United States Office of Research and EPA/600/R-01/049
Environmental Protection Development August 2001
Agency Washington, D.C. 20460
v>EPA Environmental Technology
Verification Report
PCB Detection Technology
Dexsil Corporation
L2000DX Analyzer
onvl
Oak Ridge National Laboratory
-------
THE ENVIRONMENTAL TECHNOLOGY VERIFICATION
PROGRAM
ET
AM^
tr
vvEPA
ornl
Oak Ridge National Laboratory
Verification Statement
TECHNOLOGY TYPE:
APPLICATION:
TECHNOLOGY NAME:
COMPANY:
ADDRESS:
WEB SITE:
E-MAIL:
ION SPECIFIC ELECTRODE
MEASUREMENT OF PCBs IN TRANSFORMER OIL
L2000DX Analyzer
Dexsil Corporation
One Hamden Park Drive
Hamden, CT 06517
www.dexsil.com
tblynn@dexsil.com
PHONE: (203) 288-3509
FAX: (203) 248-6523
The U.S. Environmental Protection Agency (EPA) has created the Environmental Technology
Verification Program (ETV) to facilitate the deployment of innovative or improved environmental
technologies through performance verification and dissemination of information. The goal of the ETV
Program is to further environmental protection by substantially accelerating the acceptance and use of
improved and cost-effective technologies. ETV seeks to achieve this goal by providing high-quality,
peer-reviewed data on technology performance to those involved in the design, distribution, financing,
permitting, purchase, and use of environmental technologies.
ETV works in partnership with recognized standards and testing organizations and stakeholder groups
consisting of regulators, buyers, and vendor organizations, with the full participation of individual
technology developers. The program evaluates the performance of innovative technologies by developing
test plans that are responsive to the needs of stakeholders, conducting field or laboratory tests (as
appropriate), collecting and analyzing data, and preparing peer-reviewed reports. All evaluations are
conducted in accordance with rigorous quality assurance protocols to ensure that data of known and
adequate quality are generated and that the results are defensible.
Oak Ridge National Laboratory (ORNL) is one of the verification organizations operating under the Site
Characterization and Monitoring Technologies (SCMT) program. SCMT, which is administered by
EPA's National Exposure Research Laboratory (NERL), is one of six technology areas under ETV. In
this verification test, ORNL evaluated the performance of polychlorinated biphenyl (PCB) detection
technologies. This verification statement provides a summary of the test results for Dexsil's L2000DX
instrument.
EPA-VS-SCM-46
The accompanying notice is an integral part of this verification statement.
August 2001
-------
VERIFICATION TEST DESCRIPTION
This verification test was designed to evaluate technologies that detect and measure PCBs in transformer
oil. The test was conducted at ORNL in Oak Pudge, Tennessee, from August 21 through August 23,
2000. Spiked samples of known concentration were used to assess the accuracy of the technology.
Environmentally contaminated oil samples, collected from ORNL transformers and ranging in
concentration from 0 to approximately 300 parts per million (ppm), were used to assess several
performance characteristics. Tests were conducted outdoors, with naturally fluctuating temperatures and
relative humidity conditions. The results of the oil analyses conducted by the technology were compared
with results from analyses of homogeneous replicate samples conducted by conventional EPA
methodology in an approved reference laboratory. Details of the test, including a data summary and
discussion of results, may be found in the report entitled Environmental Technology Verification Report:
PCB Detection Technology—Dexsil Corporation, L2000DX, EPA/600/R-01/049.
TECHNOLOGY DESCRIPTION
The L2000DX Analyzer (dimensions: 9 x 9.5 x 4.25 in.) is a field-portable ion-specific electrode
instrument, weighing approximately 5 Ib 12 oz, designed to quantify concentrations of PCBs, chlorinated
solvents, and pesticides in soils, water, transformer oils, and surface wipes. The L2000DX can be
operated in the field powered by a rechargeable 8-V gel cell, or in the laboratory using 120-V AC power.
To prepare a sample for analysis, 5 mL of the oil is collected in a polyethylene reaction tube. Two glass
ampules contained in the reaction tube are broken to introduce metallic sodium to the oil. The mixture is
then shaken for 10s and allowed to react for a total of 1 min. The sodium strips the covalently bonded
chlorine atoms off the PCB molecule. An aqueous extraction solution is added to the reaction tube to
adjust the pH, destroy the excess sodium, and extract and isolate the newly formed chloride ions in a
buffered aqueous solution. The aqueous layer is decanted, filtered, and collected in an analysis vial. The
ion-specific electrode is put into this aqueous solution to measure the millivolt potential. The potential is
then converted to the equivalent PCB concentration. The lowest concentration reported by the L2000DX
is typically 3 ppm. The performance of a previous version of this instrument (the L2000 PCB/Chloride
Analyzer) was verified by ETV for soil and solvent extracts in 1998.
VERIFICATION OF PERFORMANCE
The following performance characteristics of the L2000DX were observed:
Precision: Precision—based on the mean percent relative standard deviation—was 11%.
Accuracy: Accuracy was assessed using the nominal concentrations of the spiked oils. The mean percent
recovery value for the spiked samples was 112%. The L2000DX results were unbiased for both single -
Aroclor and multi-Aroclor mixtures.
False positive/false negative results: Of the 20 blank samples, Dexsil reported PCBs in 5 samples (25%
false positives). In addition, false positive and false negative results were determined by comparing the
L2000DX results with the reference laboratory results for the environmental and spiked samples. One of
the results was reported as a false positive (13% of total), and none were false negatives.
Completeness: The L2000DX generated results for all 152 oil samples, for a completeness of 100%.
EPA-VS-SCM-46 The accompanying notice is an integral part of this verification statement. August 2001
-------
Comparability: A one-to-one sample comparison of the L2000DX results and the reference laboratory
results was performed for all samples (spiked and environmental) that were reported as detections. The
correlation coefficient (r) for the comparison of the entire oil data set was 0.92 [slope (m) = 0.89]. The
reference laboratory's method was biased high for samples that contained mixtures of overlapping
Aroclors (such as a mixture of 1254 and 1260). If the samples containing mixtures of Aroclors are
removed from the data set, the r value is 0.95 and the m value is 1.1.
Sample Throughput: Operating in the field, the Dexsil team accomplished a sample throughput rate of
approximately eight samples per hour for the oil analyses. One operator prepared the samples, while
the other performed the analyses. The instrument can be operated by a single trained analyst.
Overall Evaluation: The overall performance was characterized as unbiased and precise. The
verification team found that the L2000DX was relatively simple for the trained analyst to operate in the
field, requiring less than an hour for initial setup. As with any technology selection, the user must
determine if this technology is appropriate for the application and the project data quality objectives.
For more information on this and other verified technologies, visit the ETV web site at
http://www.epa.gov/etv.
Gary J. Foley, Ph.D.
Director
National Exposure Research Laboratory
Office of Research and Development
W. Frank Harris, Ph.D.
Associate Laboratory Director
Biological and Environmental Sciences
Oak Ridge National Laboratory
NOTICE: EPA verifications are based on evaluations of technology performance under specific, predetermined criteria
and appropriate quality assurance procedures. EPA and ORNL make no expressed or implied warranties as to the
performance of the technology and do not certify that a technology will always operate as verified. The end user is solely
responsible for complying with any and all applicable federal, state, and local requirements. Mention of commercial
product names does not imply endorsement or recommendation.
EPA-VS-SCM-46
The accompanying notice is an integral part of this verification statement.
August 2001
-------
EPA/600/R-01/049
August 2001
Environmental Technology
Verification Report
PCB Detection Technology
Dexsil Corporation
L2000DX Analyzer
By
Amy B. Dindal
Charles K. Bayne, Ph.D.
Roger A. Jenkins, Ph.D.
Oak Ridge National Laboratory
Oak Ridge, Tennessee 37831-6120
Eric N. Koglin
U.S. Environmental Protection Agency
Environmental Sciences Division
National Exposure Research Laboratory
Las Vegas, Nevada 89193-3478
-------
Notice
The U.S. Environmental Protection Agency (EPA), through its Office of Research and Development (ORD),
funded and managed, through Interagency Agreement No. DW89937854 with Oak Ridge National
Laboratory, the verification effort described herein. This report has been peer and administratively reviewed
and has been approved for publication as an EPA document. Mention of trade names or commercial products
does not constitute endorsement or recommendation for use of a specific product.
