vvEPA
United States
Environmental Protection
Agency
Environmental Monitoring
Systems Laboratory
P.O. Box 93478
Las Vegas NV 89193-3478
EPA 600/R-92/032
February 1992
Pre-Issue
Research and Development
Superfund Innovative
Technology Evaluation
(SITE) Report for the
Westinghouse Bio-Analytic
Systems Pentachlorophenol
(PCP) Immunoassays
-------
SUPERFUND INNOVATIVE TECHNOLOGY EVALUATION (SITE)
REPORT FOR THE WESTINGHOUSE BIO-ANALYTIC SYSTEMS
PENTACHLOROPHENOL (PCP) IMMUNOASSAYS
by
M. E. Silverstein, R. J. White, and R. W. Gerlach
Lockheed Engineering & Sciences Company
Las Vegas, NV 89119
and
J. M. Van Emon
Environmental Monitoring Systems Laboratory
Las Vegas, NV 89193-3478
o-
CN.
>~i Contract Numbers 68-03-3249 and 68-CO-0049
Work Assignment Manager
E. N. Koglin
Quality Assurance and Methods Development Division
Environmental Monitoring Systems Laboratory
Las Vegas, NV 89193-3478
ENVIRONMENTAL MONITORING SYSTEMS LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
LAS VEGAS, NEVADA 89193-3478
-------
NOTICE
The information in this document has been funded wholly or in part by the U.S. Environmental
Protection Agency under Contract Nos. 68-03-3249 and 68-CO-0049 to Lockheed Engineering &
Sciences Company. It has been subjected to the Agency's peer and administrative review and it has
been approved for publication as an EPA document. Mention of corporation names, trade names, or
commercial products does not constitute endorsement or recommendation for use.
-------
ABSTRACT
The results of the demonstration of two Westinghouse Bio-Analytic Systems (WBAS) immunoassay
technologies are described in this report. The immunoassays measure parts per billion concentrations
of pentachlorophenol in environmental water samples. The study was conducted under the Superfund
Innovative Technology Evaluation (SITE) Program.
The demonstration was designed to evaluate the ruggedness and utility of a semiquantitative
immunoassay field kit. The results obtained from the field kit were compared to those obtained from
a quantitative, high-sample-capacity plate immunoassay. Both techniques were compared to a
standard U.S. Environmental Protection Agency (EPA) gas chromatography/mass spectrometry
(GC/MS) procedure (EPA Method 8270) for pentachlorophenol determination. The demonstration
was performed at the MacGillis & Gibbs Superfund Site in New Brighton, Minnesota, a National
Priorities List site known to have ground water contaminated with pentachlorophenol. The
immunoassay demonstration was conducted jointly with a SITE demonstration of a bioremediation
technology. This technology was designed by BioTrol, Inc. (Chaska, Minnesota), to biodegrade
pentachlorophenol in aqueous matrixes and waste streams.
The results of the WBAS immunoassay demonstration support the conclusion that the field
immunoassay is a useful screening tool. Though the study's data quality objectives for accuracy and
precision were only partially met, most of the results were close to the expected concentrations.
Results verified that the method can provide qualitative or semiquantitative screening information.
Although the results were more variable than had been anticipated, the incorporation of additional
procedural precautions and carefully chosen quality control acceptance criteria for on-site analysis
could improve performance substantially. Both immunoassays produced results with a bias toward a
high concentration when compared to GC/MS, but the tendency was not large and may have been
partly due to loss during sample extraction (EPA Method 3510) prior to analysis by GC/MS. The
detection of structurally related compounds by the immunoassays may have also contributed to the
high bias. The results indicate that the plate immunoassay is an accurate and precise method for
quantitating pentachlorophenol in water.
This evaluation is submitted in partial fulfillment of Contract Nos. 68-03-3249 and 68-CO-0049 by
Lockheed Engineering &. Sciences Company under the sponsorship of the EPA.
111
-------
CONTENTS
Notice ii
Abstract iii
Figures viii
Tables x
Abbreviations and Acronyms xi
Acknowledgements xii
Executive Summary ." xiii
1 Introduction 1
Overview of the Immunoassay Program 1
Overview of the SITE Program 1
Overview of the WBAS SITE Demonstration 2
Project Organization 2
Scope of Document 3
2 Descriptions of Technologies 5
WBAS Kit Immunoassay 5
WBAS Plate Immunoassay 6
Previous Plate Immunoassay Laboratory Evaluation 9
Gas Chromatography/Mass Spectrometry Analysis 11
3 PCP Immunoassay Demonstration Design 12
Predemonstration Testing and Planning 12
Study Design for the Silt Demonstration and Evaluation 14
Sample Collection Procedures 15
Sample Tracking 17
Quality Assurance Design 17
QA/QC and Bioreactor Sample Analysis 19
Data Quality Objectives 19
Ensuring QA Design Conformity 20
Data Management 20
4 Method Results and Comparisons 22
Kit Immunoassay Comparisons ;.... 25
Kit Immunoassay Comparison to the GC/MS 25
Kit Immunoassay Comparison to the Plate Immunoassay 28
On-Site Kit Immunoassay Comparison to Off-Site
Kit Immunoassay 30
Overall Kit Immunoassay Comparison 30
-------
CONTENTS (continued)
Plate Immunoassay Comparisons 30
Plate Immunoassay Comparison to the GC/MS 30
Comparison of EMSL-LV to WBAS Plate
Immunoassay Analyses 32
Overall Plate Immunoassay Comparison 32
SAIC GC/MS to EMSL-LV GC/MS Comparison 34
5 Quality Assurance and Quality Control Results 36
Quality Assurance and Quality Control Sample Results 36
QA Audit Samples 37
QC Performance Samples 38
Duplicate and Method Split Samples 41
Negative Control Samples : 44
Field Blank Samples 44
Matrix Spike Samples 46
Assessment of Data Quality 46
Data Quality in Terms of Five Data Quality Elements 46
QA Problems and Resolutions 53
Changes to the QA Plan 54
Results of the On-Site Systems Audit 54
Additional QA/QC Observations and Conclusions 55
Changes in Optical Density Levels of Kit Immunoassay Standards 55
