vvEPA
EPA/600/R-14/008| January 2014 | www.epa.gov/research
United States
Environmental Protection
Agency
The Verification of a Method for
Detecting and Quantifying
Diethylene Glycol, Triethylene
Glycol, Tetraethylene Glycol,
2-Butoxyethanol and
2-Methoxyethanol in Ground
and Surface Waters
RESEARCH AND DEVELOPMENT
-------
-------
The Verification of a Method for
Detecting and Quantifying Diethylene
Glycol, Triethylene Glycol,
Tetraethylene Glycol,
2-Butoxyethanol and
2-Methoxyethanol in Ground and
Surface Waters
Prepared by
Brian A. Schumacher
U.S. Environmental Protection Agency
National Exposure Research Laboratory Las Vegas, NV 89119
and
Lawrence Zintek
U.S. Environmental Protection Agency
Region 5 Chicago Regional Laboratory
Chicago, IL 60605
U.S. Environmental Protection Agency
Office of Research and Development
Washington, DC 20460
-------
-------
Table of Contents
Table of Contents
List of Figures v
List of Tables vii
Notice ix
The EPA Quality System and the HF Research Study ix
Acknowledgments ix
Executive Summary xi
List of Acronyms xiii
1.0 Introduction 1
1.1 Background 1
1.2 Project Description and Objectives 1
2.0 Method Verification Procedure 3
2.1 Laboratory Initial Demonstration of Capability and Unknown Samples 3
2.2 Blind Sample Description and Spike Concentrations 4
2.3 Statistical Analyses 4
3.0 Quality Assurance 7
3.1 Initial Calibration 7
3.2 Instrument Blank 8
3.3 Laboratory Control Sample/Laboratory Control Sample Duplicate (LCS/LCSD) 8
3.4 Laboratory Fortified Matrix/Laboratory Fortified Matrix Duplicate (MS/MSD) 9
3.5 Laboratory Replicate (Duplicate) 9
3.6 Quality Control Check Standard (QCCS) 9
3.7 Continuing Calibration Verification (CCV) 9
3.8 Holding Times 9
3.9 Audits 9
4.0 Results and Discussion 11
4.1 Tetraethylene Glycol 11
4.2 Triethylene Glycol 16
4.3 Diethylene Glycol 20
4.4 2-Butoxyethanol 24
4.5 2-Methoxyethanol 28
5.0 Summary and Conclusions 33
6.0 Recommendations 35
7.0 References 37
Appendix A 39
in
-------
Table of Contents
IV
-------
List of Figures
List of Figures
Figure 1. Low Concentration (100 |jg/L Spike) Average Recovery for Tetraethylene
Glycol among the Analytical Laboratories. (Closed Circle is Mean
Concentration and Whisker is One Standard Error.) 12
Figure 2. High Concentration (200 |jg/L Spike) Average Recovery for Tetraethylene
Glycol among the Analytical Laboratories. (Closed Circle is Mean
Concentration and Whisker is One Standard Error.) 12
Figure 3. Precision among Low Concentration (100 |jg/L) Sample Recoveries for
Tetraethylene Glycol among the Analytical Laboratories. (Closed Circle is
Mean Concentration and Whisker is 95% Confidence Limits.) 13
Figure 4. Precision among Low Concentration (100 |jg/L) Sample Recoveries for
Tetraethylene Glycol among the Analytical Laboratories Excluding
Laboratory 6 Data. (Closed Circle is Mean Concentration and Whisker is
95% Confidence Limits. The Points with No Bottle Numbers are for High
Concentration Samples.) 14
Figure 5. Low Concentration (100 ug/L) Average Recovery for Tetraethylene Glycol
by Matrix. (Closed Circle is Mean Concentration and Whisker is One
Standard Error.) 15
Figure 6. Low Concentration (100 ug/L) Average Recovery for Tetraethylene Glycol
by Matrix Excluding Laboratory 6 Data. (Closed Circle is Mean
Concentration and Whisker is One Standard Error.) 15
Figure 7. High Concentration (200 ug/L) Average Recovery for Tetraethylene Glycol
by Matrix. (Closed Circle is Mean Concentration and Whisker is One
Standard Error.) 16
Figure 8. Low Concentration (80 ug/L Spike) Average Recovery for Triethylene
Glycol among the Analytical Laboratories. (Closed Circle is Mean
Concentration and Whisker is One Standard Error.) 17
Figure 9. High Concentration (200 ug/L Spike) Average Recovery for Triethylene
Glycol among the Analytical Laboratories. (Closed Circle is Mean
Concentration and Whisker is One Standard Error.) 17
Figure 10. Precision among Low Concentration (80 ug/L) Sample Recoveries for
Triethylene Glycol among the Analytical Laboratories. (Closed Circle is
Mean Concentration and Whisker is 95% Confidence Limits. The Points
with No Bottle Numbers are for High Concentration Samples.) 18
Figure 11. Low Concentration (80 ug/L) Average Recovery for Triethylene Glycol by
Matrix. (Closed Circle is Mean Concentration and Whisker is One Standard
Error.) 19
Figure 12. High Concentration (200 ug/L) Average Recovery for Triethylene Glycol by
Matrix. (Closed Circle is Mean Concentration and Whisker is One Standard
Error.) 19
Figure 13. Low Concentration (10 ug/L Spike) Average Recovery for Diethylene Glycol
among the Analytical Laboratories. (Closed Circle is Mean Concentration
and Whisker is One Standard Error.) 21
-------
List of Figures
Figure 14. High Concentration (100 |jg/L Spike) Average Recovery for Diethylene
Glycol among the Analytical Laboratories. (Closed Circle is Mean
Concentration and Whisker is One Standard Error.) 21
Figure 15. Precision among Low Concentration (10 |jg/L) Sample Recoveries for
Diethylene Glycol among the Analytical Laboratories. (Closed Circle is
Mean Concentration and Whisker is 95% Confidence Limits.) 22
Figure 16. Low Concentration (10 |jg/L) Average Recovery for Diethylene Glycol by
Matrix. (Closed Circle is Mean Concentration and Whisker is One Standard
Error.) 23
Figure 17. High Concentration (100 |jg/L) Average Recovery for Diethylene Glycol by
Matrix. (Closed Circle is Mean Concentration and Whisker is One Standard
Error.) 23
Figure 18. Low Concentration (60 |jg/L Spike) Average Recovery for 2-Butoxyethanol
among the Analytical Laboratories. (Closed Circle is Mean Concentration
and Whisker is One Standard Error.) 24
Figure 19. High Concentration (180 |jg/L Spike) Average Recovery for 2-Butoxyethanol
among the Analytical Laboratories. (Closed Circle is Mean Concentration and
Whisker is One Standard Error.) 25
Figure 20. Precision among Low Concentration (60 |jg/L) Sample Recoveries for 2-
Butoxyethanol among the Analytical Laboratories. (Closed Circle is Mean
Concentration and Whisker is 95% Confidence Limits. The Points with No
Bottle Numbers are for High Concentration Samples.) 26
Figure 21. Low Concentration (60 ug/L) Average Recovery for 2-Butoxyethanol by
Matrix. (Closed Circle is Mean Concentration and Whisker is One Standard
Error.) 26
Figure 22. High Concentration (180 ug/L) Average Recovery for 2-Butoxyethanol by
Matrix. (Closed Circle is Mean Concentration and Whisker is One Standard
Error.) 27
Figure 23. Low Concentration (40 ug/L Spike) Average Recovery for 2-Methoxyethanol
among the Analytical Laboratories. (Closed Circle is Mean Concentration and
Whisker is One standard error.) 29
Figure 24. High Concentration (100 ug/L Spike) Average Recovery for 2-Methoxyethanol
among the Analytical Laboratories. (Closed Circle is Mean Concentration and
Whisker is One Standard Error.) 29
Figure 25. Precision among Low Concentration (40 ug/L) Sample Recoveries for 2-
Methoxyethanol among the Analytical Laboratories. (Closed Circle is Mean
Concentration and Whisker is 95% Confidence Limits. The Points with No
Bottle Numbers are for High Concentration Samples.) 30
Figure 26. Low Concentration (40 ug/L) Average Recovery for 2-Methoxyethanol by
Matrix. (Closed Circle is Mean Concentration and Whisker is One Standard
Error.) 31
Figure 27. High Concentration (100 ug/L) Average Recovery for 2-Methoxyethanol by
Matrix. (Closed Circle is Mean Concentration and Whisker is One Standard
Error.) 31
VI
-------
List of Tables
List of Tables
Table 1. Laboratory Reporting Limits 3
Table 2. Source Waters 4
Table 3. Concentrations of Analytes in Unknown Sample Bottles 4
Table 4. Data Quality Indicators and their Acceptance Criteria 7
vn
-------
List of Tables
Vlll
-------
Notice
The information in this document has been funded wholly by the United States Environmental
Protection Agency. It has been subjected to the Agency's peer and administrative review 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.
The EPA Quality System and the
Hydraulic Fracturing Research Study
EPA requires that all data collected for the characterization of environmental processes and
conditions are of the appropriate type and quality for their intended use. This is accomplished through an
Agency-wide quality system for environmental data. Components of the EPA quality system can be
found at http://www.epa.gov/quality. EPA policy is based on the national consensus standard ANSI/ASQ
E4-2004 Quality Systems for Environmental Data and Technology Programs: Requirements with
Guidance for Use. This standard recommends a tiered approach that includes the development and use
of Quality Management Plans (QMPs). The organizational units in EPA that generate and/or use
environmental data are required to have Agency-approved QMPs. Programmatic QMPs are also written
when program managers and their QA staff decide a program is of sufficient complexity to benefit from a
QMP, as was done for the study of the potential impacts of hydraulic fracturing (HF) on drinking water
resources. The HF QMP describes the program's organizational structure, defines and assigns quality
assurance (QA) and quality control (QC) responsibilities, and describes the processes and procedures
used to plan, implement, and assess the effectiveness of the quality system. The HF QMP is then
supported by project-specific QA project plans (QAPPs). The QAPPs provide the technical details and
associated QA/QC procedures for the research projects that address questions posed by EPA about the HF
water cycle and as described in the Plan to Study the Potential Impacts of Hydraulic Fracturing on
Drinking Water Resources (EPA/600/R-ll/122/November 2011; www.epa.gov/hfstudv). The results of
the research projects will provide the foundation for EPA's study report.
Acknowledgments
The authors would also like to thank the following for their valuable input to the project:
• The Quality Assurance Group: Office of Research and Development/National Exposure Research
Laboratory (ORD/NERL) - George Brilis, Michelle Henderson, and Margie Vazquez; Region 3 - Jill
Bilyeu; EPA Region 5 - Angela Ockrassa; Eurofms Lancaster Laboratories - Dorothy Love; Test
America - Teresa Williams; Philadelphia Water District - Robert Eppinger.
• Sample Preparation Group: ORD/NERL - Lantis Osemwengie, Jade Morgan, Don Betowski,
Wallace Atterberry, and Amanda DiGoregorio.
• Method Testing Group: ORD/NERL - Patrick DeArmond and Jody Shoemaker; EPA Region 3 -
Jennifer Gundersen; Eurofms Lancaster - Charles Neslund; Test America Inc - Charlie Carter;
Philadelphia Water District - Earl Peterkin; Metropolitan District of Southern California - Rich Yates.
• Data Verification and Data Analysis Group: ORD/NERL - Patrick DeArmond and Maliha Nash.
IX
-------
-------
Executive Summary
This verification study was a special project designed to determine the efficacy of a draft standard
operating procedure (SOP) developed by US EPA Region 3 for the determination of selected glycols and
glycol ethers in drinking waters that may have been impacted by active unconventional oil and gas
operations utilizing hydraulic fracturing (HF) extraction. HF has become increasingly prevalent as a
method of extracting energy resources from "unconventional" reservoirs, such as coalbeds, shales, and
tight sands. Concerns have been raised about the potential for hydraulic fracturing fluid chemical
additives to enter ground water aquifers that, in turn, may be used as drinking water sources.