-------
Table of Contents
List of Figures v
List of Tables vii
Acknowledgments ix
Abbreviations and Acronyms xi
1. INTRODUCTION 1
2. TECHNOLOGY DESCRIPTION 2
General Technology Description 2
Oil Sample Preparation 2
Instrument Calibration 2
Sample Analysis 2
3. VERIFICATION TEST DESIGN 4
Objective 4
Testing Location and Conditions 4
Sample Descriptions 4
ORNL Transformer Oil Samples 4
Quality Control Samples 4
Sample Randomization 5
Summary of Experimental Design 5
Description of Performance Factors 6
Precision 6
Accuracy 6
False Positive/Negative Results 6
Completeness 6
Comparability 6
Sample Throughput 7
Ease of Use 7
Cost 7
Miscellaneous Factors 7
4. REFERENCE LABORATORY ANALYSES 8
Background 8
Reference Laboratory Selection 8
Reference Laboratory Method 9
Reference Laboratory Performance 9
5. TECHNOLOGY EVALUATION 11
Objective and Approach 11
Precision 11
Accuracy 11
False Positive/False Negative Results 11
Completeness 11
Comparability 11
Sample Throughput 13
Ease of Use 13
in
-------
Cost Assessment 13
L2000DX Costs 15
Labor 15
Equipment 16
Reference Laboratory Costs 16
Sample Shipment 16
Labor, Equipment, and Waste Disposal 16
Cost Assessment Summary 17
Miscellaneous Factors 17
Summary of Performance 17
6. REFERENCES 19
Appendix A — Dexsil's L2000DX Results Compared with Reference
Laboratory Results 21
Appendix B — Data Quality Objective (DQO) Example 25
IV
-------
List of Figures
1 L2000DX Analyzer 2
2 L2000DX PCB results versus reference laboratory results for all samples 14
3 L2000DX PCB results versus reference laboratory results, excluding samples
containing mixtures of Aroclors 14
4 Range of percent difference (%D) values for PCB results 15
-------
-------
List of Tables
1 Summary of Performance Evaluation (PE) Oil Analyses 4
2 Summary of Oil Sample Analyses 5
3 Accuracy of Reference Laboratory in Analyzing Performance Evaluation Samples 10
4 Precision of Reference Laboratory for Oil Samples 10
5 Summary of Reference Laboratory False Positive (fp) Performance on Blank Samples 10
6 Summary of the L2000DX Precision 11
7 Summary of the L2000DX Accuracy for Performance Evaluation Oils 12
8 Summary of L2000DX False Positive Performance on Blank Samples 12
9 Summary of the L2000DX Detect/Nondetect Performance Relative to the
Reference Laboratory Results 12
10 Comparison of L2000DX Results with Reference Laboratory Data 13
11 Estimated analytical costs for PCB-contaminated samples 16
12 Performance Summary forthe L2000DX 18
vn
-------
-------
Acknowledgments
The authors wish to acknowledge the support of all those who helped plan and conduct the verification test,
analyze the data, and prepare this report. In particular, we recognize the technical expertise of Viorica
Lopez-Avila (Midwest Research Institute), who was the peer reviewer of this report. For sample collection
support, we thank Kevin Norris, Charlie Bruce, and Fred Morgeson (Oak Ridge National Laboratory). The
authors also acknowledge the participation of Dexsil Corporation, in particular, Ted Lynn and Dave Balog,
who performed the analyses during the test.
For more information on the PCB Detection Technology Verification contact
Eric N. Koglin
Project Technical Leader
Environmental Protection Agency
Environmental Sciences Division
National Exposure Research Laboratory
P.O. Box 93478
Las Vegas, Nevada 89193-3478
(702) 798-2432
koglin.eric@epa.gov
For more information on Dexsil's L2000DX Analyzer contact
Ted Lynn
Dexsil Corporation
One Hamden Park Drive
Hamden, CT06517
(203)288-3509
tblynn@dexsil. com
www.dexsil.com
Roger A. Jenkins
Program Manager
Oak Ridge National Laboratory
Chemical and Analytical Sciences Division
P.O. Box 2008
Oak Ridge, TN 37831-6120
(865) 574-4871
j enkinsra@ornl .gov
-------
-------
Abbreviations and Acronyms
AL action level
DQO data quality objective
EPA U.S. Environmental Protection Agency
ERA Environmental Resource Associates
ETV Environmental Technology Verification (EPA program)
FA false acceptance decision error rate
fn false negative result
fp false positive result
FR false rejection decision error rate
ID identifier
N number of samples
NR number of replicate sets
NERL National Exposure Research Laboratory (EPA)
ORNL Oak Ridge National Laboratory
PCB polychlorinated biphenyl
PE performance evaluation
ppm parts per million (equivalent units: |_lg/mL for oils)
Pr probability
QA quality assurance
QC quality control
RSD relative standard deviation (percent)
RT regulatory threshold
SCMT Site Characterization and Monitoring Technologies Center (ETV)
SD standard deviation
TSCA Toxic Substances Control Act
UPSI United Power Services, Inc.
%D percent difference
XI
-------
-------
Section 1 — Introduction
The U.S. Environmental Protection Agency (EPA)
created the Environmental Technology Verification
Program (ETV) to facilitate the deployment of
innovative or improved environmental technologies
through performance verification and dissemination
of information. The goal of the ETV Program is to
further environmental protection by substantially
accelerating the acceptance and use of improved and
cost-effective technologies. ETV seeks to achieve
this goal by providing high-quality, peer-reviewed
data on technology performance to those involved in
the design, distribution, financing, permitting,
purchase, and use of environmental technologies.
ETV works in partnership with recognized standards
and testing organizations and stakeholder groups
consisting of regulators, buyers, and vendor
organizations, with the full participation of
individual technology developers. The program
evaluates the performance of innovative
technologies by developing verification test plans
that are responsive to the needs of stakeholders,
conducting field or laboratory tests (as appropriate),
collecting and analyzing data, and preparing peer-
reviewed reports. All evaluations are conducted in
accordance with rigorous quality assurance (QA)
protocols to ensure that data of known and adequate
quality are generated and that the results are
defensible.
ETV is a voluntary program that seeks to provide
objective performance information to all of the
participants in the environmental marketplace and to
assist them in making informed technology
decisions. ETV does not rank technologies or
compare their performance, label or list technologies
as acceptable or unacceptable, seek to determine
"best available technology," or approve or
disapprove technologies. The program does not
evaluate technologies at the bench or pilot scale and
does not conduct or support research. Rather, it
conducts and reports on testing designed to describe
the performance of technologies under a range of
environmental conditions and matrices.
The program now operates six centers covering a
broad range of environmental areas. ETV began
with a 5-year pilot phase (1995-2000) to test a wide
range of partner and procedural alternatives in
various technology areas, as well as the true market
demand for and response to such a program. In these
centers, EPA utilizes the expertise of partner
"verification organizations" to design efficient
processes for conducting performance tests of
innovative technologies. These expert partners are
both public and private organizations, including
federal laboratories, states, industry consortia, and
private sector entities. Verification organizations
oversee and report verification activities based on
testing and QA protocols developed with input from
all major stakeholder/customer groups associated
with the technology area. The verification described
in this report was administered by the Site
Characterization and Monitoring Technologies
(SCMT) Center, with Oak Ridge National
Laboratory (ORNL) serving as the verification
organization. (To learn more about ETV, visit
ETV's Web site at http://www.epa.gov/etv.) The
SCMT Center is administered by EPA's National
Exposure Research Laboratory (NERL),
Environmental Sciences Division, in Las Vegas,
Nevada.
The verification of a field analytical technology for
detection of polychlorinated biphenyls (PCBs) in
transformer oil is described in this report. The
verification test was conducted at ORNL in Oak
Ridge, Tennessee, from August 21 through August
23, 2000. The performance of the Dexsil
Corporation's L2000DX Analyzer was determined
under field conditions. The technology was
evaluated by comparing its results with those
obtained using a recognized reference laboratory
analytical method, EPA Method 600/4-81-045.
-------
Section 2 — Technology Description
In this section, the vendor (with minimal editorial changes by ORNL) provides a description of the
technology and the analytical procedure used during the verification testing activities.
General Technology Description
The L2000DX Analyzer (dimensions: 9 x 9.5 x
4.25 in.; see Figure 1) is a field-portable ion-
specific electrode instrument, weighing
approximately 5 Ib 12 oz, designed to quantify
concentrations of PCBs, chlorinated solvents, and
pesticides in soils, water, transformer oils, and
surface wipes. The L2000DX can be operated in
the field powered by a rechargeable 8-V gel cell,
or in the laboratory using 120-V AC power. In this
verification test, the lowest reported concentration
of PCBs in transformer oil was 3 ppm. The
performance of a previous version of this
instrument (the L2000 PCB/Chloride Analyzer)
was verified by ETV for soil and solvent extracts
in 1998 (EPA 1998).
Oil Sample Preparation
Sample preparation begins by collecting 5 mL of
the oil in a polyethylene reaction tube. Two glass
ampules contained in the reaction tube are broken,
introducing metallic sodium to the oil. The
mixture is then shaken for 10 s and allowed to
react for a total of 1 min. The sodium strips the
covalently bonded chlorine atoms off the PCB
molecule. An aqueous extraction solution is added
to the reaction tube to adjust the pH, destroy the
excess sodium, and extract and isolate the newly
formed chloride ions in a buffered aqueous
solution. The aqueous layer is decanted, filtered,
and collected in an analysis vial. The ion-specific
electrode is put into this aqueous solution to
measure the millivolt potential. The potential is
then converted to the equivalent PCB
concentration.
Instrument Calibration
A one-point calibration is performed prior to
sample analysis. The analyst simply follows the
menu-driven instructions prompted in the display.
When prompted, the instrument will ask if the
calibration solution is ready. The analyst inserts
the ion-specific electrode into the 50-ppm chloride
solution and then pushes the "yes" button. The
instrument will then prompt the user when the
calibration is completed. Additional calibration is
required when the instrument prompts the user,
approximately every 15 min.
Sample Analysis
To begin analysis of the sample, the analyst
chooses the appropriate Aroclor from the
Figure 1. L2000DX Analyzer.
-------
programmed menu. If the Aroclor is not known or "enter" button. After approximately 30 s, the PCB
if there is a mixture of Aroclors, Aroclor 1242 concentration of the samples (in ppm) is displayed
should be chosen for the most conservative by the L2000DX.
results. The analyst then places the electrode into
the aqueous extract solution and pushes the
-------
Section 3 — Verification Test Design
Objective
The purpose of this section is to describe the
verification test design. It is a summary of the test
plan (ORNL 2000).
Testing Location and Conditions
The verification of field analytical technologies for
PCBs was conducted on the grounds outside of
ORNL's Building 5507, in Oak Ridge, Tennessee.
The temperature and relative humidity were
monitored during field testing. Over the three days
of testing, the average temperature was 84°F, and
temperatures ranged from 63 to 98°F. The average
relative humidity was 55%, and relative humidity
ranged from 27 to 90%.