Kit Immunoassay Results-Hand-Drawn Versus
Computer-Calculated 55
Instrument Cross Calibration 55
6 Conclusions and Recommendations 59
Kit Immunoassay Conclusions and Recommendations 59
Kit Immunoassay Conclusions 59
Kit Immunoassay Recommendations 61
Plate Immunoassay Conclusions and Recommendations 63
Plate Immunoassay Conclusions 63
Plate Immunoassay Recommendations 63
Joint SITE Demonstration Conclusions 64
References 65
Appendices
A Silt Demonstration of Biological Treatment of Groundwater by
BioTrol, Inc., at a Wood Preserving Site in New Brighton, MN 67
vi
-------
CONTENTS (continued)
B Examples of Kit and Plate Immunoassay Standard Curves and
Sample Placement Layout for the Plate Immunoassay 78
C Data Qualifier Flags and Definitions Applied to Kit Immunoassay
Data During Data Verification 84
D Algorithms Used to Determine Pentachlorophenol Concentrations
in Samples Used in Method Comparisons of the Kit
and Plate Immunoassays 87
E Participating Personnel 91
Vll
-------
FIGURES
Number Page
1 Organizational structure for the SITE demonstration of the WBAS
immunoassays at the MacGillis & Gibbs Superfund Site 4
2 WBAS field kit immunoassay 6
3 WBAS kit immunoassay analysis steps for determining
pentachlorophenol in aqueous samples 7
4 Portable spectrometer used for the WBAS
kit immunoassay demonstration 8
5 WBAS plate immunoassay analysis steps for determining
pentachlorophenol in aqueous samples 10
6 Sample flow and analysis scheme for the SITE demonstration of the
WBAS immunoassays for pentachlorophenol 16
7 Pentachlorophenol concentrations for all influent samples
by all methods and analysis sites over time 23
8 Pentachlorophenol concentrations for all effluent samples
by all methods and analysis sites over time 24
9 Comparison of results from bioreactor samples analyzed for PCP on site
by the kit immunoassay and at SAIC by GC/MS 27
10 Comparison of results from bioreactor samples analyzed for PCP on site
by the kit immunoassay and at EMSL-LV by the plate immunoassay 29
11 Comparison of results from bioreactor samples analyzed for PCP at EMSL-LV
by the plate immunoassay and at SAIC by GC/MS 31
12 Comparison of results from bioreactor samples analyzed for PCP at EMSL-LV
and WBAS by the plate immunoassay 33
13 Comparison of results from bioreactor samples analyzed for PCP at EMSL-LV
and SAIC by GC/MS 35
viu
-------
FIGURES (continued)
Number Page
14 QAA sample concentrations (ppm PCP), determined by the
kit immunoassay, by analysis site over time 40
15 Mean versus difference in results from pairs of analyses for kit
immunoassay duplicate and split field samples diluted into
calibration range (ppb PCP) 42
16 Mean versus difference in results from pairs of analyses for kit
immunoassay duplicate and split field samples with dilution
factors applied (ppm PCP) 43
17 Optical densities of the low and high standards for the
on-site kit immunoassay analyses 57
18 Comparison of the results of sample concentrations determined
on-site by graph paper and straight-edge ruler plotting versus a least
squares fit of the same samples 58
IX
-------
TABLES
Number Page
1 Cross-Reactivity of Anti-Pentachlorophenol Antibodies 13
2 Analysis Site and Method Comparison Summary 26
3 Kit and Plate Immunoassay Bias Versus GC/MS Results :. 28
4 Performance Audit Sample Results for QAA Samples 38
5 Performance Audit Sample Results for QAB Samples 39
6 Kit Immunoassay QC Performance Sample Results 41
7 Kit Immunoassay Negative Control Sample Results 45
8 Field Blank Analyses Results 45
9 Kit Immunoassay Matrix Spike Results 47
10 Plate Immunoassay Matrix Spike Results 47
11 Types and Numbers of Field Samples Analyzed by Analysis Site 52
12 Summary of Quantifiable Data for the Kit Immunoassay 52
13 On-Site Systems Audit Checklist 56
14 Comparison of Method Performances for PCP Analysis in Aqueous Samples 62
-------
ABBREVIATIONS AND ACRONYMS
AFMMP
ANOVA
BATS
CLP
CV
DNP
DQO
EMSL-LV
EPA
GC
GC/MS
GLP
gpm
L
LESC
mL
MMTP
n
N
NC
nm
OD
OMMSQA
PAH
PCP
pH
ppb
ppm
QA
QAA
QAB
QAPjP
QC
RREL
SAIC
SARA
SAS
SITE '
SOP
WBAS
Advance Field Monitoring Methods Program
analysis of variance
Biological Aqueous Treatment System
Contract Laboratory Program
coefficient of variation
dinitrophenol
data quality objective
Environmental Monitoring Systems Laboratory-Las Vegas
U.S. Environmental Protection Agency
gas chromatography
gas chromatography/mass spectrometry
good laboratory practice
gallons per minute
liter
Lockheed Engineering & Sciences Company
milliliter
Monitoring and Measurement Technologies Program
number
normal
negative control
nanometer
optical density
Office of Modeling, Monitoring Systems and Quality Assurance
polynuclear aromatic hydrocarbon
pentachlorophenol
negative log of the hydrogen ion activity
parts per billion
parts per million
quality assurance
quality assurance performance sample, type A
quality assurance performance sample, type B
quality assurance project plan
quality control
Risk Reduction Engineering Laboratory
Science Applications International Corporation
Superfund Amendments and Reauthorization Act
Statistical Analysis System
Superfund Innovative Technology Evaluation
standard operating procedure
Westinghouse Bio-Analytic Systems
XI
-------
ACKNOWLEDGEMENTS
This demonstration and the subsequent preparation of this report required the services of personnel
from the U.S. Environmental Protection Agency, Environmental Monitoring Systems Laboratory
(Las Vegas, Nevada); the U.S. Environmental Protection Agency, Risk Reduction Engineering
Laboratory (Cincinnati, Ohio); Westinghouse Bio-Analytic Systems (Rockville, Maryland); Science
Applications International Company (Paramus, New Jersey, San Diego, California; and McLean,
Virginia); BioTrol Inc., (Chaska, Minnesota); and Lockheed Engineering & Sciences Company (Las
Vegas, Nevada). Names of individuals and their functional roles on behalf of these organizations are
included in Appendix E.