One group of hydraulic fracturing fluid chemical additives that concerns have been raised about
includes the additives: 2-methoxyethanol (2-ME), 2-butoxyethanol (2-BE), diethylene glycol (Di-EG),
triethylene glycol (Tri-EG), and tetraethylene glycol (Tetra-EG). The primary objective of this study was
to verify the performance of the draft standard operating procedure developed by US EPA Region 3 in
multiple laboratories. This study verified a simple and rapid high performance-liquid
chromatography/tandem mass spectrometry (HPLC/MS/MS) method for the quantitation of these five
chemical additives in aqueous samples. The draft method was quick, required little to no sample
preparation, and utilized the sensitivity that HPLC/MS/MS provides.
The verification of the draft SOP included volunteer federal, state, municipal, and commercial
analytical laboratories. Each laboratory tested the efficacy of the draft SOP using the HPLC/MS/MS
instrumentation present in their laboratories. Four different water matrices were used to spike batches of
samples at various concentrations. Three source matrix waters were collected from bulk water samples
acquired from three drinking water source wells (prior to treatment) around the country in areas where
active shale oil and gas operations are occurring or where they may occur in the future. The source
matrix waters were collected at Avella, Pennsylvania, Raleigh, North Carolina, and Ada, Oklahoma.
Laboratory deionized water, from Las Vegas, Nevada, was used as a fourth matrix. Batches of 36 blind
samples, prepared by EPA, were distributed to the volunteer laboratories for analysis following the draft
SOP.
To ensure that data of known and documentable quality are generated by the participating
analytical laboratories, data quality indicators (DQIs) were defined to examine key parameters and to
determine if the key parameters met their acceptance criterion. The key parameters included: verification
of the calibration curves, determination of any laboratory blank contamination issues, examination of the
precision and accuracy of the laboratory control and matrix spike samples, substantiation of sample
precision from duplicate samples, second source check standard verification, confirmation of reporting
limits and appropriate method detection limits, and continuing calibration verification after each batch of
samples.
To determine if the draft glycol SOP could be followed and meet the performance criteria, blind
samples submitted to the analytical laboratories and several key factors were examined. The key factors
included: accuracy (defined as the difference between the known and measured concentration) within and
among the laboratories, precision within the analytical laboratory, and investigating whether matrix
effects from the four waters used were present.
The data generated by the analytical laboratories following the draft SOP were statistically
analyzed to determine if differences existed among the laboratories as related to key factors used to
determine the analytical performance of the draft SOP. Accuracy, determined by comparing the
measured result to the known spiked concentration, met the performance criteria but a few statistical
outliers were identified. The precision within the analytical laboratories met the performance criteria
indicating that reproducible results were being generated. Matrix effects between the four different water
XI
-------
matrices were not identified in any of the analytical laboratories for any of the compounds of interest,
indicating that the method could produce the same results in the four water matrices tested.
The reporting limits and calibration ranges of the draft SOP were similar among different
instruments used at the analytical laboratories with reporting limits typically ranging between 5 and 10
(ig/L. However, a few discrepancies were noted. Some laboratories had better sensitivity on their
instruments, wider calibration ranges; and had different optimum calibration fits (i.e., a linear calibration
fit was specified in the draft SOP but a quadratic fit of the calibration point data yielded better results).
Differences in instrument sensitivity were found where 2-BE could not be successfully detected at one
laboratory and two laboratories could not successfully detect 2-ME.
Overall, the draft glycol SOP presented a method that was accurate and precise by meeting the
established performance criterion nearly all the time. No matrix effects on the chemical recoveries were
exhibited for the four waters tested when the compounds of interest were detectable. The multi-
laboratory verification of the draft SOP resulted in the generation of several recommendations in order to
construct an improved analytical method including: allowing best calibration fit parameters,
incorporating surrogate spikes, conducting a sample preservation and holding time study, conducting a
filtering unit study, allowing for greater adjustment of chromatographic conditions, initiating second
source verification, and applying calibration check verification concentration consistency.
xn
-------
2-BE
2-ME
CCV
CV
Di-EG
DQI
FS
GC
HF
HPLC
HPLC/MS/MS
LC/MS/MS
LCS/LCSD
MDL
MS
MS/MS
MS/MSD
N/A
NERL
ORD
QA
QA/QC
QC
QAPP
QCCS
QMP
RPD
RSD
SOP
Tetra-EG
Tri-EG
US EPA
List of Acronyms
2-Butoxyethanol
2-Methoxyethanol
Continuing Calibration Verification
Coefficient of Variation
Diethylene Glycol
Data Quality Indicator
Field Sample
Gas Chromatography
Hydraulic Fracturing
High Performance Liquid Chromatography
High Performance Liquid Chromatography/Mass Spectrometry/Mass
Spectrometry
Liquid Chromatography/Mass Spectrometry/Mass Spectrometry
Laboratory Control Sample/ Laboratory Control Sample Duplicate
Method Detection Limit
Mass Spectrometry
Mass Spectrometry/Mass Spectrometry
Matrix Spike/Matrix Spike Duplicate
Not Applicable
National Exposure Research Laboratory
Office of Research and Development
Quality Assurance
Quality Assurance/Quality Control
Quality Control
Quality Assurance Project Plan
Quality Control Check Sample
Quality Management Plan
Relative Percent Difference
Relative Standard Deviation
Standard Operating Procedure
Tetraethylene Glycol
Triethylene Glycol
United States Environmental Protection Agency
xni
-------
XIV
-------
1.0 Introduction
Glycols and glycol ethers are solvents and chemical intermediates commonly used during the
production of many resins, plasticizers, adhesives, surfactants, and cosmetics. Due to their useful
properties, many glycols and glycol ethers, including 2-methoxyethanol (2-ME), 2-butoxyethanol (2-BE),
diethylene glycol (Di-EG), triethylene glycol (Tri-EG), and tetraethylene glycol (Tetra-EG), have been
classified as high-production volume chemicals by the United States Environmental Protection Agency
(US EPA)1. Additionally, these compounds have frequently been used during oil and gas production. For
example, ethylene glycol, Di-EG, Tri-EG, and Tetra-EG are commonly used during the dehydration
processes of natural gas2, and glycol ethers are used as foaming agents and in breaker fluids during
hydraulic fracturing3.
1.1 Background
Hydraulic fracturing (HF) has become increasingly prevalent as a method of extracting energy
resources from "unconventional" reservoirs, such as coalbeds, shales, and tight sands. Concerns have
been raised about the potential for hydraulic fracturing fluid chemical additives to enter ground water
aquifers that, in turn, may be used as drinking water sources. Of concern for this project are diethylene
glycol (CASRN #111-46-6), triethylene glycol (CASRN #112-27-6), tetraethylene glycol (CASRN #112-
60-7), 2-butoxyethanol (CASRN #111-76-2), and 2-methoxyethanol (CASRN #109-86-4). In response to
this concern, the US EPA Region 3 Environmental Science Center in Fort Meade, Maryland developed a
quick, draft method for the determination and quantification of these compounds. This draft method,
prepared in the form of a standard operating procedure (SOP; Appendix A), needed to be verified to
determine its efficacy in determining these compounds in laboratory and various drinking water matrices.
1.2 Project Description and Objectives
The Multi-laboratory Verification of Diethylene Glycol, Triethylene Glycol, Tetraethylene Glycol,
2- Butoxyethanol and 2-Methoxyethanol in Ground and Surface Waters by Liquid
Chromatography/Tandem Mass Spectrometry study was designed to determine the efficacy of a draft
method developed by US EPA Region 3 for the determination of glycols and glycol ethers in drinking
waters collected from drinking water wells. The objectives of this study were to verify a simple and rapid
liquid chromatography/tandem mass spectrometry (LC/MS/MS) method for the quantitation of 2-ME, 2-
BE, Di-EG, Tri-EG, and Tetra-EG in aqueous samples by examining three key performance areas: (1)
can the analytical laboratory follow the SOP, (2) can the analytical laboratory meet the SOP requirements
as related to quality assurance/quality control measures, and (3) how well did the method perform in
terms of bias/accuracy, precision, and the absence of matrix effects. This report describes the
interlaboratory verification results for the described method.
For the verification of the method, eight analytical laboratories were invited to participate in the
analyses of a series of multiple blind samples (spiked and unspiked) in multiple matrices (e.g., laboratory
waters and drinking well waters).
The following laboratories were invited to participate:
1. US EPA National Exposure Research Laboratory, Environmental Sciences Division, Las Vegas, NV,
2. US EPA National Exposure Research Laboratory, Microbiological & Chemical Exposure Assessment
Research Division, Cincinnati, OH,
3. US EPA Region 3 Environmental Science Center, Fort Meade, MD,
-------
4. US EPA Region 5 Chicago Regional Laboratory, Chicago, IL,
5. Eurofins Lancaster Testing Laboratories, Lancaster, PA,
6. TestAmerica, Inc, Arvada, CO,
7. Philadelphia Water Department, Philadelphia, PA, and
8. Metropolitan Water District of Southern California, La Verne, CA.
Two laboratories had instrument issues or business time constraints and were unable to participate. All
participating laboratories did so a gratis.
To ensure that these study objectives were met, all participating laboratories strictly adhered to
the requirements that:
• Each laboratory verified and optimized the liquid chromatography/mass spectrometry/mass
spectrometry conditions in sections 10 and 11 of the draft SOP (Appendix A) on their
instrumentation and determined the reporting limits on their LC/MS/MS systems.
• Each laboratory followed all analytical and quality control procedures in the approved quality
assurance project plan (QAPP)4.
• Each laboratory documented any deviations from the SOP or QAPP.
• All data produced were capable of being verified by an independent person reviewing the
analytical data package.
• Each laboratory had a verifiable QA program, equal to or exceeding EPA requirements, in
placeand operating throughout the study to ensure that the data produced are of appropriate and
documented quality.
-------
2.0 Method Verification Procedure
2.1 Laboratory Initial Demonstration of Capability and Unknown Samples
For the verification study, each participating laboratory was sent a copy of the draft SOP as
Appendix A of the QAPP. The conditions in the SOP were to be used as a starting point in order to
optimize each LC/MS/MS instrument within the limits identified in the SOP. The goal of optimization
was to: familiarize the analyst with the analytes and method, determine the range of the calibration
curves, and determine the method detection and reporting limits on their instrument. At least seven
replicates at a low level were used in order to determine a method detection limit (MDL) for each analyte
in each laboratory (40 CFR Part 136 Appendix B).
Once optimized, the reporting limit was calculated to be at least 3 times the MDL and may be different
between the laboratories because of varying sensitivities of the LC/MS/MS systems used. A reporting
limit target of 5 |lg/L was established for this study. Each laboratory determined their reporting limits
(Table 1). If the actual reporting limit, as calculated using the MDL, was determined to be less than the
reporting limit target at the laboratory, the target 5 |ig/L reporting limit was used. If the determined
reporting limit was greater than the target reporting limit of 5 |lg/L, the new value was reported and used
for this study. All concentrations reported were required to be at or above the reporting limit for
statistical analysis.
Table 1. Laboratory Reporting Limits.
Analyte
2-ME
2-BE
Di-EG
Tri-EG
Tetra-EG
Laboratory 1
Non-Detect
Non-Detect
5^g/L
5^g/L
5|ig/L
Laboratory 2
5^g/L
5[lg/L
5^g/L
5^g/L
5|ig/L
Laboratory 3
Non-Detect
lO^ig/L
5^g/L
5^g/L
5|ig/L
Laboratory 5
25 ng/L
10 ng/L
5^g/L
5[lg/L
5^g/L
Laboratory 6
20 ng/L
8^g/L
8^g/L
8M£/L
8^g/L
LaboratoryS
5|ig/L
5^g/L
5|ig/L
5^g/L
5^g/L
* Non-Detect indicates that the analytical laboratory could not detect this compound.