Sample Descriptions
PCBs (C^HK^Q,.) are a class of compounds that are
chlorine-substituted linked benzene rings. There are
209 possible PCB compounds (also known as
congeners). PCBs were commercially produced as
complex mixtures for use in transformers,
capacitors, paints, pesticides, and inks beginning in
1929 (Erickson 1997). Monsanto Corporation
marketed products that were mixtures of 20 to 60
PCB congeners under the trade name Aroclor.
Aroclor mixtures are identified by a number (e.g.,
Aroclor 1260) that represents the mixture's chlorine
composition as a percentage (e.g., 60%). The
samples used in this study were brought to the
testing location for evaluation by the vendor.
ORNL Transformer Oil Samples
Oils contaminated with various levels of PCBs were
collected from active and inactive transformers at
ORNL. These transformers have been in service for
decades. Because of the lack of computerized
records, historical information about these oils (such
as when the PCBs were added and the chemical
characteristics of the oils) is unavailable. It is
believed that all these oils are mineral oil. The
concentration of PCBs in these samples ranges from
<5 ppm to nearly 50 ppm, with the PCBs consisting
of single and multiple Aroclor mixtures (primarily
1242, 1254, and 1260, although other Aroclors may
be present). Because most of the native total PCB
concentrations in these samples were less than 50
ppm, ORNL augmented the Aroclor concentration
of several of these samples to increase the total PCB
concentration. The augmentation procedure is
described in the Sample Preparation section, below.
Quality Control Samples
Performance evaluation (PE) spiked samples and
certified blanks were obtained from Environmental
Resource Associates (ERA). ERA purchased the oil
used as blanks and as the spiking material from
Calumet Lubricants (Princeton, Louisiana). The
accompanying Material Safety Data Sheet calls the
material a transformer oil, with its chemical name
being a "severely hydrotreated light naphthenic
petroleum oil," CAS # 64742-53-6. As described in
Table 1, PE samples were prepared at
concentrations ranging from 5 to 175 ppm,
containing single Aroclors (1254 or 1260) and 50:50
mixtures of 1254 and 1260.
Table 1. Summary of Performance Evaluation
(PE) Oil Analyses
Nominal PCB
concentration
(ppm)
5
25
40
50
60
75
100
175
Aroclor(s)
1254
1260
1254/1260
1254/1260
1254/1260
1260
1254
1254/1260
Ratio in
mixture
n/a
n/a
50/50
50/50
50/50
n/a
n/a
50/50
Sample Preparation
The oil samples did not require homogenization.
The samples, contained in 4-oz glass jars, were split
into 10-mL aliquots using a disposable plastic
syringe. Replicate splits of each oil sample were
prepared for the vendor and the reference
-------
laboratory. Three sets of archives were also
prepared.
As mentioned previously, the ORNL transformer oil
samples originally contained PCB concentrations of
<50 ppm. Several of the transformer oils were aug-
mented with additional Aroclors (up to -200 ppm),
so that a larger dynamic range could be tested. To
spike the samples, -250 mL of oil was poured into a
1-L wide-mouth jar. A stir bar was added; then the
jar was placed on a magnetic stirrer. While the oil
was being stirred, hexane solutions with known
concentrations of Aroclors were added to increase
the total PCB concentration. One Aroclor was added
to each augmented transformer oil. Typically, the
Aroclor already present in the sample was the one
that was added. For example, if the native PCB
concentration was 3 ppm of Aroclor 1260, then
50 ppm of Aroclor 1260 was added to the oil.
The concentrations of all samples used in the study
were confirmed by an ORNL in-house method. The
oil samples were prepared by diluting 1 g of oil in
10 mL of hexane. The hexane extract was analyzed
on a Hewlett Packard 6890 gas chromatograph
equipped with an electron capture detector and an
autosampler. The analytical method used was a
slightly modified version of EPA's SW-846 dual-
column Method 8081 (EPA 1994).
Sample Randomization
The samples were randomized in two stages. First,
the order in which the filled jars were distributed
was randomized so that the vendor did not always
receive the first jar filled for a given sample set.
Second, the order of analysis was randomized so
that Dexsil and the reference laboratory analyzed the
same set of samples, but in a different order. Each
jar was labeled with a sample number. Replicate
samples were assigned unique (but not sequential)
sample numbers. Spiked materials and blanks were
labeled in the same manner, such that these quality
control (QC) samples were indistinguishable from
other samples. All samples were analyzed blindly by
both the vendor and the reference laboratory.
Summary of Experimental Design
The distribution of samples is shown in Table 2.
A total of 152 oil samples were analyzed, with
approximately 65% of the samples being naturally
contaminated and augmented transformer oils, and
the remaining 35% being PE samples and blanks.
Four replicates were analyzed for each sample type.
For example, four environmental samples were
analyzed in the 50.1- to 75.0-ppm concentration
range, indicating that 16 individual samples were
used in the study.
Table 2. Summary of Oil Sample Analyses
Target concentration range
(ppm)
Blank
<5.0
5.1-25.0
25.1-40.0
40.1-50.0
50.1-75.0
75.1-100.0
>100
All samples, incl. 4 replicates each
Number of samples"
Environmental
2
2
4
4
3
4
2
4
100
PE
5
1
1
1
1
2
1
1
52
" Four replicates were analyzed for each sample.
-------
Description of Performance Factors
In Section 5, technology performance is described in
terms of precision, accuracy, completeness, and
comparability, which are indicators of data quality
(EPA 1996). False positive and negative results,
sample throughput, and ease of use are also
described. Each of these performance characteristics
is defined in this section.
Precision
Precision is the reproducibility of measurements
under a given set of conditions. Standard deviation
(SD) and relative standard deviation (RSD) for
replicate results are used to assess precision, using
the following equation:
RSD = (SD/average concentration) x 100% .
(Eq. 1)
The overall RSD is characterized by three summary
values:
mean — i.e., average;
median — i.e., 50th percentile value, at which
50% of all individual RSD values are below and
50% are above; and
range — i.e., the highest and lowest RSD values
that were reported.
The average RSD may not be the best representation
of precision, but it is reported for convenient
reference. RSDs greater than 100% should be
viewed as indicators of large variability and possibly
non-normal distributions.
Accuracy
Accuracy represents the closeness of the tech-
nology's measured concentrations to known (in this
case, spiked/PE) values. Accuracy is assessed in
terms of percent recovery, calculated by the
following equation:
% recovery = (measured concentration/
known concentration) x 100% .
(Eq. 2)
As with precision, the overall percent recovery is
characterized by three summary values: mean,
median, and range.
False Positive/Negative Results
A false positive (fp) result is one in which the
technology detects PCBs in the sample when there
actually are none (Berger, McCarty, and Smith
1996). A false negative (fn) result is one in which
the technology indicates that no PCBs are present in
the sample when there actually are (Berger,
McCarty, and Smith 1996). The evaluation of fp and
fn results is influenced by the actual concentration
in the sample and includes an assessment of the
reporting limits of the technology.
False positive results are assessed in two ways.
First, the results are assessed relative to the blanks
(i.e., the technology reports a detected value when
the sample is a blank). Second, the results are
assessed on environmental and spiked samples
where the analyte was not detected by the reference
laboratory (i.e., the reference laboratory reports a
nondetect and the field technology reports a
detection).
False negative results, also assessed for
environmental and spiked samples, indicate the
frequency with which the technology reported a
nondetect (i.e., less than reporting limits) and the
reference laboratory reported a detection.
The reference laboratory results were validated by
ORNL so that fp/fn assessment would not be
influenced by faulty laboratory data. The reporting
limit is considered in the evaluation. For example, if
the reference laboratory reported a result as
0.9 ppm, and the technology's paired result was
reported as below reporting limits (<1 ppm), the
technology's result was considered correct and not a
false negative result.
Completeness
Completeness is defined as the percentage of
measurements that are judged to be usable (i.e., the
result is not rejected). The acceptable completeness
is 95% or greater.
Comparability
Comparability refers to how well the field
technology and reference laboratory data agree. The
difference between accuracy and comparability is
that accuracy is judged relative to a known value,
and comparability is judged relative to the results of
a standard or reference procedure, which may or
may not report the results accurately. The reference
-------
laboratory result is not assumed to be the "correct"
result. This evaluation is performed to compare the
result from the field analytical technology with what
a typical fixed analytical laboratory might report for
the same sample. A one-to-one sample comparison
of the technology results and the reference
laboratory results is performed in Section 5.
A correlation coefficient quantifies the linear
relationship between two measurements (Draper and
Smith 1981). The correlation coefficient is denoted
by the letter r; its value ranges from -1 to +1, where
0 indicates the absence of any linear relationship.
The value r = -1 indicates a perfect negative linear
relation (one measurement decreases as the second
measurement increases); the value r = +1 indicates a
perfect positive linear relation (one measurement
increases as the second measurement increases).
The slope of the linear regression line, denoted by
the letter m, is related to r. Whereas r represents the
linear association between the vendor and reference
laboratory concentrations, m quantifies the amount
of change in the vendor's measurements relative to
the reference laboratory's measurements. A value of
+1 for the slope indicates perfect agreement. (It
should be noted that the intercept of the line must be
close to zero [i.e., not statistically different from
zero], in order for the slope value of+1 to indicate
perfect agreement.) Values greater than 1 indicate
that the vendor results are generally higher than
those of the reference laboratory, while values less
than 1 indicate that the vendor results are usually
lower than the values from the reference laboratory.
In addition, a direct comparison between the field
technology and reference laboratory data is
performed by evaluating the percent difference
(%D) between the measured concentrations, defined
as
%D = ([field technology] - [reflab\)l(reflab)
x 100% . (Eq. 3)
The range of %D values is summarized and reported
in Section 5.