XII
-------
EXECUTIVE SUMMARY
This evaluation report presents the results of a demonstration designed to assess the capabilities of
two immunoassay technologies to measure pentachlorophenol (PCP) in water. The technologies, a
semiquantitative field kit immunoassay and a quantitative plate immunoassay, were both developed by
Westinghouse Bio-Analytic Systems (WBAS) of Rockville, Maryland. The demonstration was
conducted under the Monitoring and Measurement Technologies Program as part of the U.S.
Environmental Protection Agency (EPA) Superfund Innovative Technology Evaluation (SITE)
Program. The demonstration was conducted under the guidance of the EPA Environmental
Monitoring Systems Laboratory in Las Vegas, Nevada (EMSL-LV).
Immunoassays are analytical techniques based on protein molecules (antibodies). The binding of a
specific antibody to its target analyte can be used to quantitatively or qualitatively determine the extent
of contamination in environmental samples. Specific antibodies have been developed to detect single
analytes or groups of related compounds. The WBAS kit immunoassay, based on rabbit polyclonal
antisera adsorbed on 8-well, polystyrene, microtiter strips, has a reported detection limit for PCP of 3
parts per billion (ppb), a linear dynamic range of 3 to 40 ppb, and a total analysis time of 30 minutes
per sample. It requires minimal logistical requirements for on-site analyses. The WBAS 96-well plate
immunoassay, based on a rat monoclonal antibody, has a reported detection limit of 30 ppb and a
linear dynamic range of 50 to 400 ppb. It requires 3 hours of hands-on analysis time per plate (10 to
20 samples run in triplicate) and involves certain logistical considerations (e.g., a mobile laboratory)
for on-site analyses. A previous evaluation by the EMSL-LV compared the plate immunoassay to GC
results for PCP analysis: data from this SITE immunoassay demonstration complements the previous
study.
The WBAS kit immunoassay was demonstrated under field (on site) and laboratory (off site)
conditions to determine its ruggedness, reliability, and potential for use as a rapid, on-site, analytical
tool in the Superfund Program. The results obtained from the kit immunoassay analyses performed
on site and off site were compared to those generated off site by the plate immunoassay; both
immunoassay techniques were compared to standard EPA gas chromatography/mass spectrometry
(GC/MS) methods for the analysis of PCP (EPA Method 3510 sample extraction followed by EPA
Method 8270 analysis by GC/MS).
The on-site demonstration took place in July and August, 1989, at the MacGillis & Gibbs Superfund
Site in New Brighton, Minnesota, a National Priorities List site known to have ground water
contaminated with PCP. The immunoassay demonstration was coordinated through RREL and
conducted jointly with a SITE demonstration of a bioremediation technology designed by BioTrol,
Inc. (Chaska, Minnesota), to biodegrade PCP in aqueous waste streams. The design of the
immunoassay Silt demonstration involved several planning components: predemonstration tests, an
experimental'design, a sampling and analysis design, quality assurance and quality control (QA/QC)
planning, and data base management.
Xlll
-------
Field samples for the immunoassay demonstration were obtained from three sampling points in the
bioreactor system: influent samples collected before pretreatment (nutrient addition and pH
adjustment), influent samples collected after pretreatment, and effluent samples collected before
filtration. Samples were collected over three 1-week periods which coincided with a 1-, 3-, and 5-
gallon-per-minute flow rate of the ground water through the bioreactor. Composite and grab samples
were collected, homogenized, split, and analyzed on site with the kit immunoassay. Sample splits were
analyzed at the WBAS and EMSL-LV laboratories with the kit and plate immunoassays and with
GC/MS by Science Applications International Corporation in San Diego, California. Comparison of
the analysis results by each method and analysis site was a critical component in the evaluation of the
immunoassays.
A rigorous QA project plan was implemented at all sites involved in the study. This plan included the
analysis of a battery of QA/QC samples, including duplicate, split, matrix spike, QA audit, QC
performance, field blank, and negative control (NC) samples and cross-calibration standards. The
QA/QC samples were used to assess the performance characteristics of the two immunoassay methods
and to test the capabilities of the technologies to meet the stated data quality objectives (DQOs) of the
demonstration; the most critical DQO was that the immunoassay sample results had to be within a
factor of two (50 to 200 percent) of the GC/MS results. Traditional methods such as GC/MS have
interlaboratory biases of 30 percent or more in addition to other sources of variability. Thus, the use
of a factor of two for the immunoassay implies a slightly greater (but quite usable) variability than one
might expect from the more traditional methods. All bioreactor and QA/QC sample data from all
analysis locations were subjected to EMSL-LV QA review and verification. The data were then
entered and stored in a documented data base.
The immunoassay technologies were assessed by comparing the analyses of the bioreactor influent and
effluent samples. Because of the differences in sample ranges of the influent and the effluent samples,
results from these sample types were treated separately in data evaluation. The most critical method
and analysis site comparisons were: (1) the on-site kit immunoassay to the GC/MS, (2) the on-site kit
immunoassay to the plate immunoassay, and (3) the plate immunoassay to the GC/MS.
Results from the on-site kit immunoassay compared favorably to the GC/MS results. There was good
relative (rank order) agreement between the two methods. Fourteen of the 16 influent samples
analyzed on site were within the factor-of-two DQO over a concentration range of approximately 1 to
60 ppm PCP. The effluent samples analyzed by the two methods were in the same general
concentration range (kit immunoassay = 0.2 to 2.3 ppm; GC/MS = 0.008 to 0.9 ppm). Results of
influent and effluent samples indicated a consistent tendency for the kit immunoassay data to have a
high bias when compared to the GC/MS data. This bias may be due to extraction inefficiency of EPA
Method 3510, cross-reactivity of tetrachlorophenol in the immunoassay, or a combination of these and
other factors. Kit immunoassay results for influent samples averaged from 65 to 119 percent higher
than GC/MS results, depending on analysis site. Effluent sample bias was small in practical (ppm)
terms. The positive bias suggests that the kit immunoassay has a minimal tendency to generate false
negative responses.