For the verification study, four water matrix sets of nine samples each were prepared for a total of
36 blind samples. Samples were prepared by an independent scientist (i.e., one not involved with the
draft glycol method verification study). The various blind samples were prepared from bulk water
samples acquired from multiple drinking water source wells around the country in areas where active
shale oil and gas operations are occurring or may occur in the future. Several gallons of each bulk water
matrix were collected in clean, capped amber glass containers and labeled with the source and date of
sampling. The matrix waters were collected from the drinking water system prior to any treatment at the
source. Bulk samples were stored at 4 °C ± 2 °C. The matrix waters were collected at Avella,
Pennsylvania, Raleigh, North Carolina, and Ada, Oklahoma. Laboratory deionized water, from Las
Vegas, Nevada, was used as a fourth matrix. The sample identifiers, FS-1 through FS-4, were utilized
throughout this report to reference the different sample matrices (Table 2).
-------
Table 2. Source Waters.
Sample Identifier*
FS-1
FS-2
FS-3
FS-4
Source
Avella, PA Drinking Well Water
Raleigh, NC Well Water
Ada, OK Ground Water
EPA-Las Vegas, NV Deionized Laboratory
Water
* FS = Field Sample.
2.2 Blind Sample Description and Spike Concentrations
The blind spiked samples were produced in bulk such that all equivalent samples came from the
same volumetric flask, and were bottled and shipped to the participating analytical laboratories on the
same day to ensure that each participating laboratory received the same samples. Each analytical
laboratory received 9 samples per matrix. Seven samples were spiked and 2 samples (bottles 8 and 9)
were blank (non-spiked source water). Of the seven spiked bottles (bottles 1-7), five bottles had low-mid
concentration of each target analyte, one bottle had a high concentration of each target analyte, and one
bottle was not spiked with one of the analyte s (Table 3). Actual concentrations of the low-mid and high
concentration samples were varied among the analytes in an effort to avoid pattern recognition among the
laboratories (i.e., all concentrations for all analytes were the same in all samples). Additionally, the one
sample that was not spiked with one of the target analytes was used to ensure that the laboratory was
confident in their analyses and reported a non-detect even though the remaining four analytes were
present in the sample.
Table 3. Concentrations of Analytes in Unknown Sample Bottles.
Bottles
1
2
3
4
5
6
7
8
9
Di-EG (llg/L)
10
100
10
10
Blank
10
10
Blank
Blank
Tri-EG (jlg/L)
80
80
80
80
200
80
Blank
Blank
Blank
Tetra-EG Qig/L)
100
100
200
100
100
Blank
100
Blank
Blank
2-BE (jlg/L)
60
60
Blank
180
60
60
60
Blank
Blank
2-ME (jlg/L)
40
40
40
Blank
40
100
40
Blank
Blank
2.3 Statistical Analyses
The total number measurement values generated was 840 (6 laboratories x 5 analyte x 4 sample x
7 bottle) per each type of calibration curve (i.e., linear or quadratic curves) used. Values less than or
equal to the reporting limit (RL) were excluded prior to statistical analyses. The concentrations that are
below reporting limit do not reflect precise or accurate measurements. However, the removal of the data
points and the resulting unbalanced design can be overcome by applying the general linear model (GLM).
Additionally, not all laboratories were able to provide measurements on all analytes because of the
detecting capability on their analytical instruments. This limitation is described within each section for the
analyte. A split plot design ANOVA5 was assembled for this data set as:
-------
Measurement = n + Lab+SW + Lab * SW + Spike + Spike * Lab + Spike * SW + e Eqn 1
where: /j, is the overall mean of the observations; Lab, SW, and Spike are class variables; Lab for
analytical laboratory, .SWfor source waters sample, and Spike is unknown concentration of the compound.
The error for SWisLab*SW and others were tested against the residual errors (mean square error or e).
The above model was used to statistically analyze all analytes except for diethylene glycol, which
was determined using the linear calibration curve for the low concentration samples. For diethylene
glycol, Laboratory 6 did not report any data for the low level spiked samples; hence, spike level is no
longer a class variable. With only one spike level, a randomized block design for ANOVA was applied as
presented in Eqn 2:
Measurement = n + Lab + SW + e Eqn 2
Reliability of the study can be measured by the coefficient of variability (CV), which is the
overall experimental error standardized by the overall mean (grand mean) of all measurements (Eqn 3):
CV(%) = * 100
granamean
where ^JMSE is the square root of mean square error. The value of CV can be used as an index for model
reliability. Model reliability increases as the value of CV decreases. If CV > 30%, then caution has to be
taken when describing model reliability and model output6. All of our models exhibited low values of
CV, where they ranged from 1.7 to 12.7.
Diagnostic checking on residuals was carried on for each model with the outliers removed.
Model residuals were tested for normality where probability of Shapiro-Wilk test was > 0.05 for all
models. Means, standard error, and the 95% confidence limit for each class level, or their combinations,
are presented in figures to explain statistical differences.
The statistical analyses were performed using a general linear model (Proc GLM) in SAS® with
the least-square means (LSMEANS) option to account for the missing values in the unbalanced design.
The probability of t-statistics was used for a multiple comparison of means between class variables and
their combinations. Mean and standard error values in figures with ±20% and ±30% thresholds were
determined using Proc Means in SAS®. The significance level was 0.05 for all statistical analyses.
-------
-------
3.0 Quality Assurance
The QAPP entitled, "Quality Assurance Project Plan for the Multi-Laboratory Verification of
Diethylene Glycol, Triethylene Glycol, Tetraethylene Glycol, 2-Butoxyethanol and 2-Methoxyethanol in
Ground and Surface Waters by Liquid Chromatography/Tandem Mass Spectrometry", was approved on
March 5, 2013. The Data Quality Indicators (DQIs) and their acceptance criteria for the measurement of
the data generated by the laboratories consisted of seven key parameters that were requested during the
verification study (Table 4). Some of the DQI key parameters did not meet the acceptance criteria in the
QAPP and these deviations are documented and explained in this section. The data generated from the
blind samples with quality assurance/quality control (QA/QC) resulted in variances that appeared to be
random with the exception of negative biases in the initial calibration curves (see section 3.1). Where the
variances occurred, the majority of the associated QA/QC acceptance parameters were met so the data
were used to assess the method performance.
Table 4. Data Quality Indicators and their Acceptance Criteria.
QC Check
5 -Point Initial Calibration
Instrument Blank
Laboratory Control Sample/ Laboratory
Control Sample Duplicate (LCS/LCSD)
Laboratory Matrix Spike/Laboratory
Matrix Spike Duplicate (MS/MSD)
Laboratory Replicate
Quality Control Check Standard (QCCS)
Continuing Calibration Verification
Sample (CCV)
Precision*
N/A
N/A
RPD < 30%
RPD < 30%
RPD < 30%
N/A
N/A
Accuracy
Correlation Coefficient
r2 > 0.99
< Reporting Limit
±30% of Known Value
Recovery Between 70 and 130%
of Spike Concentration
N/A
±20% of Known Value
±30% of Known Value
* RPD = relative percent difference; N/A = not applicable.
3.1 Initial Calibration
Linear initial calibration curves were required to be determined for each chemical. Although
not specified in the draft SOP, quadratic calibration curves were also requested since: (a) the allowance to
use both forms of initial calibration is being incorporated into the revised SOP, and (b) fitting the
calibration points to a quadratic calibration curve is a simple, non-time consuming process for the
analytical laboratories. In most cases, using the quadratic calibration curve reduced/removed most of the
low calibration point(s) biases. The initial calibrations met the acceptance criteria of having a r2 value >
0.99 for both linear and quadratic fits in most cases.
Laboratory 6 did not meet the correlation coefficient parameter for 2-BE (r2 = 0.979). A
negative bias for 2-BE at the low concentration level (-144% at the 5 (ig/L concentration) in the
calibration curve using the linear calibration fit was observed. The remainder of the calibration curve,
whether linear or quadratic, displayed a range in positive and negative biases (-45 to 77%) among the
calibration points. This broad range in biases is believed to be the cause for the failure of Laboratory 6 to
meet the r2 > 0.99 criterion for 2-BE.
-------
Similar to the negative bias seen in Laboratory 6 for 2-BE, Laboratory 6 also had negative
biases (-75 to -267%) at the lower concentration levels (5 and 10 (ig/L) in the linear calibration curves for
Tetra-EG, Tri-EG and Di-EG. These biases were not as pronounced in the quadratic calibration curve for
these compounds. Laboratory 6 had a negative bias (-74 to -104%) in both the linear and quadratic
calibration fits for 2-ME at the lower concentration levels (5 and 10 (ig/L) forcing the low level
calibration standard to be dropped and the reporting limit to be raised to 20 |ig/L (Table 1).
The linear calibration curves in Laboratory 1 met the correlation coefficient requirement;
however, Tri-EG and Tetra-EG had negative biases (-67 and -51%, respectively) at the 5 (ig/L
concentration standard. The calibration curves for Di-EG, Tri-EG and Tetra-EG were established from 5-
100 (ig/L requiring dilution of the samples that were over the calibration curve. All reported Laboratory 1
results were used in this verification study and were in statistical agreement with the participant
laboratory's data.
Laboratory 5 linear calibration curves all met the correlation coefficient requirement but 4 out
of 5 of the target analytes had negative biases near the reporting limit. The Di-EG calibration curve did
not display a negative bias. The 5 and 10 (ig/L calibration standards were not incorporated in the linear
calibration curve due to very strong negative bias for 2-ME resulting in a raised reporting limit to 25 (ig/L
(Table 1). The Tetra-EG higher concentration standards of the calibration curve were not used due to
negative bias resulting in an abbreviated calibration curve from 5-100 (ig/L. The abbreviated calibration
curve; however, still contained the -62% negative bias near the reporting limit (5 (ig/L). Some results
were reported above the calibration curve and they were considered to be semi-quantitative. The results
were used in this verification study as they were in statistical agreement with the other participating
laboratories data.
Linear calibration curves for Laboratories 2, 3, and 8 met the correlation coefficient
requirement; however, some negative biases were seen near the reporting limit. In Laboratory 2, Tri-EG
and Tetra-EG had negative biases (-65 and -51%, respectively) at the reporting limit (5 (ig/L). In
Laboratory 3, Di-EG, Tri-EG and Tetra-EG had negative biases (-45, -60 and -106%, respectively) at
reporting limit (5 (ig/L) while in Laboratory 8, a -34% bias at 5 (ig/L was identified for Tetra-EG.
In two instances, the initial calibration was not done by the laboratories because the analytical
laboratories were not able to ionize the analytes. No calibration curves were submitted for 2-ME and 2-
BE by either Laboratory 1 nor for 2-ME by Laboratory 3.
3.2 Instrument Blank
The instrument blank results were acceptable in all cases except for Laboratory 6 which initially
had blank contamination issues with 2-BE. Laboratory 6 used an isocratic gradient which is a deviation
from the draft SOP. The generated Laboratory 6 data were accepted and included in the statistical data
evaluation.
3.3 Laboratory Control Sample/Laboratory Control Sample Duplicate
(LCS/LCSD)
Overall, the LCS data were acceptable for all the laboratories except for Laboratory 6.
Laboratory 6 had a few random exceedances of the acceptance criteria where the LCS determined value
was biased high for 2-ME and biased low for Di-EG and the LCSD was biased low for 2-ME.