Sample Throughput
Sample throughput is a measure of the number of
samples that can be processed and reported by a
technology in a given period of time. This is
reported in Section 5 as number of samples per hour
or day times the number of analysts.
Ease of Use
A significant decision factor in purchasing an
instrument or a test kit is how easy the technology is
to use. Several factors are evaluated and reported on
in Section 5:
• What is the required operator skill level (e.g.,
technician or advanced degree)?
• How many operators were used during the test?
Could the technology be run by a single person?
• How much training would be required in order
to run this technology?
• How much subjective decision-making is
required?
Cost
An important factor in the consideration of whether
to purchase a technology is cost. Costs involved
with operating the technology and the standard
reference analyses are estimated in Section 5. To
account for the variability in cost data and
assumptions, the economic analysis is presented as a
list of cost elements and a range of costs for sample
analysis. Several factors affect the cost of analysis.
Where possible, these factors are addressed so that
decision makers can independently complete a site-
specific economic analysis to suit their needs.
Miscellaneous Factors
Any other information that might be useful to a
person who is considering purchasing the
technology is documented in Section 5. Examples of
information that might be useful to a prospective
purchaser are the amount of hazardous waste
generated during the analyses, the ruggedness of the
technology, the amount of electrical or battery
power necessary to operate the technology, and
aspects of the technology or method that make it
user-friendly or user-unfriendly.
-------
Section 4 — Reference Laboratory Analyses
Background
The verification process is based on the presence of
a statistically validated data set against which the
performance of the technology may be compared.
The choice of an appropriate reference method and
reference laboratory are critical to the success of the
verification test. To assess the performance of the
PCB field analytical technology, the data obtained
from the verification test participant were compared
to data obtained using a conventional analytical
method.
Verifications of technologies for the detection and
quantification of PCBs in soil and solvent extracts
occurred under the ETV program in 1997, 1998, and
2000. EPA SW-846 Method 8081 (EPA 1994) was
the reference method used for these verifications.
Since the time of the original PCB analyses, Method
8081 has been updated to Method 8082 for PCB
analyses. When planning for the verification test of
the L2000DX Analyzer to detect PCBs in
transformer oils, we considered using Method 8082
to generate the reference results. Further
investigation into reference methods indicated that
the test method outlined in EPA 600/4-81-045, The
Determination ofPolychlorinatedBiphenyls in
Transformer Fluid and Waste Oils (EPA 1982) was
more appropriate, as this is the method that is
frequently used by the utilities industry.
The fundamental difference between Methods 8082
and 600 is based on the quantification of the total
PCB concentration. In Method 8082, Aroclor
quantifications are typically performed by selecting
three to five representative peaks, confirming that
the peaks are within the established retention time
windows, integrating the selected peaks, quantifying
the peaks based on the calibrations, and averaging
the results to obtain a single concentration value for
the multi-component Aroclor. If mixtures of
Aroclors are suspected to be present, the sample is
typically quantified as the most representative
Aroclor pattern. If the identification of multiple
Aroclors is definitive, total PCBs in the sample are
calculated by summing the concentrations of all
Aroclors.
In Method 600, more peaks (typically five to ten)
are selected as representative for each Aroclor. For
quantification, the total area of these peaks is
compared to the total area curve generated from the
calibration standards. When mixtures of Aroclors
are present, typically all Aroclors are quantified and
summed to indicate a total PCB concentration.
A direct comparison by ORNL of data generated by
Method 8082 and by Method 600 on 36 split oil
samples indicated that the two methods usually
yielded comparable results for single Aroclor oils,
but that Method 600 usually generated a higher PCB
value for samples with overlapping mixtures of
Aroclors.
Reference Laboratory Selection
United Power Services, Inc. (UPSI), of Nashville,
Tennessee, was selected to perform the reference
analyses. Part of the selection process involved a
predemonstration study in which 40 oil samples
were sent to UPSI for blind analysis. Included in the
design were replicates, blanks, spikes, and actual
transformer oil samples. Results from this study
indicated that UPSI was proficient in analyzing oils
for PCBs.
ORNL performed an on-site audit of UPSI on
September 8, 2000, during the UPSI analysis of the
verification samples. The purpose of the visit was to
observe laboratory operations while ETV oil
samples were being analyzed to verify that UPSI
maintained the level of QC needed for reference
data. UPSI is a small laboratory with one analyst
dedicated to the analysis of PCBs in oil. UPSI
analyzed approximately 19,000 oils for PCBs in
1999. In addition, the company has an extremely
quick turnaround time (2-3 weeks) and competitive
prices ($10-20 per sample). Based on observations
and interviews during the visit, ORNL staff
concluded that the analyst appeared to be very
meticulous and conscientious, and that the
laboratory manager was actively involved with the
analyses and reviewed all of the data before they
were finalized.
Although the laboratory lacked updated written
procedures for both method analysis and quality
-------
assurance, UPSI had a wealth of QC data which it
uses to evaluate method performance on a daily
basis. These data could be compiled and evaluated
to demonstrate long-term method performance, but
this task has not yet been undertaken because of
staffing constraints. Data archival is limited to
storage of one hard copy in the laboratory
warehouse. No electronic data are saved for longer
than one month. Since ETV required a hard copy of
all data from UPSI, this was not a significant
concern. The analyst performs all calculations
manually using a calculator. This may lead to some
data entry errors, but the auditors felt that the lack of
automation was not a significant hindrance, and
would in fact require that all of the data be
examined carefully. Because the scope of this work
is so limited, resolution of issues with long-term
data retrievability and insufficient written
procedures were considered satisfactory, and the
laboratory was considered acceptable for performing
these analyses.
Reference Laboratory Method
The reference laboratory's analytical method,
presented in the verification test plan, followed the
guidelines established in EPA Method 600/4-81-045
(EPA 1982). An oil sample was prepared by
pipetting 0.5 mL of sample into a test tube, weighing
the sample, and then adding 2 mL of sulfuric acid
and 4.5 mL of isooctane. The test tube was shaken,
the mixture was allowed to settle, and then 1 mL of
the isooctane layer was placed undiluted in an
autosampler vial for analysis. After an initial
analysis, the samples were diluted and reanalyzed as
appropriate.
Three calibration standards at 1, 5, and 10 ppm were
analyzed daily for Aroclors 1242, 1254, and 1260. A
linear regression curve from the total area of five to
ten representative peaks was generated for each
Aroclor. Calibration check standards at 5 ppm were
alternated every tenth analysis with a 50-ppm QC
spiked oil sample. If the calibration check or the QC
sample was outside the established acceptance limits
(±10% of nominal value), five samples before and
five samples after the calibration check or QC
sample in question were reanalyzed.
The analyses were performed on a Hewlett Packard
5890 gas chromatograph equipped with an electron
capture detector. The capillary column used was an
Alltech (Deerfield, Illinois) Pesticide column (20 m
x 0.53 mm x 0.6 |_lm film thickness). The column
temperature program was an isothermal 17-min
analysis at 185°C. The detector temperature was
350°C, and the injector temperature was 275°C.
Aroclors were identified by visually matching the
peak pattern to that of a standard Aroclor. A quanti-
tative result was generated using the total area of
representative peaks in the sample and a linear
regression equation. The lowest reported
concentration was typically 1 ppm.
Reference Laboratory Performance
ORNL validated all of the reference laboratory data
according to the procedure described in the
verification test plan (ORNL 2000). During the
validation, the following aspects of the data were
reviewed: completeness of the data package,
correctness of the data, correlation between replicate
sample results, evaluation of QC sample results, and
evaluation of spiked sample results. Each of these
categories is described in detail in the verification
test plan. The reference laboratory results met
performance acceptance requirements on all QC
samples. An evaluation of the performance of the
reference laboratory results through statistical
analysis of the data was performed and is
summarized below.
Tables 3, 4, and 5 provide summaries of the
performance of the reference laboratory. In Table 3,
the accuracy of the laboratory measurements is
presented through an analysis of results from the 32
PE samples. Half of the samples were spiked with
one Aroclor, either 1254 or 1260, while the other
samples were spiked with a 50:50 mixture of the
two. The reference laboratory's results were biased
high by approximately 30% on the PE samples with
a mixture of Aroclors, while the results were
unbiased (mean % recovery = 97%) on the single-
Aroclor samples. This bias on samples containing
Aroclor mixtures is due to the quantification
method. Aroclors 1254 and 1260 have several
analyte peaks in common. In this quantification
method, there is no attempt to compensate for
overlapping peaks. Therefore, when both Aroclors
are present, the peaks that are common to both
Aroclors are essentially over-quantified, and the
method generates a result that is biased high.
Table 4 presents the precision of the method for
samples where all four replicates were reported as a
-------
Table 3. Accuracy of Reference Laboratory in Analyzing Performance Evaluation Samples
Statistic
Average
Median
Range of results
% Recovery
Single Aroclor
(either 1254 or 1260)
(N = 16)
97
97
80-120
Mixture of Aroclors
(1254 & 1260)
(N = 16)
134
133
114-149
All data
(N = 32)
115
120
80-149
Table 4. Precision of Reference
Laboratory for Oil
Samples
Statistic
Average
Median
Range
XT "
NR
% RSD
11.4
10.6
1-41
31
" Based on sample sets where all four replicates
were reported as a detection.
TableS. Summary of Reference
Laboratory False Positive (fp)
Performance on Blank Samples
Statistic
No. of data points
No. of fp results
% of total results that were fp
Oil samples
20
0
0
detection. The mean RSD was 11%. The reference
laboratory did not report PCBs in any of the
20 blank oil samples (Table 5). Overall, ORNL
concluded that the reference laboratory results were
acceptable for comparison with the field analytical
technology because the method adequately
represents the industry standard. The high bias on
mixtures of Aroclor was considered in the
comparison with the field technology results.