The kit immunoassay results were compared to the plate immunoassay results to detect differences
between the methods and to provide insight for interpreting the performances of the immunoassays
compared to the GC/MS. There was reasonable agreement between the two immunoassay techniques;
27 of 38 (71 percent) on-site kit immunoassay influent sample results were within a factor of two of
xiv
-------
the plate immunoassay results (VVBAS and EMSL-LV analysis sites combined). For both
immunoassays, effluent samples were in the same general range (0.20 to 2.74 pprn; n = 38). Although
no significant bias was observed between the two immunoassay techniques, a significant amount of
scatter (variability) was observed.
Overall, the plate immunoassay results compared more favorably to the GC/MS than did the kit
immunoassay results. At one analysis site (EMSL-LV), 17 of 18 (94 percent) effluent sample results
were within a factor of two of the corresponding GC/MS results, while at the other site (WBAS), 12 of
18 samples were within this limit. The results from various QA/QC samples suggest that WBAS had
unusual site- or operator-specific factors affecting the quality of their analyses. As with the kit
immunoassay, the plate immunoassay exhibited a high bias when compared to the GC/MS, although
the bias was much smaller (17 to 40 percent for influent samples, depending on analysis site).
Data derived from the QA and QC samples provided insight into the intra- and intermethod
performance assessment in terms of the accuracy and precision of the kit and plate immunoassays.
Seventy-six percent of the audit samples and 74 percent of the bioreactor samples analyzed using the
kit immunoassay met the accuracy DQOs. The false negative rate was 2.6 percent (based on 76
effluent and influent sample analyses), and the false positive rate was 19 percent (based on 98 NC
samples). However, the matrix spike recoveries were unsatisfactory (-166 to +313 percent), a fact that
may be attributed to a poorly developed matrix spike protocol. Precision for the kit immunoassay was
not as good as expected. The coefficients of variability for QC performance and QA audit samples
exceeded the DQO of ±50 percent in most cases; however, results of the duplicate and split sample
analyses were reasonably good for a semiquantitative method.
Ninety-five percent of the audit samples and 81 percent of the bioreactor samples met the accuracy
DQO for the plate immunoassay. There were no false negatives (based on 78 effluent and influent
sample analyses) and no false positives (based on 21 NC samples). The matrix spike recoveries were
less than satisfactory (41 to 169 percent), but were considerably better than for the kit immunoassay
spike recovery results. Overall, precision for the plate immunoassay method was better than the kit
immunoassay method. Better precision and accuracy for the plate immunoassay was not surprising
because the kit immunoassay was designed to be a semiquantitative method while the plate
immunoassay was expected to be quantitative.
The WBAS kit immunoassay proved to be a useful and promising technology that can provide on-site,
real-time, cost-effective, semiquantitative data with a low risk of generating false negative responses.
The kit immunoassay, which is easy to learn and perform in the field, can be an effective field
screening method at Superfund sites known to have PCP-contaminated water. The plate
immunoassay exhibited better precision and accuracy than the kit immunoassay, with quantitative
results closer to those generated by the GC/MS. The plate immunoassay can be readily set up in a
field laboratory, and its sample throughput is greater than that of the kit immunoassay. The SITE
demonstration indicated that the WBAS kit and plate immunoassay technologies can provide effective
screening capabilities in the field and can be used to complement conventional laboratory methods for
measuring PCP in aqueous samples. The demonstration also underscored the need for continued
QA/QC guidelines and protocol development to improve and fully characterize the quality of
immunoassay data. Both WBAS immunoassays evaluated in this report showed promise as
measurement and monitoring tools at hazardous waste sites.
xv
-------
SECTION 1
INTRODUCTION
The performance of two immunoassay methods was assessed during a U.S. Environmental Protection
Agency (EPA) Superfund Innovative Technology Evaluation (SITE) Program demonstration. The
methods are semiquantitative and quantitative immunoassay techniques developed by Westinghouse
Bio-Analytic Systems (WBAS) of Rockville, Maryland, to detect pentachlorophenol (PCP) in water.
The immunoassays were demonstrated under field (on site) and laboratory (off site) conditions and
were compared to a standard EPA gas chromatography/mass spectrometry (GC/MS) method.
OVERVIEW OF THE IMMUNOASSAY PROGRAM
The EPA Environmental Monitoring Systems Laboratory at Las Vegas, Nevada (EMSL-LV), is
responsible for developing and evaluating immunoassays for specific environmental applications.
According to EPA guidelines for methods evaluations, this process requires the determination of
performance parameters such as precision, within- and among-laboratory biases, berween-method bias,
method detection limits, interferences, and ruggedness of the method.
To be effective as rapid screening tools, immunoassays must provide timely, cost-effective results that
complement conventional analytical methods. They must be capable of measuring the target analyte
with sufficient accuracy and precision to identify that samples are clearly above or below a critical
concentration range.
OVERVIEW OF THE SITE PROGRAM
The Superfund Amendments and Reauthorization Act of 1986 (SARA) mandated that the EPA
develop timely and cost-effective remedies at National Priorities List (i.e., Superfund) sites. As part of
the response to this mandate, the EPA established SITE, "a program of research, evaluation, testing,
development, and demonstration of alternative or innovative treatment technologies....which may be
utilized in response actions to achieve more permanent protection of human health and welfare and
the environment" (SARA, 1986). The Silt Program is comprised of two innovative technology
categories. The first category includes the demonstration of alternative treatment technologies that
can be used in Superfund site remediation. These activities are administered by the EPA Risk
Reduction Engineering Laboratory (RREL) at Cincinnati, Ohio. The second category is for the
evaluation of measurement and monitoring techniques that can withstand the rigors of field
conditions. This portion of the Silt Program, the Monitoring and Measurement Technologies
Program (MMTP), is administered by the Advanced Field Monitoring Methods Program (AFMMP)
of the EPA Office of Modeling, Monitoring Systems, and Quality Assurance (OMMSQA). The
Environmental Monitoring Systems Laboratory at Las Vegas, Nevada, is the lead laboratory for the
-------
AFMMP. The immunoassay procedures evaluated in this report represent the first measurement and
monitoring technologies to be evaluated under MMTP in the SITE Program.