-------
3.4 Laboratory Fortified Matrix/Laboratory Fortified Matrix Duplicate (MS/MSD)
The MS/MSD data were generally acceptable. The laboratories were required to decide which
samples to use for MS/MSD samples. Laboratory 2 had a positive bias for Di-EG with acceptable RPDs
between the MS and MSD. Laboratory 1 had randomly positive and negative biases for 11 out of 14
MS/MSD samples for Di-EG, Tri-EG, and Tetra-EG with acceptable RPDs between the replicates.
Laboratory 6 had: low recoveries for Di-EG with an acceptable RPD between the duplicate samples; one
biased high recovery with an exceeded RPD for 2-BE; all samples biased high with acceptable RPDs for
2-ME and Tetra-EG; and one MS with low recovery that exceeded the RPD acceptance limit for Tri-EG.
3.5 Laboratory Replicate (Duplicate)
Laboratory duplicates were not analyzed by Laboratories 5 and 6 so the MS/MSD RPDs were
used to determine duplicate reproducibility in these cases with the identified discrepancies being
explained in Section 3.4. All other duplicate data met the <30% RPD acceptance criterion.
3.6 Quality Control Check Standard (QCCS)
The laboratories chose a QCCS standard to analyze with each of their batches. The QCCS was
either purchased as a prepared diluted standard/mix or purchased as a neat material. (This is not a
comparison study of analytical standards so no sources of standards are mentioned.) Laboratory 6 did not
analyze a QCCS sample. Laboratory 2 had a slight positive bias for Di-EG while Laboratory 3 had slight
positive bias for Tetra-EG and Tri-EG. Laboratory 5 had slight positive bias for Di-EG, Tri-EG, and
Tetra-EG.
3.7 Continuing Calibration Verification (CCV)
The CCV check samples were generally within acceptance criterion with exceptions at two of the
participating analytical laboratories. At Laboratory 5, the 2-ME CCV checks exceeded the acceptance
limit for 2 out of 4 check samples with a 45% recovery of the 25(ig/L CCV sample and a 132% recovery
of the 50 (ig/L CCV sample. Laboratory 6 CCV check samples were negatively biased for Di-EG in 4 out
of 7 check samples; positively biased for 2-BE and 2-ME in 4 out of 7 check samples; and positively
biased for Tetra-EG in 3 out of 7 check samples.
3.8 Holding Times
There are no preservation or holding time studies of these analytes in reagent water or the
matrices of concern. A fourteen day holding time was used for this study. Laboratory 1 initiated the
analysis of the samples on day 14 and had a few samples that were analyzed after that due to an
instrument failure. Laboratories 3 and 6 performed their analysis on day 14 with a few samples being run
early on day 15 at Lab 3. Laboratories 2, 5, and 8 performed their analyses within 6-8 days, 2-3 days, and
2 days, respectively. With no formal holding times being established, all the data were used based on this
criterion.
3.9 Audits
Each participating laboratory was asked to perform a readiness review, surveillance audit, and
audit of data quality during the study. Reports were submitted from Laboratories 2, 3, 5, 6, and 8.
Laboratory 1 did not provide the requested audit reports. In each case, no major findings were identified.
Discrepancies, when noted, have been identified in this section of the report. In all cases, discrepancies
were deemed to not have a major effect on the resultant data due to the passage of a majority of the
quality assurance/quality control measures; therefore, all data was used during the statistical analysis (see
section 2.4) of the results.
-------
In addition to these three audits, Laboratories 3 and 8 performed additional surveillance
audits/quality assurance inspections and technical systems audits. During one of the surveillance
audits/quality assurance inspections, 2 blind sample bottles were mislabeled. Results from these samples
were removed from statistical consideration. A comprehensive quality perspective (i.e., overview) was
performed at Laboratory 8 over several months of this study with no findings being identified.
This report was reviewed by the NERL Director of Quality Assurance according to the
requirements of the QAPP, "Quality Assurance Project Plan for the Multi-Laboratory Verification of
Diethylene Glycol, Triethylene Glycol, Tetraethylene Glycol, 2-Butoxyethanol and 2-Methoxyethanol in
Ground and Surface Waters by Liquid Chromatography/Tandem Mass Spectrometry" (approved on
March 5, 2013) and deemed acceptable on January 31, 2014.
10
-------
4.0 Results and Discussion
To determine the performance characteristics of the draft glycol method, several key factors
needed to be examined. The key factors included: (a) examining the accuracy (defined as the difference
between the known and measured concentration) within and among the laboratories, (b) determining the
precision within the analytical laboratory, and (c) investigating whether matrix effects from the four
waters used were present.
When examining the accuracy within and among the laboratories, the ±20% and ±30% of the
known concentration lines will be plotted in the appropriate figures. These two percentages are
commonly cited, acceptance criteria and are the two acceptance criteria used for accuracy determinations
in this study (Table 4). The means of the data was examined first to see if it passed the ±20% criterion
which would indicate a high degree of accuracy. If the means of the data do not meet the ±20% criterion,
then they should meet the ±30% criterion to be fully acceptable.
The laboratory numbers do not coincide with the laboratory list in Section 1.2; this was done to
provide anonymity to the participating analytical laboratories.
The following sections and figures discuss the results generated from the linear calibration
curves. If differences in the statistical interpretation of the results (i.e., acceptance vs failure to meet the
acceptance criterion) occurred when the data were generated from the quadratic calibration curves, these
differences, and how they affected our interpretation of the data, are clearly delineated in the text.
4.1 Tetraethylene Glycol
All six participating laboratories provided data. All submitted blind blank samples (see Table 3)
showed no target analytes at or above the reporting limits of 5 or 8 (ig/L depending on the sensitivity of
the instrument at the participating analytical laboratory. The blind samples were spiked at 100 or 200
jig/L.
The accuracy among the analytical laboratories was within 20% of the known concentrations,
except for Laboratory 6 for the low level spiked samples which were biased high (Figures 1 and 2). The
results for any given laboratory included all samples, regardless of matrix, at the given concentration
level.
The precision within the laboratory can be determined by examining the results of the individual
bottles shipped to the analytical laboratory. In each case, the bottle number (represented as the last digit
on the x-axis identifiers) represents the mean of the four different water matrices for that bottle number
(Figure 3). By checking for significant differences among the 5 bottles spiked at the 100 (ig/L
concentration, the precision within an analytical laboratory can be determined. In general, precision met
the performance criteria among all the samples spiked at the same concentrations with a few exceptions.
For example, Laboratories 1 and 2 showed statistical differences between bottles 1 and 6 (Figure 4).
These exceptions; however, do not affect our interpretation of the study results as all of the bottles
determined concentrations were within the study's accuracy acceptance criterion.
11
-------
415
Figure 1.
Figure 2.
365
5 315
3
§265
215
0)
o
165
115 «
* •
• ----------------------------------- 20%
65 ------------------------------------ $$%
Labi Lab2 Lab3 Lab5 Lab6 Lab8
Low Concentration (100 |jg/L Spike) Average Recovery for Tetraethylene Glycol among the
Analytical Laboratories. (Closed Circle is Mean Concentration and Whisker is One Standard
Error.)
270
----------------------------------- 30%
250
--------------------------- ------- 20%
230
I210 * » _
| 190
* i *
£ 170 * T
0) •
o • ---------------------------------- 20%
u 15°
----------------------------------- 30%
130
110
Labi
Lab2
Lab3
Lab5
Lab6
Lab8
High Concentration (200 ug/L Spike) Average Recovery for Tetraethylene Glycol among the
Analytical Laboratories. (Closed Circle is Mean Concentration and Whisker is One Standard
Error.)
12
-------
400
350
300
CT
S£ 250
c.
'ro 20°
4-J
OJ
« 150
o
u
100
50
0
124561245612456124561245612456
Labi Lab2 Lab3 Lab5 Lab6 Lab8
Lab Bottle Number
Figure 3. Precision among Low Concentration (100 ug/L) Sample Recoveries for Tetraethylene Glycol
among the Analytical Laboratories. (Closed Circle is Mean Concentration and Whisker is 95%
Confidence Limits.)
13
-------
124
119 j
114 I
B>109 I T j
^104 * * I
o 1
co aa
§ 94 T
§ 89 1
11 ll Hj
84
74
12456 12456 12456 12456 12456
Labi Lab2 Lab3 Lab5 Lab8
Lab Bottle Number
Figure 4. Precision among Low Concentration (100 ug/L) Sample Recoveries for Tetraethylene Glycol
among the Analytical Laboratories excluding Laboratory 6 Data. (Closed Circle is Mean
Concentration and Whisker is 95% Confidence Limits. The Points with No Bottle Numbers are
for High Concentration Samples.)
The influence of the different matrices used to make the blind samples, upon first examination,
appear to have a strong influence on the analytical results at the low concentration level (Figure 5).
Matrices FS-1 and FS-3 have means outside the ±30% acceptance criterion. In contrast, matrices FS-2
and FS-4 fall very close to the known concentrations and well within the ±20% accuracy acceptance
criterion. These anomalous findings are the results of the influence of the high biased concentrations
found in the low-level sample concentrations from Laboratory 6 (Figure 2). Once the results from
Laboratory 6 are removed from the statistical analysis, the data indicate that there are no matrix affects
among the analytical laboratories and all mean concentrations are statistically similar (p > 0.5) and fall
within the ±20% accuracy acceptance criterion (Figure 6). Similarly, no matrix effects were observed
among the analytical laboratories at the high concentration levels (Figure 7) with all mean concentrations
falling within the ±20% accuracy acceptance criterion.
14
-------
230
210
190
"3)170
ii.
-£ 130 J: ----------------------------------------- 30%
CD ------------------------------------------ 20%
§110 .
O - 1 - i
90
------------------------------------------ 20%
70 ------------------------------------------ 30%
50
FS-1 FS-2 FS-3 FS-4
Source Water
Figure 5. Low Concentration (100 |jg/L) Average Recovery for Tetraethylene Glycol by Matrix. (Closed
Circle is Mean Concentration and Whisker is One Standard Error.)
140
130 ------------------------------------- 30%
— 120 ------------------------------------------------- 20%
"5)
3110
c
1 10° * - i - * - 1
c f
8 90
c
o
O 80 ------------------------------------- 2Q%
70 ------------------------------------- 30%
60
FS-1 FS-2 FS-3 FS-4
Source Water
Figure 6. Low Concentration (100 ug/L) Average Recovery for Tetraethylene Glycol by Matrix excluding
Laboratory 6 Data. (Closed Circle is Mean Concentration and Whisker is One Standard Error.)
15
-------
^
3.
(-
incentratio
u
o
280 i
ofin -
o/in
^4U
220 -
onn
^lUU
180 -
-i ^n
IDU "
•i/in
I4U
•ion -
<
FS
i i { i
>-1 FS-2 FS-3 F£
Source Water
3 no/
oU/o
THO/
20%
>
20%
o no/
oU/o
;-4
Figure 7. High Concentration (200 |jg/L) Average Recovery for Tetraethylene Glycol by Matrix. (Closed
Circle is Mean Concentration and Whisker is One Standard Error.)
4.2 Triethylene Glycol
All six participating laboratories provided data for triethylene glycol. All submitted blind blank
samples (see Table 3) showed no target analytes at or above the reporting limits of 5 or 8 |ig/L, depending
on the sensitivity of the instrument at the participating analytical laboratory. The blind samples were
spiked at 80 or 200 |ig/L.
The accuracy among the analytical laboratories was within 20% of the known concentrations,
except for Laboratory 6 for the high level spiked samples which fell within the ±30% acceptance criterion
(Figures 8 and 9). The results for any given laboratory included all samples, regardless of matrix, at the
given concentration level. A slight difference was seen in the results when the concentrations were
determined using either linear or quadratic calibration curves. When the concentrations of triethylene
glycol are calculated using the quadratic calibration curve, Laboratory 6 falls outside the ±20%
acceptance limit for the low level spiked sample but within the ±30% acceptance criterion while for the
high level spiked sample, Laboratory 6 falls outside the ±30% acceptance criterion.
