume, field of view size, and number of fields of view may be
varied to achieve the required value.
2 2
NOTE 1—Typical EFAs are 385 mm for a filter with a 25-mm diameter and 855-mm for a filter with a 37-mm diameter.
Misclassification of a nonasbestos structure as an asbestos structure is not treated as a false positive in the present discussion of DLs. For purposes of defining a DL,
consider only the background sources described above as contributing to false positives.
QA/QC procedures include: testing a sample of filters from a new supply before the new supply is used in the field; and diligently eliminating sources of asbestos
contamination from the laboratory.
Background estimation methods are described in 6.4.2.
The sensitivity concept also may be applied to averages of multiple measurements in situations where “a measurement” always means the average of a specified number
of independent replicate measurements. This application of sensitivity is not discussed here.
D6620 − 19
TABLE 1 Detection Limits for Different Background Means
Measurement Unit Equals Number of Structures
(Nominal α = 0.05; Power = 0.95)
NOTE 1—“Structures” applies both to structures and fibers depending
on the measurement protocol.
Background Decision Actual Type I Detection Limit
Mean (λ ) Value x Error Rate (α) (λ )
0 0 1
0.00–0.05 0 0.000–0.048 3.00
0.05–0.35 1 0.002–0.049 4.74
0.35–0.81 2 0.006–0.049 6.30
0.81–1.36 3 0.010–0.049 7.75
1.36–1.97 4 0.013–0.050 9.15
1.97–2.61 5 0.016–0.050 10.51
6.3.4 From the definition of sensitivity, it follows that the structure concentration measurement for a sample is the number of
structures counted multiplied by sensitivity:
str/cc 5 # structures *S (2)
~ !
str/cc 5 # structures *S (2)
~ !
6.4 Based on the usual assumption that the structure count from an air sample is described by the Poisson probability
distribution, equations were developed for calculating DLs. The DL is stated as a mean structure count. The mean structure count
may subsequently be translated to concentration units (str/cc) through multiplication by the sensitivity of the measurement as
shown in Eq 23.
6.4.1 Background Mean Known —Let X represent the number of structures counted in a sample based on inspection of a filter
area equal to A (A = ∑FOV ·FOVA where FOV is the number of grid openings with area FOVA ). Let λ be the true average
1 i i i
structure count. To establish the DL, set up a statistical test of the hypothesis H : λ=λ versus the alternative H : λ>λ as described
0 0 1 0
in 5.2.3. λ is the true mean count of structures for the background distribution when an area, A, of the filter is inspected. The
decision value, x , is defined as the solution to P(X>x |λ=λ , A)=α (α is the significance level or Type I error rate of the statistical
0 0 0
test). The power of the statistical test is calculated as P(X>x |λ=λ , A)=1–β. β is the Type II error rate of the test and 1–β is the
0 1
value specified as the power of the test. The DL is the value of λ that satisfies the equation for the power of the test.
6.4.1.1 The equations for calculating the DL are as follows:
Solve:
Solve
P~X.x !?λ5 ~λ , A! 5 α (3)
0 0
P X.x ?λ5 λ , A 5 α (3)
~ !
0 0
to determine the decision value, x .
...