OVERVIEW OF THE WBAS SITE DEMONSTRATION
The WBAS SITE demonstration was conducted primarily to assess the ruggedness and utility of a
semiquantitative immunoassay field analysis technique for rapid sample screening. The demonstration
also provided the opportunity to further evaluate a quantitative 96-weil, microtiter plate immunoassay
(see Section 2 for a discussion of the previous evaluation). The two immunoassay methods were
developed by WBAS to detect PCP in water under field and laboratory conditions. These techniques.
referred to in this report as "kit immunoassay" and "plate immunoassay," respectively, were compared
to each other and to a standard EPA GC/MS method for detecting PCP in water.
The demonstration took place from July 23 to August 29, 1989, at the MacGillis & Gibbs Superfund
Site at New Brighton. Minnesota. This location was well suited for the evaluation of the WBAS
immunoassays for several reasons. Ground water at the site was contaminated with PCP and
polynuclear aromatic hydrocarbons (PAHs) as the result of a wood preservative treatment operation.
In addition, a SITE demonstration at the MacGillis &. Gibbs site had been previously planned for the
summer of 1989. RREL was demonstrating the BioTrol Aqueous Treatment System (BioTrol, Inc.,
Chaska. Minnesota), a biological reactor (bioreactor) designed to biodegrade PCP and PAHs in
aqueous media into carbon dioxide, water, and sodium chloride (Stinson et al., 1991, and Appendix A).
In an effort to minimize complications involved in adding a design to one that was already in place,
the sampling and analysis scheme of the WBAS immunoassay demonstration was constructed around
the design of the bioreactor demonstration.
PROJECT ORGANIZATION
The success of the immunoassay evaluation depended on the coordinated efforts of four primary
organizations: EMSL-LV, WBAS, RREL, and BioTrol, Inc.
EMSL-LV, with assistance from its contractor, Lockheed Engineering &. Sciences Company (LESQ,
was responsible for:
• Designing, overseeing, and ensuring the implementation of the elements of the
demonstration and quality assurance (QA) plan.
• Acquiring the necessary confirmatory data.
• Performing off-site analysis by both immunoassay techniques and by GC/MS.
• Preparing and distributing QA and quality control (QC) samples.
• Evaluating and reporting on the performance of the technologies.
WBAS, the developer of the immunoassays, was responsible for:
•,
• Performing preliminary testing to assess kit immunoassay performance.
-------
• Supplying a sufficient number of field kits and plate immunoassay reagents to perform the
analyses required to conduct the demonstration.
• Providing technical assistance to the on-site personnel using the kit immunoassay.
• Performing off-site analyses by both immunoassay techniques.
RREL, through its contractor. Science Applications International Corporation (SAIC), Paramus. New
Jersey, conducted the BioTrol. Inc., bioreactor demonstration. RREL and SAIC were responsible for
the following aspects of the immunoassay demonstration:
• Performing the on-site kit immunoassay analysis on designated samples.
• Providing logistical support, including field sample collection, processing, and shipment.
• Analyzing samples by GC/MS and reporting the results.
For the immunoassay demonstration. BioTrol. Inc., was responsible for:
• Providing predemonstration test samples.
• Providing technical assistance.
Figure 1 shows the organizational structure of the immunoassay demonstration.
SCOPE OF DOCUMENT
This document includes descriptions of the methods and operating theories of the two WBAS
immunoassay technologies (Section 2), designs of the Silt demonstration and QA plans (Section 3),
and comparisons of the immunoassay methods to a standard EPA method (Section 4). The results of
intra- and intermethod and QA/QC performances are described in Section 5. Conclusions and
recommendations about the WBAS immunoassay technologies are discussed in Section 6.
-------
SITE PROGRAM
i
REMEDIATION
TECHNOLOGIES
RREL
(SAIC)
BIOREACTOR
DEMONSTRATION
(BIOTROL, INC.)
EVALUATION OF
THE TECHNOLOGY
MONITORING &
MEASUREMENT
TECHNOLOGIES
>
r
EMSL-LV
(LESC)
IMMUNOASSAY
DEMONSTRATION
(WBAS)
i
EVALUATION OF
THE TECHNOLOGY
Figure 1. Organizational structure for the SITE demonstration of the WBAS immunoassays at the
MacGillis & Gibbs Superfund Site.
-------
SECTION 2
DESCRIPTIONS OF TECHNOLOGIES
Immunoassays are analytical techniques based on protein molecules called antibodies. The binding of
a specific antibody to its target analyte can be used to detect or quantitate contamination in
environmental samples. Specific antibodies can be developed to detect either a single analyte
(e.g., compound) or groups of related compounds. Plate immunoassays have a high sample capacity,
are quantitative in nature, and may take 3 to 4 hours of preparation and analysis time to run. Kit
immunoassays, on the other hand, are usually designed for rapid qualitative or semiquantitative on-site
measurements. Both immunoassay technologies are portable, though plate immunoassays may require
more logistical and equipment considerations for use in the field. Immunoassay techniques are quite
versatile and can be applied to many analytical situations. The WBAS kit and plate immunoassay
techniques and their theories of operation are discussed in the following sections.
WBAS KIT IMMUNOASSAY
The WBAS immunoassay kit (Figure 2) is designed for on-site qualitative or semiquantitative
determination of PCP in water. In this evaluation, the kit immunoassay was used to provide
semiquantitative results. The test, which is based on rabbit polyclonal antisera, has a reported
detection limit of 3 pans per billion (ppb) and a total analysis time of approximately 30 minutes.
The method has a linear dynamic range of 3 to 40 ppb.