16
-------
105 30o/o
95 § 20%
^~
DO
3 85 . .
c
g
| 75 . I
X I
O Rr
u 65 20o/o
55 30%
45
Labi Lab2 Lab3 Lab5 Lab6 Lab8
Figure 8. Low Concentration (80 |jg/L Spike) Average Recovery for Triethylene Glycol among the
Analytical Laboratories. (Closed Circle is Mean Concentration and Whisker is One
Standard Error.)
270
3QO/0
250
20%
230
| 210
o
- 190
£ • $
-------
90
85
"M 80
c
.2 75
g 70
c
o
u
65
60
55
00
Figure 10.
12346 12346 12346 12346 12346 12346
Labi Lab2 Lab3 Lab5 Lab6 Lab8
Lab Bottle Number
Precision among Low Concentration (80 |jg/L) Sample Recoveries for Triethylene Glycol among
the Analytical Laboratories. (Closed Circle is Mean Concentration and Whisker is 95%
Confidence Limits. The Points with No Bottle Numbers are for High Concentration Samples.)
The precision within the laboratory can be determined by examining the results of the individual
bottles shipped to the analytical laboratory. In each case, the bottle number (represented as the last digit
on the x-axis identifiers) represents the mean of the four different water matrices for that bottle number
(Figure 10). By checking significant differences among the 5 bottles spiked at the 80 (ig/L concentration,
the precision within an analytical laboratory was determined. In general, precision met the performance
criteria among all the bottles spiked at the same concentrations. The results from Laboratory 6 showed
that bottle 6 was statistically different (p<0.003) than the other four bottles received at the laboratory.
The influence of the different matrices used to make the blind samples showed a slight statistical
difference for the FS-2 matrix, which was slightly lower than the FS-4 matrix samples (Figure 11). This
anomaly in mean values may be the result of very tight precision within the participating analytical
laboratories for each matrix and slight accuracy differences resulting in the appearance of meaningful
differences when none really exist. No matrix effects were seen for the high concentration blind samples
among the analytical laboratories (Figure 12) with all mean concentrations being statistically similar (p >
0.52) and falling within the ±20% accuracy acceptance criterion.
18
-------
110
30%
100
20%
cr 90
1
g so
1
4-»
c
<" -7n
o 70
c
O
U 20%
60
30%
50
FS-1 FS-2 FS-3 FS-4
Source Water
Figure 11. Low Concentration (80 |jg/L) Average Recovery for Triethylene Glycol by Matrix.
(Closed Circle is Mean Concentration and Whisker is One Standard Error.)
270
30<%
250
_____ 20%
1J5230
.1 210
+->
ro
i_ -p
| 190 o T T
§ [ " I
U 170 1
20%
150
30%
130
FS-1 FS-2 FS-3 FS-4
Source Water
Figure 12. High Concentration (200 ug/L) Average Recovery for Triethylene Glycol by Matrix.
(Closed Circle is Mean Concentration and Whisker is One Standard Error.)
19
-------
4.3 Diethylene Glycol
All six participating laboratories provided data for diethylene glycol. All submitted blind blank
samples (see Table 3) showed no target analytes at or above the reporting limits of 5 or 8 |ig/L, depending
on the sensitivity of the instrument at the participating laboratory. The blind samples were spiked at 10 or
100 |ig/L.
A non-normal data distribution was found when using both spiked sample concentrations
together; therefore, the two populations (i.e., low and high spiked datasets) were treated separately
statistically. The primary cause for the existence of the two populations is believed to be a result of the
low level spiked sample concentration being so close to the reporting limits.
The accuracy among the participating analytical laboratories was within 20% of the known
concentrations for the low level spiked samples for Laboratories 1 and 5 and within ±30% for Laboratory
3 (Figure 13). Laboratory 6 did not report any data for the low level spiked samples as the spike levels
were too close to their laboratory reporting limit. Laboratory 2 results were high and Laboratory 8 results
were low for the low level spiked samples with both laboratories exceeding the ±30% acceptance
criterion. For the high level spiked samples, all analytical laboratories fell within the ±20% acceptance
criterion (Figure 14) except Laboratory 2 which had results above the 20% acceptance criterion.
The precision within the laboratory for diethylene glycol was determined by examining the
results of the individual bottles shipped to the analytical laboratory. In each case, the bottle number
(represented as the last digit on the x-axis identifiers) represents the mean of the four different water
matrices for that bottle number (Figure 15). By checking significant differences among the 5 bottles
spiked at the 10 (ig/L concentration, the precision within an analytical laboratory was determined. In
general, precision meeting the performance criteria was seen in Laboratories 5 and 8 while variability
among all the bottles spiked existed within each of the other participating laboratories. This variability
within the laboratories may be due to the closeness of the spike concentration (10 (ig/L) to the reporting
limits (5 (ig/L).
20
-------
19
17
15
c 13
o
c
CD
O
C
o
O
30%
20%
20%
30%
Labi
Lab2
Lab3
Lab5
Lab8
Figure 13. Low Concentration (10 ug/L Spike) Average Recovery for Diethylene Glycol among the
Analytical Laboratories. (Closed Circle is Mean Concentration and Whisker is One Standard
Error.)
165
145
c 125
o
§ 105
o
O
85
65
30%
20%
t
I
3U70
b1 Lab2 Lab3 Lab5 Lab6 Lab8
Figure 14. High Concentration (100 ug/L Spike) Average Recovery for Diethylene Glycol among the
Analytical Laboratories. (Closed circle is Mean Concentration and Whisker is One Standard
Error.)
21
-------
"
O 13
c
O
U 9
19
" *'* ,
'
T
T
5
1346713467134671346713467
Labi Lab2 Lab3 Lab5 Lab8
Lab Bottle Number
Figure 15. Precision among Low Concentration (10 pg/L) Sample Recoveries for Diethylene Glycol among
the Analytical Laboratories. (Closed Circle is Mean Concentration and Whisker is 95%
Confidence Limits.)
No matrix effects were seen at both high and low diethylene glycol concentration among the
analytical laboratories (Figures 16 and 17) with no significant difference among all the mean diethyl
glycol concentrations (p > 0.6 for Figure 20 and p > 0.25 for Figure 21) and falling within the ±20%
accuracy acceptance criterion.
22
-------
14
13
12
CT
"o5 11
O 10
03
-------
4.4 2-Butoxyethanol
Five participating laboratories provided data for 2-butoxyethanol. Laboratory 1 could not detect
2-butoxyethanol on their instrument. All submitted blind blank samples (see Table 3) showed no target
analytes at or above the reporting limits of 5, 8, or 10 |lg/L, depending on the sensitivity of the instrument
at the participating analytical laboratory. The blind samples were spiked at 60 or 180 |lg/L.
The accuracy among the analytical laboratories was within 20% of the known concentrations for
both the low level and high level spiked samples in all analytical laboratories (Figures 18 and 19).
85
75
=|65
c
0
'ro 55
-t->
CD
O
§ 45
o
35
O^
SU%
____________________________________ ~>no/
2.(j/o
* . {
. 1 •
____________________________________ ono/
ZU /o
SU%
Lab2 Lab3 Lab5 Lab6 Lab8
Figure 18.
Low Concentration (60 jjg/L Spike) Average Recovery for 2-Butoxyethanol among the Analytical
Laboratories. (Closed Circle is Mean Concentration and Whisker is One Standard Error.)
The precision within the laboratory can be determined by examining the results of the individual
bottles shipped to the analytical laboratory. In each case, the bottle number (represented as the last digit
on the x-axis identifiers) represents the mean of the four different water matrices for that bottle number
(Figure 20). By checking significant differences among the 5 bottles spiked at the 60 (ig/L concentration,
the precision within each analytical laboratory was determined. No statistically significant differences
were found among the bottles within the individual participating laboratory (p>0.43).
No matrix effects (p>0.10) were seen for either concentration blind samples among the
participating laboratories (Figures 21 and 22) with all mean concentrations being statistically similar and
falling within the ±20% accuracy acceptance criterion.
24
-------
260 I
240
30o/o
220
o
"CD
200 I i •
180 $ • i
20%
160 -\
| 140 20%
8 120 3°%
100 j
80 j
gQ ^ f 1 j r !
Lab2 Lab3 Lab5 Lab6 Lab8
Figure 19. High Concentration (180 ug/L Spike) Average Recovery for 2-Butoxyethanol among the
Analytical Laboratories. (Closed Circle is Mean Concentration and Whisker is One Standard
Error.)
25
-------
90
80
QO 70
c
o
60
(I O
c
OJ
u
40
30
12567 12567 12567 12567 12567
Lab2 Lab3 Lab5 Lab6 Lab8
Lab Bottle Number
Figure 20. Precision among Low Concentration (60 ug/L) Sample Recoveries for 2-Butoxyethanol among
the Analytical Laboratories. (Closed Circle is Mean Concentration and Whisker is 95%
Confidence Limits. The Points with No Bottle Numbers are for High Concentration Samples.)
85
75
00 65
c
O
55
c
-------
240 -
220 -
— 200 -
00
i
c 180
0
03
c 160 -
OJ
o
c
o
U 140 -
-i nn ~
20%
0
I
I O
^ I
20%
30%
FS-1
FS-2 FS-3
Source Water
FS-4
Figure 22. High Concentration (180 pg/L) Average Recovery for 2-Butoxyethanol by Matrix. (Closed
Circle is Mean Concentration and Whisker is One Standard Error.)
27
-------
4.5 2-Methoxyethanol
Four laboratories provided data for 2-methoxyethanol. Laboratories 1 and 3 could not detect 2-
methoxyethanol on their analytical instruments. All submitted blind blank samples (see Table 3) showed
no target analytes at or above the reporting limits of 5, 20, or 25 |lg/L, depending on the sensitivity of the
instrument at the analytical laboratory. The blind samples were spiked at 40 or 100 |lg/L.
At the higher concentration, Laboratory 6 was significantly different from the others laboratories
(p<0.05); whereas, Laboratory 5 was significantly different only from Laboratory 2 (p<0.001). At the
lower concentration, there were no significant differences between labs (p> 0.136). The accuracy among
the analytical laboratories was within 20% of the known concentrations for both the low level and high
level spiked samples in all analytical laboratories with the exception of the 40 (ig/L samples analyzed at
Laboratory 6 (Figures 23 and 24). Laboratory 6 results were positively biased and were just slightly
higher than the ±30% acceptance limit.
28
-------
60
55
* 30%
50
20%
~c 45
O
03
£ 40
OJ
o
O T
U 35 1 I
30
30%
25
Lab2 Lab5 Lab6 Lab8
Figure 23. Low Concentration (40 ug/L Spike) Average Recovery for 2-Methoxyethanol among the
Analytical Laboratories. (Closed Circle is Mean Concentration and Whisker is One Standard
Error.)
140
130 >- 30%
120 f 20%
<> (i
i 110
c
O
TO 10°
c
o 90
c
O
u
80 20%
70 30%
60
Lab2
Lab5
Lab6
Lab8
Figure 24. High Concentration (100 ug/L Spike) Average Recovery for 2-Methoxyethanol among the
Analytical Laboratories. (Closed Circle is Mean Concentration and Whisker is One Standard
Error.)
29
-------
70
60
• 50
c
o
'-4-J
03
i_
-t-i
£ 40
(J
c
O
u
30
20
O O
12356 12356 12356
Lab2 Lab5 Lab6
Lab Bottle Number
12356
Lab8
Figure 25. Precision among Low Concentration (40 |jg/L) Sample Recoveries for 2-Methoxyethanol among
the Analytical Laboratories. (Closed Circle is Mean Concentration and Whisker is 95%
Confidence Limits. The Points with No Bottle Numbers are for High Concentration Samples.)