10
-------
Section 5 — Technology Evaluation
Objective and Approach
The purpose of this section is to present a statistical
evaluation of the L2000DX data and determine the
technology's ability to measure PCBs in transformer
oil samples. This section includes an evaluation of
comparability through a one-to-one comparison with
the reference laboratory data. Other aspects of the
technology (such as cost, sample throughput,
hazardous waste generation, and logistical
operation) are also evaluated in this section.
Appendix A contains the raw data provided by the
vendor during the verification test that were used to
assess the performance of the L2000DX. Appendix
B is a data quality objective example which
incorporates the performance information generated
during this test into a real-world scenario. During
the verification test, Dexsil was provided with
information as to which Aroclors were present in the
sample based on what was reported by the reference
laboratory in the predemonstration study. Dexsil
used this information to determine the final sample
results.
Precision
Precision is the reproducibility of measurements
under a given set of conditions. Precision was
determined by examining the results of blind
analyses for four replicate samples. Data were
evaluated only for those samples where all four
replicates were reported as a detection. For example,
NR = 31 (31 sets of four replicates) represents a total
of 124 individual sample analyses. A summary of
the overall precision of the L2000DX for the oil
sample results is presented in Table 6. The mean
RSD for the samples was 11%. A 95th percentile
value of 21% (Table 6) indicates that all but 5% of
the RSD values were 21% or below.
Accuracy
Accuracy represents the closeness of the
L2000DX's measured concentrations to the known
content of spiked samples. Table 7 presents a
summary of the L2000DX's overall accuracy for the
oil results. The table shows percent recoveries for
the single-Aroclor and mixed-Aroclor samples
separately for comparison with the reference
laboratory, since the laboratory results indicated a
significant high bias for mixtures. The percent
Table 6. Summary of the
L2000DX Precision
Statistic
Mean
Median
95th percentile
Range
%RSDfl
(NR = 31*)
11
9
21
3-67
"Calculated only from those samples where all
four replicates were reported as a detection.
6NR = number of replicate sets.
recovery values for the single Aroclor PEs and the
mixture PEs were comparable, indicating that the
L2000DX was unbiased for all PE samples. The
overall percent recovery was a mean value of 112%.
False Positive/False Negative Results
Table 8 shows the L2000DX performance for false
positive (fp) results for blank samples. Of the
20 blank transformer oils, Dexsil reported 5 samples
with detectable quantities of PCBs (25% fp).
Table 9 summarizes the L2000DX's fp and fn
results relative to the reference laboratory results.
(See Section 3 for a more detailed discussion of this
evaluation.) For the transformer and spiked oils (i.e.,
excluding the blank samples), one PCB result (12%
fp rate) was reported as a false positive relative to
eight reference laboratory ono-detect results. There
were no false negatives (i.e., where the laboratory
reported a detection and Dexsil reported a
nondetect).
Completeness
Completeness is defined as the percentage of
measurements that are judged to be usable (i.e., the
result was not rejected). Valid results were obtained
by the technology for all 152 oil samples. Therefore,
completeness was 100%.
Comparability
Comparability refers to how well the L2000DX and
reference laboratory data agreed. In this evaluation,
11
-------
Table 7. Summary of the L2000DX Accuracy for Performance Evaluation Oils
Statistic
Mean
Median
Range of results
% recovery
Single Aroclor
(either 1254 or 1260)
(N = 16)
119
104
80-315
Mixture of Aroclors
(1254 and 1260)
(N = 16 )
105
100
83-154
All data
(N = 32)
112
101
80-315
Table 8. Summary of L2000DX False Positive
Performance on Blank Samples
Statistic
No. of data points
No. of fp results
% of total results that were fp
Oil samples
20
5
25%
Table 9. Summary of the L2000DX Detect/
Nondetect Performance Relative to
the Reference Laboratory Results
Statistic
No. of results where lab
reported non-detect
No. of fp results
% of total results that were fp
No. of results where lab
reported detection
No. of fn results
% of total results that were fn
Oil Samples
8°
1
13%
124
0
0
" This evaluation does not include 20 blanks.
12
-------
the laboratory results are not presumed to be the
"correct" answers. Rather, these results represent
what a typical fixed laboratory would report for
these types of samples. A one-to-one sample
comparison of the L2000DX results and the
reference laboratory results was performed for all
transformer and spiked samples that were reported
as a detection. (Appendix A provides the raw data.
See Section 4 for a complete evaluation of the
reference laboratory results.)
In Table 10, the comparability of the results are
presented in terms of correlation coefficients (r) and
slopes (rri). The data were also evaluated excluding
those samples that the reference laboratory reported
as mixtures, since the laboratory was biased on the
analysis of these samples. The comparison was also
performed using high (MOO ppm) and low
(<100 ppm) concentrations. As shown in Table 10,
the correlation coefficients for all samples ranged
from 0.83 to 0.92 (m ranged from 0.67 to 0.89).
When the samples that were mixtures were
excluded, the range of correlation coefficients was
tighter (0.93 to 0.95), and the slope values were
closer to 1.0, indicating more consistent agreement
between the L2000DX and the reference laboratory.
Figures 2 and 3 are plots of the L2000DX PCB
results versus those for the reference laboratory for
all results (N = 124) and excluding the samples
reported as mixtures by the reference laboratory
(N = 92). These figures illustrate that Dexsil's
results generally agreed with those of the reference
laboratory. The high bias of the reference laboratory
on samples containing Aroclor mixtures most
significantly affected the comparability for
concentrations less than 100 ppm, where the r
values were 0.83 and 0.93, including and excluding
the biased reference laboratory results, respectively.
Another metric of comparability is the percent
difference (%D) between the reference laboratory
and the L2000DX results (see Section 3).The ranges
of %D values for the PCB results are presented in
Figure 4. Acceptable %D values would be between
-25% and 25%, or near the middle of the x-axis of
the plots. Approximately 41% of the results were
between -25% and 25%. The L2000DX values were
usually higher than those of the reference laboratory
on the remaining 60% of the results.
Sample Throughput
Sample throughput is representative of the estimated
amount of time required to prepare and analyze the
sample and perform the data analysis. Operating in
the field, the two-person Dexsil team accomplished
a sample throughput rate of approximately eight
samples per hour for the 152 oil analyses.
Ease of Use
Two operators were used for the test because of the
number of samples and the working conditions, but
the technology can be operated by a single person.
Users unfamiliar with the technology may need
approximately one-half day of additional training to
operate the instrument. No particular level of
educational training is required for the operator.
Cost Assessment
The purpose of this economic analysis is to estimate
the range of costs for analysis of PCB-contaminated
Table 10. Comparison of L2000DX Results with Reference Laboratory Data
Description of
sample set
All values, excluding
nondetects
< 100 ppm
>100 ppm
All data
N
124
89
35
Correlation
coefficient
(r)
0.92
0.83
0.89
Slope
(m)
0.89
0.79
0.67
Excluding samples that
reference laboratory reported
as mixtures
N
92
68
24
Correlation
coefficient
(r)
0.95
0.93
0.94
Slope
(m)
1.1
0.89
0.97
13
-------
350
I 300
&
I 250
1 200
r = 0.92
m = 0.89
intercept = 22.8 ppm
50
100
150
200
250
300
350
Reference laboratory total PCB concentration (ppm)
Figure 2. L2000DX PCB results versus reference laboratory results for all samples.
300
250
| 200
a 150 -
100 H
o
o
o
r = 0.95
m=l.l
intercept = 11.5 ppm
50 100 150 200
Reference laboratory total PCB concentration (ppm)
250
Figure 3. L2000DX PCB results versus reference laboratory results, excluding samples containing
mixtures of Aroclors.
14
-------
35 -,
30 -
25 -
I 20 '
15-1
10 -
5 -
O
O
Range of percent difference values
Figure 4. Range of percent difference (%D) values for PCB results.
oil samples using the L2000DX and a conventional
analytical reference laboratory method. The analysis
was based on the results and experience gained from
this verification test, costs provided by Dexsil, and
representative costs provided by the reference
analytical laboratory to analyze the samples. To
account for the variability in cost data and
assumptions, the economic analysis is presented as a
list of cost elements and a range of costs for sample
analysis by the L2000DX instrument and by the
reference laboratory.
Several factors affected the cost of analysis. Where
possible, these factors were addressed so that
decision makers can complete a site-specific
economic analysis to suit their needs. The following
categories are considered in the estimate:
sample shipment costs,
labor costs, and
equipment costs.
Each of these cost factors is defined and discussed
and serves as the basis for the estimated cost ranges
presented in Table 11. This analysis assumed that
the individuals performing the analyses were fully
trained to operate the technology. Costs for sample
acquisition and pre-analytical sample preparation,
tasks common to both methods, were not included in
this assessment.
L2000DXCosts
The costs associated with using the L2000DX
instrument included labor and equipment costs. No
sample shipment charges were associated with the
cost of operating the instrument because the samples
were analyzed on site.
Labor
Labor costs included mobilization and demobili-
zation, travel, per diem expenses, and on-site labor.
• Mobilization and demobilization. This cost
element included the time for one person to
prepare for and travel to each site. This estimate
ranged from zero (if the analyst is on site) to
5h, atarateof$50/h.
• Travel. This element was the cost for the
analyst(s) to travel to the site. If the analyst is
located at the site, the cost of commuting to the
site would be zero. The estimated cost for an
analyst to travel to the site for this verification
test ($1000) included the cost of airline travel
and rental car fees.