The kit immunoassay is performed in an 8-well, polystyrene, microtiter strip that is coated with anti-
PCP rabbit polyclonal antibody. Each well has a volume capacity of approximately 0.25 mL. Figure 3
depicts the steps required for analysis by the kit immunoassay. Calibration standards of 3.0, 7.1, 16.9,
and 40.0 ppb PCP are used in four of the eight wells. Environmental water and QA/QC samples are
diluted into the range (i.e., 3 to 40 ppb) for quantitation estimates in the four remaining wells. Fifty
/*L of sample or standard are placed in a well. Fifty pL of enzyme-labeled PCP analog (PCP-
Peroxidase) are then added Competition between PCP in the sample and PCP-Peroxidase occurs for
binding to the immobilized antibody in the well. After a 15-minute incubation, all unbound materials
are removed by washing with a buffer rinse. Next, 100 jiL of a 1:1 mixture of an enzyme substrate
(H2Oj) and chromogen (S^'.S.S'-tetramethylbenzidine) are added. The immobilized enzyme acts on
these substances, producing a colored end product. Fifty ^L of 0.3 N sulfuric acid are added to stop
further color production. A portable spectrometer (manufactured by Hyperion, Inc., for Dynatech
Laboratories, Inc.)(Figure 4), is used to measure the optical density (OD) of the colored end product
at 405 nm (Figure 4). Since the quantitarion is based on competition for antibody, the color intensity
is inversely proportional to the analyte concentration in the sample. Quantitation of PCP is
determined by using a standard curve. These estimates can be determined by manually plotting the
four calibration standards on serai-log graph paper, or the standard curve can be fitted by a best-fit
-------
line generated with a programmable calculator.
immunoassay are given in Appendix B.
Examples of hand-plotted standard curves for the kit
WBAS PLATE IMMUNOASSAY
The WBAS plate immunoassay for quantitation of PCP in water samples is also a competitive
inhibition enzyme immunoassay. A 96-well, polystyrene, microtiter plate coated with
2,6-dichlorophenol-protein (DCP-protein) conjugate is used for the solid phase. The free analyte in
the sample and the bound conjugate compete for binding to the anti-PCP antibody in solution. An
enzyme-labeled secondary antibody is used to quantuate the amount of anti-PCP antibody bound to
the 2.6-dichlorophenol-protein conjugate on the solid phase. The amount of bound anti-PCP
PENTACHLOROPHENOL
FIKLD ANALYSIS KIT
Figure 2. WBAS field kit immunoassay.
-------
\ / \
v
PCP-specific antibody is
passively adsorbed in the
wells of a polystyrene
microtiter strip.
2.
Enzyme-labeled PCP and PCP
from an environmental sample
compete for binding sites on
the antibody.
E E
3.
YYY
A buffer rinse removes all
unbound materials. Both the
enzyme-labeled PCP and PCP
from the sample are now bound
to the antibody.
o oo o
°0 00 O on°
E E
0*0*.
° • o • o o
O «O •
o • a • •
YYYY YYYY YYYY
C^-S-VXS:S:^^NX>X^>^XVX^^ 5. t^^c^^^^^cs^^^xs^c•^^^^^^^^x>^^f?^SS 6. cs^^^^^^^^^x^^^s^^^^x:^^cs^^^^^^^
A chromogenic substrate is
added to react with the bound
enzyme.
The reaction between the
enzyme and substrate produces
a colored end-product.
After stopping the reaction,
the absorbance is measured
by a spectrometer to quantify
the unknown amount of PCP
in the sample.
PCP-Speci
-------
Figure 4. Portable spectrometer used for the WBAS kit immunoassay demonstration.
-------
antibody is inversely related to the amount of PCP in the sample. In the final procedural step, an
enzyme substrate is added, and the colored end product OD is read with a plate reader
(HP/Genenchem3) at 405 nm. The intensity of color is inversely proportional to the concentration of
analyte in the sample. Figure 5 shows the analysis steps described above for the plate immunoassay.
The plate immunoassay has a detection level of approximately 30 ppb PCP and a linear dynamic
range from 50 to 400 ppb. If the extraction procedure in EPA Method 604 (U.S. EPA. 1982) is used
on the water samples, the minimum detectable level is below 1 ppb. The plate immunoassay
procedure involves an overnight incubation step to generate a plate coating (needing about 10 minutes
of the analyst's time); however, the actual analysis time is only about 3 hours. The method has a high
sample throughput because 10 to 20 samples can be run in triplicate on each plate, and several plates
can be run in batches in one day. An example of the sample placement layout on a typical plate
analyzed in the SITE evaluation is given in Appendix B.
Results from the 96-well plate immunoassay are calculated using a 4-parameter curve fit generated by
the TiterCalca data analysis program. Results were quantitated using the approximately linear region
of the standard curve (i.e., the region between 10 percent and 90 percent OD range). An example of a
typical standard curve is given in Appendix B. Data generated by EMSL-LV and WBAS from the
plate immunoassay were collected in hard copy form and on floppy diskettes using the TiterCalc*
software.
Previous Plate Immunoassav Laboratory Evaluation
A methods comparison was conducted at EMSL-LV prior to the SITE demonstration (Van Emon
and Gerlach, 1990) comparing the plate immunoassay and a gas chromatography (GC) detection
protocol (EPA Method 604 [U.S. EPA, 1982]). This study used environmental water samples (i.e.,
drinking water, surface water, and ground water) spiked with PCP at various concentrations to
evaluate the technology. Extracts were prepared following protocols in EPA Method 604 and
quantitated by both the plate immunoassay and GC for comparison. Extracts from a simple solid-
phase extraction technique, developed by WBAS, were also analyzed by the plate immunoassay and by
GC In addition, unextracted samples were analyzed directly by the plate immunoassay (direct plate
immunoassay) as the method does not require an extraction for relatively clean samples (i.e., not silty).
The direct plate immunoassay data were compared to the GC results obtained with the solid-phase
and EPA Method 604 extracts.
The results of this previous evaluation showed no practical difference between: (1) the plate
immunoassay and GC detection of Method 604 extracts, (2) the plate immunoassay and GC detection
of solid-phase extracts, (3) immunoassay results from WBAS and EMSL-LV, following the WBAS
solid-phase or EPA Method 604 extraction protocols, and (4) precision of the direct plate
immunoassay obtained by WBAS and EMSL-LV. Overall, this study generated a 9 percent false
positive rate (based on blank sample analysis; n = 115) and a 0 percent false negative rate (based on
spiked environmental water samples; n = 192). It is important to emphasize that the plate
immunoassay could be performed directly on the environmmental water samples without an
extraction, although the direct method had a higher variability than immunoassay following either
extraction technique. Thus, the method could be performed in a field laboratory as a quantitative,
high-sample capacity, analytical methodology.