The precision within the laboratory was determined by examining the results of the individual
bottles shipped to the analytical laboratory. In each case, the bottle number (represented as the last digit
on the x-axis identifiers) represents the mean of the four different water matrices for that bottle number
(Figure 25). By checking significant differences among the 5 bottles spiked at the 40 (ig/L concentration,
the precision within an analytical laboratory was determined. No statistically significant differences
(p>0.14) were found among the bottles within the individual participating laboratory except for
Laboratory 6 which showed statistical differences, but not significant differences, between bottle 3 and
bottles 2 (p>0.22) and 5 (p>0.14; Figure 25).
No matrix effects were seen for either concentration blind samples (p >0.083). Overall mean
concentrations for water matrices are within the ±20% accuracy acceptance criterion (Figures 26 and 27).
30
-------
55
50
45
40
03
o
u
35
30
25
FS-1
FS-2 FS-3
Source Water
30%
20%
20%
30%
FS-4
Figure 26. Low Concentration (40 ug/L) Average Recovery for 2-Methoxyethanol by Matrix. (Closed
Circle is Mean Concentration and Whisker is One Standard Error.)
140
oo
110
c
o
'•4-1
03
100
§ 90
c
o
U 80
_
20%
70
60
FS-1
30%
FS-2 FS-3
Source Water
FS-4
Figure 27. High Concentration (100 ug/L) Average Recovery for 2-Methoxyethanol by Matrix. (Closed
Circle is Mean Concentration and Whisker is One Standard Error.)
31
-------
32
-------
5.0 Summary and Conclusions
Several key factors were used to determine the performance characteristics of the draft glycol
method including: detectability of the compounds, accuracy, precision, and the presence/absence of
matrix effects. A batch of 36 blind samples spiked with the five compounds of interest in four different
water matrices was submitted to six analytical laboratories. Each laboratory was provided with a copy of
the draft glycol SOP and asked to optimize their LC/MS/MS system to perform the analyses of the blind
samples. Once optimized, the blind samples were analyzed following the QA/QC requirement identified
in the project QAPP. Each laboratory provided the resultant data as determined using both linear and
quadratic calibration curves. The data were statistically analyzed, after checking the data for normalcy,
by running least square means analyses to determine if statistical differences existed among the
laboratories and if the data fit within the acceptance criterion for the key factors of interest.
Diethylene glycol, triethylene glycol, and tetraethylene glycol were detectable at all participating
analytical laboratories. 2-butoxyethanol could not be successfully detected at Laboratory 1 while
Laboratories 1 and 3 could not successfully detect 2-methoxyethanol.
While statistical differences between laboratories were present, the accuracy of the laboratory
analyses were typically within the acceptance criteria of either ±20 or 30% of known value. A few
notable exceptions were identified including:
a) Laboratory 6 was not within ±30% for tetraethylene glycol with a positive bias at the low
concentration,
b) Laboratory 6 was not with ±20% for triethylene glycol with a negative bias at the high
concentration,
c) Laboratory 2 was not within ±30% for diethylene glycol with a positive bias at the low
concentration,
d) Laboratories 3 and 8 were not within ±20% and ±30%, respectively, for diethylene glycol with a
negative bias at the low concentration, and
e) Laboratory 6 was not within ±30% for 2-methoxyethanol with a positive bias at the low
concentration.
Precision within the analytical laboratories met the performance criteria indicating that
reproducible results were being produced at the analytical laboratories. All replicate bottles were
statistically the same with the exceptions of diethylene glycol across the analytical laboratories and 2-
methoxyethanol in Laboratory 6. The variability within the laboratories during the analysis of diethylene
glycol was most likely due to the closeness of the spike concentration (10 (ig/L) to the reporting limits (5
or 8 (ig/L). Laboratory 6 did not report any concentrations for diethylene glycol as a result of the spiked
concentration of the low level sample being too close to their reporting limit.
Matrix effects were not identified for the tested water matrices at any of the analytical laboratories
indicating that the method could produce the same results in the four water matrices tested.
33
-------
Overall, the draft glycol SOP presented a method that was accurate and precise, when the
compounds of interest were detectable, by meeting the established performance criterion nearly all the
time. Further, the draft method exhibited no matrix effects in the four waters tested for any of the
compounds of interest. The detection of 2-ME and 2-BE was problematic and instrument/laboratory
dependent. With the few method variations and QA/QC deviations that were identified throughout the
study among the analytical laboratories, a strong QA/QC program to monitor the resultant data is
essential. The QA/QC program should incorporate blind samples of known concentrations to ensure
quality of the resultant data and that the results are within the specified accuracy acceptance limits.
34
-------
6.0 Recommendations
The interlaboratory verification of the draft glycol SOP resulted in some recommendations in
order to further improve the analytical method. These recommendations are presented below:
a) The draft glycol SOP states that at least a 5-point initial linear calibration curve should be used. This
restriction was found to limit the effectiveness of the method. Many calibration curves generated
during the study were best fit using quadratic formulae yet the linear curves were also acceptable.
The deviations from the calibration curve were lessened, especially at the low concentration standards
when incorporating a quadratic fit. A minimum 6-point quadratic fit should be allowed with
stipulations on the allowable deviations of the calibration points from the curve, such as ±25%
deviation from the curve with a correlation coefficient, r2 > 0.99.
b) Surrogate spikes should be incorporated in order to have recovery data with every sample. There are
now commercially available isotopically labeled 2-BE and Di-EG that do not require custom
synthesis. Recovery limits should be set in reagent water and matrix waters for the surrogate spikes
with correlations back to the native analytes.
c) There are no preservation or holding time studies of these analytes in reagent water or matrices of
concern. A fourteen day holding time was used for this study. Studies should be conducted to
determine if holding at 4 ±2° C is adequate to preserve sample integrity. The multi-laboratory study
did, however, demonstrate that the matrices studied did not affect the integrity of the spiked sample
from the time of collection to the time of analysis in the participating laboratories.
d) The preparation of samples in the laboratory does not require a filtering procedure but leaves it as an
option to use a 0.45 (im Teflon® filter unit. There are no data presented demonstrating the
performance of a filter unit and if target analyte contamination is an issue. Various glycols may be
used in the manufacturing or cleaning of the filter units and may cause bias in the results. A filter unit
study should be conducted to determine their performance and effects on contaminant concentrations.
e) The liquid chromatography conditions in the draft glycol SOP rise to a maximum of 15% acetonitrile.
This may result in a build-up of organic contaminants on the column that will lessen the performance
over time. A higher organic content gradient for the acetonitrile should be used in order to elute non-
target analytes from the column with each injection cycle.
f) Mass calibration/tuning appear to be required annually in the draft glycol SOP. This should be
checked routinely and re-calibration should be required before analysis if mass shift is noticed that
will affect the sample results.
g) Second source standards are an issue with this method. Standard concentrations need to be verified
between the different vendors.
h) The final CCV should have a concentration near the mid-point of the calibration curve. During this
verification study, varying CCV concentrations covering the calibration curve range were used and
depending on the concentration selected, different recovery biases may be encountered.
35
-------
36
-------
7.0 References
(1) US EPA Office of Pollution Prevention and Toxics: High Production Volume (HPV)
Challenge, http://www.epa.gov/hpv/pubs/general/opptsrch.htm, Accessed October 25, 2012.
(2) Sorensen, J. A.; Gallagher, J. R.; Hawthorne, S. B.; Aulich, T. R., JV Task 3 - Gas Industry
Groundwater Research Program: Final Report. DOE NETL, 2000.
(3) US EPA Office of Water: Evaluation of Impacts to Underground Sources of Drinking Water by
Hydraulic Fracturing of Coalbed Methane Reservoirs (816-R-04-003). Washington, DC, 2004.
(4) Quality Assurance Project Plan for the Multi-Laboratory Verification of Diethylene Glycol,
Triethylene Glycol, 2-Butoxyethanol and 2-Methoxyethanol in Ground and Surface Waters by
Liquid Chromatography/Tandem Mass Spectrometry.
http://www2.epa.gov/sites/production/files/documents/glvcol-qapp-revO-0.pdf
(5) Peterson R.G. 1985. Design and Analysis of Experiments. Marcel Dekker, Inc. New York.
1985.
(6) Federer W. T. (1955) Experimental design. Oxford & IBH publishing CO. PVT. LTD. New
Delhi.
37
-------
38
-------
Appendix A
Glycol Analysis of Aqueous Samples by Direct Injection HPLC/MS/MS
39
-------
40
-------
DRAFT R3QA239-xxxxxxxx
Glycol Analysis by HPLC/MS/MS
Effective Date: 2012
Page 1 of 15
Glycol Analysis by HPLC/MS/MS
Effective Date: XXXXX2012
EPA Region 3
Office of Analytical Services and Quality Assurance
701 MapesRoad
Fort Meade, Maryland 20755
Approved by:
Jill Bilyeu Date
Quality Assurance Officer
Prepared by: Jennifer L. Gundersen, Ph.D.
Laboratory Branch
Revi ewed by: xxxxxxx
Laboratory Branch
The controlled official version of this document is the electronic version viewed on-
line. If this is a printed copy of the document, it is an uncontrolled working copy.
-------
DRAFT
R3QA239-xxxxxxxx
Glycol Analysis by HPLC/MS/MS
Effective Date: 2012
Page 2 of 15
Updates Table
Peer reviewer's initials indicate that changes meet the NELAC and regulatory requirements described in
Section 9.4 in SOP R3-QA060
Responsible
Person
Date
Description of Change
Peer
Reviewer
Date
New SOP
-------
DRAFT
R3QA239-xxxxxxxx
Glycol Analysis by HPLC/MS/MS
Effective Date: 2012
Page 3 of 15
1 Scope and Application
1.1 This SOP is based on EPA SW-846 Method 8321B, 8000C and ASTM D7731-11E1. See Table 1 for
analytes.
Table 1.
Analyte
Diethylene glycol
Triethylene glycol
Tetraethyleneglycol
2-Butoxyethanol
2-Methoxyethanol
CAS#
111-46-6
112-27-6
112-60-7
111-76-2
109-86-4
MDL
(aqueous, ug/1)
In prep
In prep
In prep
In prep
In prep
NQL
(aqueous, ug/1)
25
25
25
5
10
2 Summary of the Method
2.1 The method employs high performance liquid chromatography (HPLC) coupled with positive
electrospray ionization (ESI+) tandem mass spectrometry (MS/MS) for the determination of a suite of
glycols in aqueous matrices.
2.2 There is no extract!on.The sample is directly injected into the HPLC/MS/MS system. Quantitation is
performed through linear, external standard, calibration.
2.3 Target compounds are identified by retention time and one or more MRM (Multiple Reaction Monitoring)
transition.
3 Definitions
3.1 Refer to the ESC Quality Manual for applicable definitions
4 Interferences
4.1 Suspended solids in the sample can clog frits in the sample management system and on the column. If
site history suggests, samples may be filtered prior to introduction to the HPLC/MS/MS system.
4.2 Matrix interferences may be caused by contaminants in the sample.
4.3 All reusable glassware must be cleaned according to procedures for cleaning glassware used in organic
compound analyses. R3QA-054 Glassware Preparation for Organic Analyses.
-------
DRAFT R3QA239-xxxxxxxx
Glycol Analysis by HPLC/MS/MS
Effective Date: 2012
Page 4 of 15
5 Safety
5.1 Before beginning any procedures, refer to the Chemical Hygiene Plan (CHP) in the OASQA Quality
Assurance Manual for general safety precautions and guidelines.
5.2 All sample prep work should be conducted in a fume hood.
5.3 The toxicity or carcinogenicity of each reagent used in this method may not have been fully established.
Each chemical should be regarded as a potential health hazard and exposure should be as low as
reasonably achievable.