15
-------
Table 11. Estimated analytical costs for PCB-contaminated samples
Analysis method: L2000DX
Analyst/manufacturer: Dexsil Corporation
Sample throughput: 8 samples/h
Cost category Cost ($)
Sample shipment 0
Labor
Mobilization/demobilization 0-250
Travel 0-1,000 per analyst
Per diem expenses 0- 1 50/day per analyst
Rate 30-7 5/h per analyst
Equipment
Mobilization/demobilization 0-150
Instrument purchase price 3500
Instrument lease price 500 per month
Reagents/supplies 5 per sample
Analysis method:
Analyst/manufacturer:
Typical turnaround:
Cost category
Sample shipment
Labor
Overnight shipping
Labor
EPA 600/4-8 1-045
Reference laboratory
14-30 working days
Cost ($)
100-200
50-150
Mobilization/demobilization Included "
Travel
Per diem expenses
Rate
Equipment
Included
Included
10-26 per sample
Included
""Included" indicates that the cost is included in the labor rate.
• Per diem expenses. This cost element included
food, lodging, and incidental expenses. The
estimate ranged from zero (for a local site) to
$150/day for each analyst.
Rate. The cost of the on-site labor was estimated
at a rate of $30-75/h, depending on the required
expertise level of the analyst. This cost element
included the labor involved during the entire
analytical process, comprising sample
preparation, sample management, analysis, and
reporting.
Equipment
Equipment costs included mobilization and
demobilization, rental fees or purchase of
equipment, and the reagents and other consumable
supplies necessary to complete the analysis.
• Mobilization and demobilization. This included
the cost of shipping the equipment to the test
site. If the site is local, the cost would be zero.
For this verification test, the cost of shipping
equipment and supplies was estimated at $150.
• Instrument purchase or lease. The instrument
can be purchased for $3500. This price includes
enough reagents for 40 tests. The instrument can
also be leased for $500 per month. Leasing the
instrument requires a prepaid, refundable $2000
deposit.
• Reagents and supplies. Reagents and supplies
for the transformer oil analysis are
approximately $5 per sample.
Reference Laboratory Costs
Sample Shipment
The costs of shipping samples to the reference
laboratory included overnight shipping charges as
well as labor charges associated with the various
organizations involved in the shipping process.
Labor. This cost element included all of the
tasks associated with shipping the samples to the
reference laboratory. Tasks included packing the
shipping coolers, completing the chain-of-
custody documentation, and completing the
shipping forms. The estimate to complete this
task ranged from 2 to 4 h, at $50 per hour.
• Overnight shipping. The overnight express
shipping service cost was estimated to be $50
for one 50-lb cooler of samples.
Labor, Equipment, and Waste Disposal
The labor bids from commercial analytical reference
laboratories that offered to perform the reference
analysis for this verification test ranged from $10 to
$26 per sample. The bid was dependent on many
factors, including the perceived difficulty of the
sample matrix, the current workload of the
16
-------
laboratory, data packaging, and the competitiveness
of the market. This rate was a fully loaded analytical
cost that included equipment, labor, waste disposal,
and report preparation.
Cost Assessment Summary
An overall cost estimate for use of the L2000DX
instrument versus use of the reference laboratory
was not made because of the extent of variation in
the different cost factors, as outlined in Table 11.
The overall costs for the application of any
technology would be based on the number of
samples requiring analysis, the sample type, and the
site location and characteristics. Decision-making
factors, such as turnaround time for results, must
also be weighed against the cost estimate to
determine the value of the field technology's
providing immediate answers versus the reference
laboratory's provision of reporting data within
30 days of receipt of samples.
Miscellaneous Factors
The following are general observations regarding
the field operation and performance of the
L2000DX instrument:
• The L2000DX required no electrical power and
worked continuously through a 10-h workday
without the need for recharging the battery.
The Dexsil team was ready for its first set of
samples within 2 h of arriving on site.
• The Dexsil team used information on which
Aroclors were in the samples to determine the
final sample result (based on the instrumental
response for each Aroclor). If the Aroclor had
been unknown, the calibration curve for Aroclor
1242 would have been used or the result would
have been reported as total chloride
concentration.
Tests with the L2000DX generated the
following waste: 57 L of TSCA-regulated solids
and 2 L of nonregulated liquid/aqueous waste.
Careful segregation of the waste could have
reduced the volume of TSCA-regulated waste to
4L.
Summary of Performance
A summary of performance is presented in Table 12.
Precision, defined as the mean RSD, was 11% for
the oil analyses. Accuracy, defined as the mean
percent recovery relative to the spiked
concentration, was 112%. Of the 20 blank oils,
Dexsil reported PCBs in 5 sample (25% false
positives). In addition, false positive and false
negative results were determined by comparing the
L2000DX results with the reference laboratory
results for the environmental and spiked samples.
One of the results was reported as a false positive
(13% fp), but none were false negatives. A one-to-
one matching of the L2000DX and reference
laboratory results indicates that the results were
comparable, with an overall correlation coefficient
of 0.92 and a slope value of 0.89. The reference
laboratory results were biased high for samples
which were mixtures of Aroclors. If those samples
are removed from the comparison, the correlation
coefficient and slope values improve to 0.95 and
1.1, respectively.
The verification test found that the L2000DX
instrument was relatively simple for a trained
analyst to operate in the field, requiring less than an
hour for initial setup. The sample throughput of the
L2000DX was eight samples per hour. Two
operators analyzed samples during the verification
test, but the technology can be run by a single
trained operator. The overall performance of the
L2000DX for the analysis of PCBs in transformer
oil was characterized as unbiased and precise.
17
-------
Table 12. Performance Summary for the L2000DX
Feature/parameter
Precision
Accuracy
False positive results on blank
samples
False positive results relative to
reference laboratory results
False negative results relative to
reference laboratory results
Comparison with reference
laboratory results (all data,
excluding suspect values)
Completeness
Weight
Sample throughput (2 operators)
Power requirements
Training requirements
Cost
Waste generated
Overall evaluation
Performance summary
MeanRSD: 11%
Mean recovery: 112%
25%
13%
None
Median
r m Absolute% D
All values: 0.92 0.89 31%
Excluding mixtures: 0.95 1.1 31%
100% of 152 oil samples
61b
8 samples/h
battery operated (8 V gel cell)
One-half day instrument-specific training
Purchase: $3,500
Lease: $500 per month (plus $2,000 refundable deposit)
Reagents/Supplies: $5 per oil sample
57 L of TSCA-regulated solids (which could have been reduced to
4 L by segregation of the waste)
2 L of nonregulated liquid/aqueous waste
(Total number of samples analyzed: 152)
Precise
Unbiased
18
-------
Section 6 — References
ASTM (American Society for Testing and Materials). 1997a. Standard Practice for Generation of
Environmental Data Related to Waste Management Activities: Quality Assurance and Quality Control
Planning and Implementation. D5283-92.
ASTM (American Society for Testing and Materials). 1997b. Standard Practice for Generation of
Environmental Data Related to Waste Management Activities: Development of Data Quality Objectives.
D5792-95.
Berger, W., H. McCarty, and R-K. Smith. 1996. Environmental Laboratory Data Evaluation. Genium
Publishing Corp., Schenectady, N.Y.
Draper, N. R., and H. Smith. 1981. Applied Regression Analysis. 2nd ed. John Wiley & Sons, New York.
EPA (U.S. Environmental Protection Agency). 1982. Test Method: The Determination ofPolychlorinated
Biphenyls in Transformer Fluid and Waste Oils. EPA/600/4-81-045. U.S. Environmental Protection Agency,
Washington, B.C., September.
EPA (U.S. Environmental Protection Agency). 1994. "Method 8081: Organochlorine Pesticides and PCBs as
Aroclors by Gas Chromatography: Capillary Column Technique." In Test Methods for Evaluating Solid
Waste: Physical/Chemical Methods (SW-846). 3d ed., Final Update II. Office of Solid Waste and Emergency
Response, Washington, B.C., September.
EPA (U.S. Environmental Protection Agency). 1996. Guidance for Data Quality Assessment. EPA QA/G-9;
EPA/600/R-96/084. EPA, Washington, B.C., July.
EPA (U.S. Environmental Protection Agency). 1998. Environmental Technology Verification Report:
Electrochemical Technique/Ion Specific Electrode, Dexsil Corporation, L2000 PCB/Chloride Analyzer.
EPA/600/R-98/109. U.S. Environmental Protection Agency, Washington, B.C., August.
Erickson, M. B. 1997. Analytical Chemistry of PCBs. 2nd ed. CRC Press/Lewis Publishers, Boca Raton, Fla.
ORNL (Oak Ridge National Laboratory). 2000. Technology Verification Test Plan: Evaluation of
Poly chlorinated Biphenyl (PCS) Field Analytical Techniques. Chemical and Analytical Sciences Bivision,
Oak Ridge National Laboratory, Oak Ridge, Tenn., August.