-------
i.S
The DCP-protein coating antigen is
passively adsorbed in the wells of
a polystyrene microtiter plate. PCP
from the sample and PCP-specific
antibody are added.
The PCP in the sample and the
immobilized DCP-protein coating
antigen compete (or binding
sites on the antibody.
A buffer rinse removes all
unbound materials. Only the
antibody bound to the
immobilized DCP-protein
coating antigen remains.
An enzyme-labeled conjugate is
added and binds only to the
PCP-specific antibody.
000000
5.
After a buffer rinse removes all
unbound enzyme-labeled
conjugate, a chromogenic
substrate is added.
o o • <
• o • o o,
The reaction between the enzyme
and chromogenic substrate
produces a colored end-product.
The absorbance is measured by
a spectrometer to quantify
the amount of PCP in the sample.
Y
PCP-Specific Antibody
DCP-Protein Conjugate
(Coating Antigen)
•• Enzyme-Labeled
Second Antibody
(IgG Alkaline
Phosphates)
PCP (Standard or
from Sample)
O = Chromogenic Stubstraie
• = Colored End-Product
<£ = Light Path
Polystyrene Microtiter
Plate
Figure 5: WBAS plate immunoassay analysis steps for determining pentachlorophenol in aqueous
samples.
10
-------
GAS CHROMATOGRAPHYMASS SPECmOMETRY ANALYSIS
The GC/MS method used in the WBAS immunoassay SITE demonstration was performed in two
main steps: (1) sample preparation and (2) analysis by GC/MS. Samples were liquid extracted
according to EPA Method 3510 (OSWER. 1986), a procedure for isolating and concentrating organic
compounds from aqueous samples. Sample extracts were then analyzed using EPA Method 8270
(OSWER. 1986), a GC/MS method for semivolatile organics, including PCP and PAHs. This method
is designed for analysis of extracts prepared from all types of solid waste, soil, and ground-water
matrices. Method 8270 contains detailed analysis instructions, QC guidelines, and performance data
for PCP analysis by GC/MS. The practical quantitation limit of the method is 50 ppb for PCP in
ground water. In addition, the method states that the experience of the analyst performing the
GC/MS analyses is invaluable to the success of the method (OSWER. 1986). These methods were
chosen as the most appropriate for the goals of the bioreactor demonstration because a wide spectrum
of PAHs were of interest in addition to PCP. This particular GC/MS method was one of several valid
EPA techniques available for the analysis for PCP.
11
-------
SECTION 3
PCP IMMUNOASSAY DEMONSTRATION DESIGN
By conducting the immunoassay demonstration in conjunction with the BioTrol, Inc.. bioreactor
demonstration, the EPA was presented with an excellent opportunity to simultaneously test the
effectiveness of remediation and monitoring and measurement technologies. The bioreactor
demonstration had been in the negotiation and planning stages for a SITE demonstration far in
advance of the plan to evaluate the immunoassay. In an effort to minimize complications, the
sampling and analysis scheme of the WBAS immunoassay demonstration was constructed around the
activities planned for the bioreactor demonstration.
The design of the SITE demonstration to evaluate the capabilities and limitations of the WBAS kit
and plate immunoassays included several components. Predemonstration testing provided insights
that were incorporated into the final design. The immunoassay sampling and QA plans were
structured to ensure the collection of enough data to make the necessary method and statistical
comparisons and to assess the quality of the data. In addition, a data management system was
developed to ensure reliable collection, integration, analysis, presentation, and storage of the data
generated during this Silt demonstration.
PREDEMONSTRATION TESTING AND PLANNING
From late May through mid July of 1989, WBAS conducted a variety of preliminary performance
checks on the kit immunoassay in a controlled laboratory environment. The preliminary checks were
valuable in finalizing the overall design of the immunoassay demonstration. Performance data were
generated using specified concentrations of PCP spiked into laboratory-grade water, ground water, and
bioreactor-matrix water samples. The ground-water and bioreactor samples were provided to WBAS
by BioTrol, Inc. These samples were not collected from the MacGillis & Gibbs site, but were used to
simulate samples that would be obtained during the demonstration.
The predemonstration tests provided insight into such operating and data quality parameters as:
linear dynamic range of the calibration curve; matrix effects and interferences, especially for the
effluent samples; QC checks (e.g., sample dilution, pretreatment, and procedural precautions);
estimates of precision from replicate and dilution analyses; bias resulting from different spectrometers;
and the stability of PCP in the field.samples. The conclusions obtained from the predemonstration
tests are presented below.
• Cross-reactivity--A variety of structurally related compounds were tested for their cross-
reactivity in the plate and kit immunoassays. The results are summarized in Table 1. For
.the plate immunoassay, a 42 percent response relative to PCP was found for 2,3,5,6-
tetrachlorophenol, and two trichlorophenols yielded about a 10 percent relative response.
12
-------
Other tested compounds gave much less or negligible responses. Similarly, the highest
cross-reactivity for the kit immunoassay was 19 percent for 2,3,5,6-tetrachlorophenol.
Several trichlorophenols showed cross-reactivities in a range of 7 to 11 percent. As with
the plate immunoassay, much less cross-reactivity was observed for the other tested
compounds.
TABLE 1. CROSS-REACTIVITY OF ANTI-PENTACHLOROPHENOL ANTIBODIES
Percent cross-reactivity2
Compound
2,3,5 ,6-Tetrachlorophenol
2,4,6-TrichlorophenoI
2,3,6-TrichIorophenol
2,6-Dichlorophenol
Tetrachlorohydroquinone
2,3,4-Trichlorophenol
2,3,5-Trichiorophenol
2,4-Dichlorophenoi
2.5-Dichlorophenol
3.5-Dichlorophenol
3,4-Dichlorophenol
2,3-Dichlorophcnol
4-Chlorophenol
Phenol
Pentachloroaniline
Pentachlorobeozeae
2,3-Dinitrotolucnc
2,4-Dinitrotolucnc
2,4.5-Trichloronitrobenzene
Plate
42.0
110
8.8
1.8
0.8
OJ
0.5
0
0
0
0
0
0
0
0
0
0
0
0
Kit
19
7
7
0.4
0.7
11.0
2J
N/A
0.1
0.1
0.1
0.1
0.2
0.1
0.1
N/A
0.1
0.1
0.1
a[IC50PCP/IC<0 compound] x 100, where ICj0 is the molar concentration of compound that inhibits 50 percent antibody binding
in immunoassays.