5.4 Material Safety Data Sheets (MSDS) must be maintained in the facility for all reagents used in the
laboratory. This information must be made available to all personnel prior to the performance of this SOP
and upon staff request. The MSDS (hard copies) are currently located in the library as well as
electronically on CD-ROM and online.
5.5 All applicable safety and compliance guidelines set forth by the EPA and by federal, state, and local
regulations must be followed during the performance of this SOP. In addition, all procedures outlined in
the ASQAB Chemical Hygiene Plan must be adhered to. Stop all work in the event of a known or
potential compromise to the health and safety of any person and immediately notify the Safety Officer,
and other appropriate personnel as outlined in the CHP.
5.6 All laboratory waste must be handled in accordance with guidelines established in the CHP and the
appropriate waste disposal procedures identified in Section 15.0 (Waste Management).
5.7 Analysts must be cognizant of all instrumental hazards (i.e. dangers from electrical shock, heat or
explosion etc.).
5.8 All chemicals used in the performance of this SOP, as well as the samples, should be handled with
caution. Adequate protective gear should be worn. At a minimum, this includes ANSI approved safety
glasses and a lab coat to protect from chemical splashes, and powderless gloves made from acid resistant
materials such as nitrile, latex, neoprene, butyl or PVC.
5.9 Spill procedures: Follow the procedures outlined in the ESC Occupant Emergency Plan (OEP),
Hazardous Material Spills section. For minor spills (which can be handled by the analyst) wear safety
glasses, lab coat, and gloves to clean up the material. For significant spills, immediately contact the
SHEM Manager.
6 Equipment and Supplies
6.1 HPLC/MS/MS system: Analytical instrument and accessories suitable for automated injection of samples
onto analytical HPLC columns and fragmentation and detection by a tandem mass spectrometer.
6.2 System used at R3-ESC: Waters (Milford, MA) TQD HPLC/MS/MS system: equipped with a 1 to 50 uL
or 1 to 100 jiL loop injector and electrospray (ESI) tandem mass spectrometer (MS/MS) capable of
-------
DRAFT R3QA239-xxxxxxxx
Glycol Analysis by HPLC/MS/MS
Effective Date: 2012
Page 5 of 15
multiple reaction monitoring (MRM) and negative and positive ion mode.
6.3 HPLC column: Waters (Milford, MA) Atlantis dC18 3|im, 2.1 x 150mm. Other columns may be used if
they provide sufficient retention and separation of the target analytes.
6.4 Data System: Computer system with software capable of accepting and processing raw detector data from
the HPLC/MS/MS. The system must have the following capabilities:
Integrate peaks from raw data.
Provide peak height and peak area information.
Calculate and store calibration information.
Identify peaks of interest by retention time.
Quantitate peaks of interest using calibration obtained.
Produce chromatograms.
Allow overlay and comparison of chromatograms.
Produce reports with quantitation information.
Provide a vehicle for storing data.
Define manually integrated data on report.
The current system for operation and processing is Waters Empower! (current revision)
6.5 Disposable 0.45um syringe tip filters, Teflon, if needed.
6.6 Disposable luer tip syringes, sized as appropriate, if needed.
6.7 Volumetric flasks - Class A glass: sized as appropriate
6.8 Micro syringes or Class A graduated (to deliver) pipets, sized as appropriate
6.9 Autosampler vials- Glass, 2 mL crimp top or screw top with Teflon-lined septum
6.10 Graduated cylinders, sized as appropriate
6.11 Disposable Pasteur pipets
7 Reagents and Standards
7.1 Acetonitrile - HPLC grade or equivalent. Optima grade is preferred.
7.2 Organic-free, deionized water: ASTM Type III water provided and monitored in-house according to R3-
QA065 (current revision) and further polished at a point of use Millipore unit to a resistivity of 18 MQ-
cm and a total organic carbon of less than 50 ppb.
7.3 Nitrogen gas, provided by liquid nitrogen dewars
-------
DRAFT R3QA239-xxxxxxxx
Glycol Analysis by HPLC/MS/MS
Effective Date: 2012
Page 6 of 15
7.4 Argon gas, provided by liquid argon dewars
7.5 Formic Acid, reagent grade.
7.6 Sodium Cesium Iodide, NaCsI. For instrument tuning. Provided annually by manufacturer with system
preventive maintenance (PM) kit.
7.7 Mobile phase: Reservior Al: H2O with 0.1% formic acid, Reservior B1: Acetonitrile with 0.1% formic
acid.
7.8 All standards are to be labeled with the Element standard number and the preparer's initials. This is a
unique identifier and all standard information is referenced in Element. Other information may include:
expiration date, concentration, and manufacturer.
7.9 Standards must be stored in glass containers at 4 +/-2°C.
7.10 Stock standard solution 100 mg/L (ppm) glycol mix - This solution can be purchased commercially as a
certified standard. Stock standards should be stored at 4-6°C or according to manufacturer's suggestions
until manufacturer's expiration. Expiration dates should be clearly specified on the label.
7.11 Intermediate standard solution (1.0 and 10 mg/L glycol mix) - Prepared by dilution of stock standard
solution to 10 or 100 mL with reagent water. Intermediate standards may be stored at 4±2 °C for a period
of up to 6 months. Expiration dates should be clearly specified on the label.
7.12 Calibration standards - Prepare dilutions of the intermediate standard solution. A minimum of 5
calibration standards is recommended. A sufficient number of standards should be analyzed in order to
allow an accurate calibration curve to be established. Due to the varied responses of the analytes,
recommended standard concentrations for establishing a calibration curve are: 5, 10, 25, 50, 100, 200, and
400ug/L (ppb). This range may be extended provided that the linear response can be adequately verified
through satisfaction of all calibration criteria and quality control requirements. The low standard must be
equivalent to or below the lowest result to be reported. All reported results must be within the calibration
range.
8 Sample Collection, Preservation and Storage
8.1 Samples must be stored in tightly sealed glass at 4 +/- 2°C in a designated sample refrigerator.
8.2 Analyze samples within 14 days of collection.
8.3 Samples extracted outside of holding time should be noted in the case narrative as qualified according to
the lab QM.
-------
DRAFT
R3QA239-xxxxxxxx
Glycol Analysis by HPLC/MS/MS
Effective Date: 2012
Page 7 of 15
9 Quality Control
9.1 Batch QC. The following are relevant QC criteria for this method taken from the OASQA Laboratory
Quality Manual (current revision).
NELAC Requirement
Minimum Frequency
Acceptance Criteria
Corrective Action
Method Blank-BLK
(clean matrix processed)
One per sample preparation
batch1
Fails if the concentration of a
targeted analyte in the blank is
at or above the reporting limit,
AND is greater than 1/10 of the
amount measured in any sample.
Criteria do not apply to sample
results reported as less than
values and mandated methods
that require correction for
blanks.
If outside acceptance criteria
reprep affected samples or
qualify sample results.
Laboratory Control
Sample (LCS) - BS
(clean matrix spiked with
analytes of interest)
One per sample preparation
batch
±20% of expected value for
aqueous samples. As per 8000C.
LCS/BS is equivalent to CCV
and SCV because there is no
extraction. Sec 11.7
If outside acceptance criteria,
first re-analyze the failed QC
to verify difficulty. If still
failing, perform corrective
actions and reprep. affected
samples or qualify results.
Matrix Spike - MS
(spiked or fortified
sample)
One per sample preparation
batch
Selection of sample
±30% of expected value for
aqueous samples. As per
SOOOC.Sec 9.5.4.
If outside acceptance criteria,
qualify the sample associated
with failing QC results.
Matrix Spike Duplicate
-MSB
(analysis of second
fortified aliquot,
processed)
One per 20 samples per
matrix and site
Selection of sample
Relative percent difference: 25,
as per Method 8000C. ±30% of
expected value for aqueous
samples. Sec 9.5.4. RPD<25
If outside acceptance criteria,
qualify the sample associated
with failing QC results.
Initial Calibration
STB
At least two calibration
standards with one at the
Level of Quantitation (not
to include the blank) unless
fewer standards are
specified by a mandated
method.
r2^ 0.99 as per Method
SOOOC.Sec 9.3.2. Minimum of
5 concentrations Method 800C
Secll.4.1.1
If the initial instrument
calibration results are outside
established acceptance criteria,
corrective actions must be
performed. Results associated
with an unacceptable initial
instrument calibration must be
qualified. Results of samples
not bracketed by initial
instrument calibration
standards (within calibration
range) must be reported as
having less certainty.
Second Source Quality
Control Standard (QCS)
- SCV (material is from a
second source; source
independent of
One per initial calibration
±20% of expected value as per
Method SOOOC.Sec 9.3.6.
If outside acceptance criteria,
first re-analyze or reprep. the
failed QC to verify difficulty.
If still out, correct problem
then recalibrate or qualify
-------
DRAFT
R3QA239-xxxxxxxx
Glycol Analysis by HPLC/MS/MS
Effective Date: 2012
Page 8 of 15
NELAC Requirement
Minimum Frequency
Acceptance Criteria
Corrective Action
calibration standards, not
processed)
results.
Continuing Instrument
Calibration Verification
-ccv
One at beginning, end and
every 20 samples
(analytical batch).
Only one per analytical
batch is needed if using
internal standards.
±20 of expected value as per
Method 8000C. Sec 11.7.6
If outside acceptance criteria,
first re-analyze or reprep the
failed QC to verify difficulty.
If reanalysis passes the first
time, then continue run. If
reanalysis fails but routine
corrective actions correct the
problem, then there must be
two consecutive passing QCs
before continuing the run. If it
still fails, then recalibrate and
reanalyze all samples since the
last acceptable CCV or stop
analysis (additional analyses
shall not occur) and if any
samples in the batch can not be
re-analyzed report data
specifying the direction of the
bias if clearly indicated.
Selectivity - Retention
Time
All chromatography
methods
All analytes in initial calibration
standards, LCS-BS, SCV and
CCV within windows
established per method or in-
house limits. The Empower
software processing method
currently sets the retention time
window at ±5% of the
If outside acceptance criteria,
first re-analyze or reprep. the
failed QC to verify difficulty.
If still out, correct problem
then recalibrate or qualify
results.
Surrogate - SUR
Organic only - All samples,
standards, QC (Surrogate
compounds as per SOP and
mandated methods). Not
currently used, may be
added at a future date.
Not currently used, may be
added at a later date.
If outside acceptance criteria,
qualify results associated with
failing QC.
Tuning
Mass spectrometry methods -
before each analytical batch '
ASTMD7731-11 states that
tuning should be done
according to manufacturer's
directions. Because hardware
tuning is done with NaCsI,
tuning is recommended to be
done yearly with the PM so
that salts do not build up on the
quadrupole.
According to manufacturer's
directions.
Perform instrument
maintenance and rerun tuning
standard. Data associated with
an unacceptable tune shall not
be reported.
Batch: environmental samples that are prepared and/or analyzed together with the same process and personnel, using the same lot(s)
of reagents. A preparation batch is composed of one to 20 environmental samples of the same NELAC-defmed matrix, meeting the
above mentioned criteria and with a maximum time between the start of processing of the first and last sample in the batch to be 24
hours. An analytical batch is composed of prepared environmental samples (extracts, digestates or concentrates) which are analyzed
-------
DRAFT R3QA239-xxxxxxxx
Glycol Analysis by HPLC/MS/MS
Effective Date: 2012
Page 9 of 15
together as a group. An analytical batch can include prepared samples originating from various environmental matrices and can exceed
20 samples. (NELAC Quality Systems Committee)
The components to be spiked shall be as specified by the mandated test method. Any permit specified analytes, as specified by
regulation or client requested shall also be included. If there are no specified components, the laboratory shall spike per the following:
For those components that interfere with an accurate assessment such as spiking simultaneously with technical chlordane, toxaphene
andPCBs, the spike should be chosen that represents the chemistries and elution patterns of the components to be reported.