19
-------
-------
Appendix A
Dexsil's L2000DX Results Compared with
Reference Laboratory Results
Sample ID
101
101
101
101
102
102
102
102
103
103
103
103
104
104
104
104
105
105
105
105
106
106
106
106
107
107
107
107
108
108
108
108
109
109
109
109
Replicate
1
2
o
6
4
1
2
3
4
1
2
o
5
4
1
2
3
4
1
2
o
3
4
1
2
3
4
1
2
o
5
4
1
2
3
4
1
2
o
3
4
Total PCB
L2000DX
<3.0
<3.0
<3.0
<3.0
<3.0
3
<3.0
<3.0
10.4
10
9.3
9.4
9.1
11.1
12
7.5
23.9
25.1
19.6
15.2
38.8
41.7
39.3
38.8
54.2
53.7
55.4
61.9
39.4
42.1
40.4
41
70.8
74.2
73.2
79.4
cone, (ppm)
Reference lab
<1.0
<1.0
<1.0
<1.0
<1.0
<1.0
<1.0
<1.0
o
5
2
1
2
4
o
6
2
2
8
9
12
10
15
11
14
16
21
21
23
23
23
25
20
26
26
27
24
32
Aroclor(s)
reported by
lab
1254
1254
1254
1254
1254
1254
1254
1254
1260
1260
1260
1260
1260
1260
1260
1260
1260
1260
1260
1260
1260
1260
1260
1260
1260
1260
1260
1260
1260
1260
1260
1260
1242/1254
1242/1254
1242/1254
1242/1254
Dexsil analysis
order8
1108
1010
1067
1027
1123
1001
1005
1054
1035
1137
1053
1056
1023
1149
1074
1118
1078
1085
1141
1046
1068
1036
1013
1063
1031
1016
1044
1145
1058
1026
1143
1019
1049
1037
1042
1127
21
-------
Sample ID
110
110
110
110
111
111
111
111
112
112
112
112
113
113
113
113
114
114
114
114
115
115
115
115
116
116
116
116
117
117
117
117
118
118
118
118
119
119
119
119
Replicate
1
2
o
5
4
1
2
3
4
1
2
o
3
4
1
2
3
4
1
2
o
5
4
1
2
3
4
1
2
o
3
4
1
2
3
4
1
2
o
3
4
1
2
3
4
Total PCB
L2000DX
100.2
85.8
87.9
101
103.1
101.6
108.3
102.4
58.3
61.4
71.5
64.8
89.8
96.2
81.1
75.4
60.6
59
65
64.6
85.4
84.8
89.4
125.4
93.1
84
102.2
97.6
103.5
84.8
79.6
89.4
98.2
97.8
90.8
86.8
178.5
166.2
168.9
159.8
cone, (ppm)
Reference lab
40
41
32
33
37
45
44
39
43
35
42
40
80
70
93
65
64
65
58
56
66
90
67
64
58
59
64
75
68
81
69
69
79
102
83
102
125
131
129
130
Aroclor(s)
reported by
lab
1254/1260
1254/1260
1254/1260
1254/1260
1260
1260
1260
1260
1260
1260
1260
1260
1260
1260
1260
1260
1260
1260
1260
1260
1260
1260
1260
1260
1254/1260
1254/1260
1254/1260
1254/1260
1260
1260
1260
1260
1254/1260
1254/1260
1254/1260
1254/1260
1260
1260
1260
1260
Dexsil analysis
order"
1092
1065
1034
1077
1055
1017
1151
1015
1121
1122
1079
1002
1091
1080
1073
1097
1018
1032
1113
1114
1105
1009
1096
1088
1022
1103
1136
1152
1084
1048
1098
1130
1131
1116
1147
1138
1117
1125
1095
1059
22
-------
Sample ID
120
120
120
120
121
121
121
121
122
122
122
122
123
123
123
123
124
124
124
124
125
125
125
125
126
126
126
126
127
127
127
127
128
128
128
128
129
129
129
129
Replicate
1
2
o
5
4
1
2
3
4
1
2
o
3
4
1
2
3
4
1
2
o
5
4
1
2
3
4
1
2
o
3
4
1
2
3
4
1
2
o
3
4
1
2
3
4
Total PCB
L2000DX
103.5
88.7
87.5
85.1
98.3
105.8
96.4
97.1
88.1
97.6
110.4
111.6
245.8
245.8
241.6
268.7
212.9
258.3
208.3
228.1
233.3
225.4
227
270.6
4.7
<3.0
6.4
28.4
3.0
<3.0
<3.0
6
<3.0
<3.0
<3.0
<3.0
<3.0
<3.0
<3.0
<3.0
cone, (ppm)
Reference lab
78
77
77
84
82
83
92
92
88
79
98
78
287
272
248
300
171
169
166
169
194
214
194
196
<1.0
<1.0
<1.0
<1.0
<1.0
<1.0
<1.0
<1.0
<1.0
<1.0
<1.0
<1.0
<1.0
<1.0
<2.0
<1.0
Aroclor(s)
reported by
lab
1242
1242
1242
1242
1260
1260
1260
1260
1242
1242
1242
1242
1260
1260
1260
1260
1260
1260
1260
1260
1260
1260
1260
1260
1254
1254
1254
1254
1254
1254
1254
1254
1254
1254
1254
1254
1254
1254
1254
1254
Dexsil analysis
order"
1087
1109
1142
1039
1051
1129
1030
1033
1104
1062
1133
1140
1050
1029
1057
1135
1069
1093
1020
1111
1014
1064
1043
1086
1004
1107
1090
1083
1072
1012
1119
1075
1128
1146
1115
1132
1038
1102
1071
1052
23
-------
Sample ID
130
130
130
130
131
131
131
131
132
132
132
132
133
133
133
133
134
134
134
134
135
135
135
135
136
136
136
136
137
137
137
137
138
138
138
138
Replicate
1
2
o
5
4
1
2
3
4
1
2
o
3
4
1
2
3
4
1
2
o
5
4
1
2
3
4
1
2
o
3
4
1
2
3
4
1
2
o
3
4
Total PCB
L2000DX
<3.0
<3.0
<3.0
3.4
4
5.1
7
15.7
24.2
22.3
24.2
26.4
33.1
39
44.3
39.7
45.8
51.1
54.2
49.1
59.6
63.8
56.7
52
64.6
84.4
98.3
85.2
107.4
98.8
94.1
130.6
175.4
271.9
233.9
186.2
cone, (ppm)
Reference lab
<1.0
<1.0
<1.0
<1.0
6
4
5
6
21
21
20
25
51
54
50
54
57
74
61
67
78
76
79
78
83
64
73
73
86
86
108
109
260
261
255
249
Aroclor(s)
reported by
lab
1254
1254
1254
1254
1254
1254
1254
1254
1260
1260
1260
1260
1254/1260
1254/1260
1254/1260
1254/1260
1254/1260
1254/1260
1254/1260
1254/1260
1254/1260
1254/1260
1254/1260
1254/1260
1260
1260
1260
1260
1254
1254
1254
1254
1254/1260
1254/1260
1254/1260
1254/1260
Dexsil analysis
order"
1008
1124
1025
1076
1110
1099
1139
1081
1150
1070
1007
1089
1106
1126
1066
1120
1061
1040
1144
1047
1148
1041
1024
1060
1006
1134
1101
1003
1112
1028
1021
1100
1045
1094
1082
1011
"These are the sample numbers from which the analysis order can be discerned. For example, 1001 was the first sample analyzed,
then 1002, etc.
24
-------
Appendix B
Data Quality Objective (DQO) Example
Disclaimer
The following hypothetical example serves to demonstrate how the information provided in this report may
be used in the data quality objectives (DQO) process. This example serves to illustrate the application of
quantitative DQOs to a decision process but cannot attempt to provide a thorough education in this topic.
Please refer to other educational ortechnical resources for further details (ASTM 1997a, b; EPA 1996). In
addition, since the focus of this report is on the analytical technology, this example makes the simplifying
assumption that the sample will be homogeneous. In the real world, however, this assumption is seldom
valid, and matrix heterogeneity constitutes a source of considerable uncertainty that must be adequately
evaluated if the overall certainty of a site decision is to be quantified.
Background and Problem Statement
An industrial company discovered a warehouse of transformers which were filled with PCB-containing oils.
The contaminated oil was transferred into waste drums. Preliminary characterization determined that the PCB
concentration in a single drum was homogenous but that PCB concentrations varied greatly from drum to
drum. The company's DQO team was considering the use of Dexsil's L2000DX to measure the PCB
concentration in each drum. The DQO team decided that drums would be disposed of by incineration if the
PCB concentration was >50 ppm ("hot"). The concentration of 50 ppm is the Toxic Substances Control Act
(TSCA) regulatory threshold (RT) for this environmental problem. Those drums with PCB concentrations
<50 ppm would be put into a landfill because incineration of oil is very expensive. With regulator agreement,
the DQO team determined that a decision rule for disposal would be based on the average concentration of
PCBs in each drum.
General Decision Rule
If average PCB Concentration < 50 ppm, then send the oil drum to the landfill.
If average PCB Concentration > 50 ppm, then send the oil drum to the incinerator.
DQO Goals
The DQO team's primary goal was to calculate how many samples would need to be analyzed by the
L2000DX Analyzer in order to confidently make a decision about disposing of the oil, given the uncertainties
of the technology's results. Because the team decided that inadvertently sending oil to the landfill that
exceeded the 50 ppm concentration of PCBs was the worst possible mistake, the number of samples
measured was primarily related to this false rejection decision error rate. A secondary decision error would
be to incinerate unnecessarily the drum that contained PCB concentrations <50 ppm, which would be a false
acceptance decision error. Both the false rejection decision error and the false acceptance decision error were
taken into account to determine the final sampling plan.
The team required that the error rate for sending a "hot" drum to the landfill (i.e., the false rejection error rate
for the decision) could not be more than 5%. Therefore, a sufficient number of samples had to be taken from
each drum to ensure that the false rejection decision error rate (FR) would be 0.05 (or less) that the true drum
concentration is 50 ppm. This scenario represents a 5% chance of sending a drum containing PCB
concentrations >50 ppm to the landfill.
25
-------
The DQO team did not want to send an excessive number of drums to the incinerator if the average PCB
concentration was <50 ppm because of the expense. In this situation, a false acceptance decision is made
when it is concluded that a drum is "hot," when in actuality, the drum contains oil with <50 ppm PCBs.
Therefore, the team recommended that the false acceptance decision error rate (FA) be 0.10 if the true drum
concentration was 40 ppm. That is, there would be a 10% probability of sending a drum to the incinerator
(denoted as PrjTake Drum to Incinerator]) if the true PCB concentration for a drum was 40 ppm.
Permissible FR and FA Error Rates and Critical Decision Points
FR: Pr[Take Drum to Landfill] <0.05 when true PCB concentration = 50 ppm
FA: Pr[Take Drum to Incinerator] <0.10 when true PCB concentration = 40 ppm
Use of Technology Performance Information to Implement the Decision Rule
Technology performance information is used to evaluate whether a particular analytical technology can
produce data of sufficient quality to support the site decision. Because the DQO team was considering the
use of the L2000DX, the performance of this technology (as reported in this ETV report) was used to assess
its applicability to this project. Two questions arise:
1. How many samples are needed from a single drum to permit a valid estimation of the true average
concentration of PCBs in the drum to the specified certainty? Recall that the simplifying assumption was
made that the PCB distribution throughout the oil within a single drum is homogeneous and thus, matrix
heterogeneity will not contribute to overall variability. The only variability to be considered in this
example, then, is the variability in the L2000DX's analytical method, which is determined by precision
studies.
2. What is the appropriate action level (AL) for using the Dexsil L2000DX to make decisions in the field?
After the required number of samples have been collected from a drum and analyzed, the results are
averaged together to get an estimate of the "true" PCB concentration of the drum. When using the
L2000DX, what is the value (here called "the action level for the decision rule") to which that average is
compared, to decide if the drum is "hot" or not? This method-specific or site-specific action level is
derived from evaluations of the method's accuracy using an appropriate quality control regimen.
L2000DX Analyzer Accuracy
The ETV verification results indicated that the PCB concentrations determined by the L2000DX were
unbiased when compared with concentration values for performance evaluation samples. Also, the
concentration measurements with the L2000DX had a strong linear correlation (r = 0.92) with the
concentration values measured by the reference laboratory using EPA Method 600/4-81-045 (EPA 1982).
Figure B-l shows a linear relationship between L2000DX's PCB concentrations and the certified PCB values
for the performance evaluation samples, which were in the concentration range of 0 to 60 ppm. The equation
for the PCB prediction line is:
L2000DXresult = -2.51 + 0.93 x (CertifiedPE Value) (Eq. B-l)
There is no significant difference at the 5% significance level between the intercept value and zero, and the
slope value and 1, in Eq. B-l. The L2000DX measurements were therefore unbiased for this concentration
range.
26
-------
60-
, so-
Determining the Number of
Samples
With the critical decision points selected, the
team could then determine the number of
samples needed from each drum to calculate the
drum's "true" average PCB concentration. For a
homogeneous matrix, the number of samples
required depends on the precision of the
analytical method.
As can be seen in Figure B-2, the standard
deviations do not increase with increasing
certified concentrations for the range of interest
from 5 to 60 ppm. Therefore, the precision of the
L2000DX can be represented by a pulled
standard deviation of 4.2 ppm within the
concentration range of 5 to 60 ppm (see
Figure B-2). The pulled standard deviation is
calculated by first calculating the average
variance and then taking the square root. This
estimate of analytical variability (precision) is
used to calculate the number of oil samples
required to be analyzed from each drum to
achieve the DQOs as determined above under
"DQO Goals." A formula is provided in EPA's Guidance for Data Quality Assessment (EPA 1996, pp. 3.2-3,
Box 3.2-1) that can be adapted to this example for calculating the number of samples required to meet the FR
and FA requirements:
30
20-
10-
10
20
30
60
Certified PCB Concentration (ppm)
Figure B-l A linear model for predicting L2000DX PCB
concentrations from certified PCB concentrations
with 95% confidence intervals (dashed lines).
N =
(Zl-
FR
(RT - CFA)
(Eq. B-2)
where
N
S2
RT
C
FR
FA
•FA
number of samples from a drum to
be measured
variance for the measurement [e.g.,
S2 = (4.2)2]
regulatory threshold (e.g., RT =
50 ppm)
concentration at which FA is
specified (e.g., CFA = 40 ppm)
false rejection decision error rate
(e.g., FR = 0.05)
false acceptance decision error rate
(e.g., FA = 0.10)
2-
0 10 20 30 40 50 60
Certified PCB Concentration (ppm)
Fisure B-2 Regressi°n °f L2000DX standard deviations
versus certified PCB concentrations.
27
-------
Z!_P = the (I-/?)* percentile of the standard normal distribution (see EPA 1996, Appendix A,
Table A-l). Example: Z(1_fR) = Z095 = 1.645 and Z(l_FA) = Z090 = 1.282.
Incorporating the appropriate values for the Dexsil L2000DX Analyzer into Eq. B-2 gives
__ (4.2)3 (L645 + L282)2 + 2 = 2 86 a 3
(50-40)2
Therefore, three samples from each drum would be analyzed by Dexsil's L2000DX to meet the criteria
established by the DQO process. Note that, to be conservative, one would round the sample size up to the
next integer. These three samples are averaged (by taking the arithmetic mean) to produce an L2000DX value
for a drum's PCB concentration.
Determining the Action Level
Now that the number of samples that need to be analyzed from each drum to meet the DQO goals has been
determined, the action level (AL) can be calculated. The AL is the decision criterion (or "cut-off value) that
will be compared with the unbiased average PCB concentration determined for each drum. The AL for the
decision rule is calculated on the basis of regulation-driven requirements (the TSCA regulatory threshold of
50 ppm) and on the basis of controlling the FR established in the DQO process. Recall that the team set the
permissible FR error rate at 5%.
The formula to compute the AL (EPA 1996) is
c
AL = RT - Zj_^ x — (Eq B.4)
V"
Computing the AL in this instance, we find the following:
42
AL = 50 ppm - (1.645) x _!_ = 46.0 ppm . /ga g 5)
4'
To summarize, three random samples from each drum are analyzed. The three results are averaged to produce
the average PCB concentration for the drum, which is then compared to the AL for the decision rule
(46.0 ppm). Therefore, the decision rule using Dexsil's L2000DX to satisfy a 5% FR and a 10% FA is as
shown in the following box.
28
-------
Decision Rule for FR = 5% and FA = 10%
If the average PCB concentration of three random oil samples from a drum is <46.0 ppm, then
send the drum to the landfill.
If the average PCB concentration of three random oil samples from a drum is >46.0 ppm, then
send the drum to the incinerator.
1.0 -i
The decision performance curve calculates the
probability of sending a drum to the incinerator
for different values of true PCB oil concentration
in a drum (for more information, see EPA 1996,
pp. 34-36). Figure B-3 shows that the decision
performance curve has the value of Pr[ Take
Drum to Incinerator] = 0.95 for True = 50 ppm.
This indicates that the decision rule meets the
DQO team's FR of 5%. The Pr[ Take Drum to
Incinerator] = 0.007 for True = 40 ppm, which is
better than the FA of 10% that the DQO team
had originally specified. This improved
performance is due to rounding up the number of
samples to the next integer in the calculation of
number of samples required.
Alternative FR Parameter
Because of random sampling and analysis error,
there is always some chance that analytical
results will not accurately reflect the true nature
of a decision unit (such as a drum, in this
example). Often, 95% certainty (a 5% FR) is
customary and sufficient to meet stakeholder comfort. But suppose that the DQO team wanted to be even
more cautious about limiting the possibility that a drum might be sent to a landfill when its true value is
50 ppm. If the team wanted to be 99% certain that a drum was correctly sent to a landfill, the following
describes how changing the FR requirement from 5% to 1% would affect the decision rule.
Using FR = 0.01, the sample size is calculated to be 6 and the AL is calculated as 46.0 ppm. (That this
calculation is the same AL as before is coincidental.) The decision performance curve has the value of
Pr[Take Drum to Incinerator] = 0.99 for True = 50 ppm. This indicates that the decision rule meets the DQO
team's FR of 1%. The Pr[ Take Drum to Incinerator] = 0.0002 for True = 40 ppm is better than the FA of
10% that the DQO team had specified. This improved performance is due to rounding up the number of
samples to the next integer in the calculation of number of samples required. The decision rule for the lower
FR would be as shown in the following box.
True PCB Concentration (ppm)
Fi g_3 Decision performance curve for PCB drum
example.
29
-------
Decision Rule for FR = 1% and FA = 10%
If the average PCB concentration of six random oil samples from a drum is <46.0 ppm, then
send the drum to the landfill.
If the average PCB concentration of six random oil samples on a drum is >46.0 ppm, then
send the drum to the incinerator.
Comparison with Reference Laboratory
For purposes of comparison, we subjected the reference laboratory data generated in the ETV verification
study to the same statistical analyses as the L2000DX data. A statistical analysis of the results from the
reference laboratory over the range 0 to 60 ppm gave an estimated standard deviation of SK{ = 3.7 ppm.
Decision rules can be calculated on the basis of this standard deviation. Table B-l compares the decision
rules for the L2000DX Analyzer with those of the reference laboratory. This comparison shows that the
L2000DX analysis and the fixed laboratory method would be comparable in performance for this example.
Table B-l. Comparison of Decision Rule Estimates for L2000DX and Reference Laboratory
Measurements
Analysis
method
L2000DX
Reference
lab
FR = 5% and FA = 10%
N
3
3
AL
(ppm)
46.0
46.5
FR = 1% and FA = 10%
N
6
5
AL
(ppm)
46.0
46.2
Cost per
sample
$5"
$10-20
Turnaround
time
8 samples/h
14-21 days
"Plus $3500 instrument purchase.
30
------- |