For those test methods that have extremely long lists of analytes, a representative number may be chosen using the following criteria for
choosing the number of analytes to be spiked. However, the laboratory shall insure that all targeted components are included over a two
year period.
For methods that include 1-10 targets, spike all components.
For methods that include 11-20 targets, spike at least 10 or 80%, whichever is greater.
For methods that include 21 or more targets, spike at least 16 components.
(NELAC, Section D.I.1.3.Ic)
The selected sample shall be rotated among client samples so various matrix problems may be noted and/or addressed.
10 Calibration and Standardization
10.1 Refer to the Batch QC table for calibration criteria.
10.2 While many mass spectrometry methods require daily tuning to assure proper mass identification prior to
each sample batch, ASTM Method D7731-11 states that tuning/mass calibration should be according to
manufacturer's directions. According to the TQD Operator Manual, unless problems are noted, this
system is only required to be tuned for proper mass identification annually with the system PM. Tuning
is done with a NaCsI solution and repeated introduction of NaCsI can cause buildup of salt in the system
and result in reduced sensitivity and will necessitate frequent cleaning.
10.3 Tuning to determine the correct system settings (cone voltage, desolvation temperature, source
temperature etc) for a particular analyte is done as needed and according to manufacturer's directions.
Representative settings for the analtyes in this method are listed in Section 11.
10.4 Records of the annual system PM are maintained in the instrument maintenance log.
10.5 Suggested concentrations for the initial calibration levels are 5.0 to 400.0 ppb. If a wider calibration
range is needed, more standard levels should be added provided the calibration curve remains linear.
10.6 Linear calibration may be used if the r2 > 0.99 and all continuing calibrations and calibration verifications
pass.
10.7 The average of the retention times of the mid-level concentrations is to be used in the processing method
as the analyte retention time.
10.8 Certificates of analysis are stored in G201.
11 Procedure
11.1.1 Transfer sample to an autosampler vial using a glass Pasteur pipet. If necessary, filter the sample through
-------
DRAFT
R3QA239-xxxxxxxx
Glycol Analysis by HPLC/MS/MS
Effective Date: 2012
Page 10 of 15
a 0.45|im syringe tip filter and dispense into autosampler vial.
11.1.2 Prepare matrix spike samples in a 10.0 mL volumetric flask. Fill to about 50% with sample; add an
appropriate volume of spike solution to achieve the needed concentration. The volume of spike added
should not be more than 100-200ul (1-2% of the total sample volume) or it could affect the concentration
in the source sample. Fill the volumetric flask to the mark with sample and mix by inverting several
times. If necessary, filter the sample through a 0.45|im syringe tip filter and dispense into autosampler
vial.
11.2 HPLC/MS analysis
11.2.1 Calibrate the HPLC/MS/MS with NaCsI, according to manufacturer's directions, during annual
preventive maintenance. More frequent calibration with NaCsI can leave residue on the quadrupoles and
should only be done following significant instrument repair.
11.2.2 Appropriate MRMs were determined during method development (see 11.2.6 below) but can be
reevaluted as needed, by tuning with authentic, individual standards to determine the most abundant
MRMs. Tuning may be done via the Waters Intellistart™ automated tuning program or manually through
the tune page.
11.2.3 Mobile phases.
11.2.3.1 For 2-methoxyethanol, isocratic elution at 0.3ml/min at 98% Al and 2% B1 is used.
11.2.3.2 For the other analytes a gradient is used.
Time (min)
Initial
3.0
10.5
12.5
13
13-19
Flow rate ml/min
0.4
0.4
0.4
0.4
0.4
0.4
%A1
98
98
85
85
98
98
%B1
2
2
15
15
2
2
Curve
Linear
Linear
Linear
Linear
Linear
Equilibration before
next injection
11.2.4 The typical injection volume is 30 jiL.
11.2.5 The gradient may be modified to achieve separation of target analytes in one run.
11.2.6 The following MRMs are monitored but may be adjusted depending on instrument response. The MRM
marked * has a higher response and is used as the primary MRM for calibration and quantitation. The
second MRM may be monitored and for supplementary confirmation but due to the lower response,
cannot be used to confirm concentrations at the lower portions of the calibration curve. ASTM D7731-11
uses only one MRM per analyte.
-------
DRAFT
R3QA239-xxxxxxxx
Glycol Analysis by HPLC/MS/MS
Effective Date: 2012
Page 11 of 15
Diethylene Glycol, Time: 0-5min, span: 0.2 Da, retention time (RT): l.Stnin
Precursor (Da)
106.94
106.94
Product (Da)
44.9*
88.4
Dwell (sec)
0.2
0.2
Cone voltage (V)
18
18
Collision energy V)
48
22
Triethylene Glycol, Time:0-5min, span 0.2 Da, RT: 2.9min
Precursor (Da)
150.97
150.97
Product (Da)
45.10*
89.00
Dwell (sec)
0.2
0.2
Cone voltage (V)
24
24
Collision energy V)
26
24
Tetraethylene Glycol: Time 5-13min, span 0.2 Da, RT: 5.6 min
Precursor (Da)
195.05
195.05
Product (Da)
45.10*
89.00
Dwell (sec)
0.2
0.2
Cone voltage (V)
22
22
Collision energy V)
22
20
2-Butoxyethanol: Time, 5-13min, span 0.2 Da, RT: 10.6min
Precursor (Da)
118.93
118.93
Product (Da)
57.10
63.00*
Dwell (sec)
0.2
0.2
Cone voltage (V)
16
16
Collision energy V)
20
14
2-Methoxyethanol: Time 0-4min, span 0.2 Da, RT: 2.6min
Precursor (Da)
76.91
Product (Da)
59.10*
Dwell (sec)
0.2
Cone voltage (V)
12
Collision energy V)
8
11.2.7 MS/MS settings may be adjusted to meet quantitation limit requirements but are generally as follows:
Desolvation temperature
Source temperature
Collision gas flow (Argon)
Cone gas
Desolvation gas
Ion Mode
Column temperature
Sample chamber
Inter-channel delay
Inter- scan delay
Capillary
2-methoxyethanol
350°C
150°C
O.lml/min
25 L/hr
600 L/hr
Electrospray postive (ESI+)
30°C
4°C
0.005s
0.005s
3.40
All other analytes
400°C
150°C
O.lml/min
25 L/hr
800 L/hr
Electrospray postive (ESI+)
30°C
4°C
0.005s
0.005s
3.40
12 Data Analysis and Calculations
12.1 Refer to the current version of the Laboratory QM for Quality Control related equations and the policy on
reporting significant figures.
-------
DRAFT R3QA239-xxxxxxxx
Glycol Analysis by HPLC/MS/MS
Effective Date: 2012
Page 12 of 15
12.2 Refer to R3QA-067 (current revision) for policies on manual integration.
12.3 Identify and confirm the presence of target analytes in the samples by matching the retention time of the
MRM
Compare the retention time of the MRM with the retention time determined during the initial calibration.
The retention times should not be more than 5% different from the initial calibration average.
12.4 If used, the internal standard calculation Response Factor (RF) can be calculated acocording to the Lab
QM.
12.5 Linear (external) calibration may be used if the r2 > 0.99.
12.6 Water samples
Final result (|ig/L C1O4") = (C)(D)
Where:
C = Concentration from IS calibration or calibration curve (|ig/L CICV) D = Dilution factor (if needed)
13 Method Performance
13.1 Method performance is evaluated based on the criteria in Table 2.
13.2 DOC accuracy and precision data and MDL study data are maintained in the OASQA Central QS files.
13.3 NQLs are listed in Section 1. There are no problematic compounds associated with this method
14 Pollution Prevention
14.1 This method has been developed to generate 10 mL or less of waste per aqueous sample. As this SOP is
routinely performed, the analyst will consider other methods to reduce the use and generation of
hazardous chemicals/waste.
14.2 Resource Management: Water Conservation. Laboratory personnel should be mindful of water
consumption, and whenever possible, employ practices that minimize water use.
15 Waste Management
15.1 Waste type code: Will vary with sample. Record the WO # on sample waste containers.
15.2 All laboratory waste must be handled in accordance with guidelines established in the ESC Chemical
Hygiene Plan (current revision).
15.3 The waste flow chart is on file with the SHEM Office.
-------
DRAFT R3QA239-xxxxxxxx
Glycol Analysis by HPLC/MS/MS
Effective Date: 2012
Page 13 of 15
15.4 Amount of waste per sample: Approximately lOmL or less of waste will be generated per sample.
16 References
16.1 SW-846 Method 8321B,Solvent-extractable nonvolatile compounds by high-performance liquid
chromatography/thermospray mass spectrometry or ultraviolet detection (rev 2, Feb 2007)
16.2 SW-846 Method 8000C, Determinative Chromatographic Procedures, (rev 3, March 2003)
16.3 ASTMD7731-11E1, Standard Test Method for Determination of Dipropylene Glycol Monobutyl Ether in
Sea Water by Liquid Chromatography/Tandem Mass Spectrometry. (August 2011)
16.4 Waters ACQUITY TQD Empower 2154 customer Familiarization Guide. Waters Corp. (2008) Milford
MA.
16.5 EPA Region 3 OASQA Laboratory Quality Manual (QM), Current Revision.
16.6 EPA Region 3 OASQA Chemical Hygiene Plan, Current Revision.
16.7 EPA Region 3 OASQA Occupant Emergency Plan, Current Revision.
16.8 EPA Region 3 OASQA, Laboratory Notebook Policy, Current Revision.
16.9 TQD Maintenance logbook: SNB 357.
16.10 Waters TQD System Run Log: PNB 207
16.11 Certificates of analysis notebook: SNB 114
16.12 R3-QA067. Procedures for Manual Integration, Current revision.
16.13 R3-QA054. Glassware Preparation for Organic Analyses. Current revision.
16.14 R3-QA065. Calibration, Verification and Maintenance of Laboratory Support Equipemt. Current revision.
16.15 NEL AC Standard. Current revi si on
17 Tables, Diagrams, Flowcharts and Validation Data
17.1 Waste handling flow chart is on file with the SHEM office.
17.2 QA/QC data is on file with the OASQA Quality Assurance Officer.
17.3 Attachment 1. EPA Internal Technical Review Checklist
-------
DRAFT
R3QA239-xxxxxxxx
Glycol Analysis by HPLC/MS/MS
Effective Date: 2012
Page 14 of 15
Attachment 1: Glycols by LC/MS (R3-QA239) Technical Review Checklist (TRC) Checklist
For Internal Use Only
wo#
Site Name:
Analyst: Date given to Reviewer:_
Matrix (circle): Aqueous / Other
Program (circle): Superfund / RCRA / WPD (NPDES) / SDWA / Other:
The signature below indicates the following:
• This data meets the needs of the customer according to the request.
• The analysis was performed as per the SOP, or exceptions documented.
• All documentation needed to recreate the analyses has been reviewed.
• Data Review status set to Peer Reviewed in Element.
Peer Reviewer signature
Date accepted
If any data for this case is stored with another case file, give Site Name and WO#_
Peer Reviewer Completes Section Below:
General:
Raw data is identified with sample IDs, site name,
WO#, analyst name, date of analysis.
Quality Control:
YES
NO N/A
Comments
NaCsI cal according to mfg recommendation within
year
Initial calibration: r2 > 0.99
Holding time: 14 days to analysis
Method Blank
-------
DRAFT R3QA239-xxxxxxxx
Glycol Analysis by HPLC/MS/MS
Effective Date: 2012
Page 15 of 15
Analyst ensures that the data case file is complete and accurate as per SOP R3QA-066:
Bench sheet or Work Order list Appropriate TV sheets / Certificates of Analysis
Sample Prep logs Element Peer Review report
Instrument run log Raw data
Standard/Reagent Prep log Data status set to analyzed
Additional Comments by Analyst on data issues:
------- |