Source: Courtesy of WBAS
• Calibration standards—A series of calibration standards were used to evaluate the linear
dynamic range of the kit. Analysis of the standards revealed that the linear part of the
curve was from 3 to 40 ppb instead of the 3 to 100 ppb range that was originally indicated
by WBAS. The narrowing of the linear range changed a variety of design components,
including matrix spiking concentrations, sample dilution schemes, and protocol
specifications.
• Replicate analyses-Replicate analyses (n = 12) of 4 ppb and 30 ppb PCP standards,
representing the low and high portions of the linear dynamic range, were used to evaluate
kit immunoassay accuracy and precision. The 4 ppb standard yielded a mean
concentration of 5.2 ppb (15 percent coefficient of variation [CV]), and the 30 ppb
standard yielded a mean concentration of 44 ppb (36 percent CV). The results of these
, analyses indicated the potential for a high bias.
13
-------
• Holding time study-Results of a holding time study, which involved the storage of splits of
influent and effluent samples for 6 days at room temperature and at 4 °C. indicated that
changes caused by this temperature storage differential were negligible.
• Range-finding and dilution checks-Field samples were diluted in order to evaluate the
expected range of PCP concentrations. Three serial dilutions were made of an influent
sample containing approximately 18 ppm PCP so that the dilution results fell on three
areas in the linear calibration range. The assessment of the results of these range-finding
and dilution checks proved useful in assessing the approximate dilution factors to be used
during the demonstration.
• Pipetting Accuracy-WBAS discovered a high degree of inaccuracy in the squeeze dropper
bottles that were supplied with the kit immunoassay. To rectify the problem, WBAS
suggested the use of a mechanical pipettor (Rainin8) in the kit analysis.
• Blank samples-Blank, or negative control, samples distributed over 12 strips were used to
estimate solid phase variability (i.e.. the variability among microtiter wells and strips).
Results indicated that there was approximately an 11 percent CV on these measurements.
• Matrix spike analyses-On two separate days, matrix spike analyses were performed on
four effluent samples diluted to 10 ppb and spiked with 15 ppb PCP. This spike level was
chosen so that all samples would fall within the linear range of the method. The mean
recoveries of the analyses for each day were 163 percent and 120 percent, respectively.
These results indicated a potential problem with matrix spike analysis for the kit
immunoassay, and problems with matrix spike analysis were encountered in the
demonstration (see Section 5).
Based on the above predemonstration data and on verbal and written information provided by WBAS,
RREL, SAIC, and BioTrol, Inc., EMSL-LV developed demonstration and QA plans (Silverstein et al.,
1991) and field instructions. EMSL-LV also prepared, verified, and distributed all QA and QC sample
materials. WBAS conducted a 1-day training session for the SAIC field personnel designated to
perform the on-site kit immunoassay analyses.
STUDY DESIGN FOR THE SITE DEMONSTRATION AND EVALUATION
The field samples for the Lmmunoassay SITE demonstration were collected in three 1-week (6-day)
intervals, alternating weekly for a period of 6 weeks. Samples were collected and split on site; one
split was analyzed at the field site by the kit immunoassay. The on-site analyses were performed in a
field trailer at the MacGillis & Gibbs site. This trailer was also used for various sample preparation
and shipment activities for both the immunoassay and bioreactor demonstrations. Split samples were
shipped to the EMSL-LV and WBAS laboratories for analysis by both the kit and plate immunoassay
methods. Sample splits were also sent for analysis by GC/MS at the SAIC laboratory in San Diego,
California. In addition, a selected group of samples was analyzed by GC/MS at the EMSL-LV facility.
EPA Method 8270 (OSWER, 1986), after sample extraction by EPA Method 3510 (OSWER, 1986),
was used as the confirmatory and comparative method. The sample flow and analysis scheme is
presented in Figure 6.
14
-------
The MacGillis & Gibbs site was ideal for the application of the immunoassay technologies because the
ground water was known to contain high levels of PCP (approximately 50 ppm). The ground water
was also relatively clean, requiring no cleanup before entering the bioreactor. Although some samples
were turbid, colored yellow or brown, or had a precipitate, no solid phase or centrifugation cleanup
was required for the immunoassay analyses. The bioreactor was operational for 6 weeks, which were
divided into three 2-week periods that coincided with changes in flow rate of the ground water
pumped through the bioreactor. Flow rates of 1, 3, and 5 gallons per minute (gpm), or periods A, B,
and C, respectively, delineated these 2-week periods. Samples were collected for the immunoassay
demonstration 6 days within each flow period (18 total days) in order to get a diverse set of field
samples while minimizing logistical effects on the bioreactor SITE demonstration activities. The
EMSL-LV and WBAS laboratories only analyzed samples from the first and third (A and C) weeks of
the demonstration by kit immunoassay. There was a concern regarding a potential shortage of
reagents due to the short lead time of the study. As a result, about one-third fewer samples were
available for comparison. For various logistical reasons, not all of the analysis sites analyzed all
samples. However, enough samples were analyzed by each method and at each site to conduct the
necessary method comparisons between the kit immunoassay, the plate immunoassay, and the
established GC/MS method (see sections 4 and 5).
The PCP-contaminated ground water was pumped into the bioreactor and discharged as treated
effluent. Data from the site characterization showed that the raw ground water contained
approximately 50 ppm PCP. Preliminary data from the bioreactor indicated the treated effluent would
contain 1 ppm or less. Two daily sampling points were selected at the entry and exit points of the
bioreactor. A diagram of the bioreactor and the sampling points is shown in Figure 2 of Appendix A
The first daily sampling point is denoted as "No. 2" in the figure. Samples collected at this point
consisted of ground water, which had been conditioned by pH adjustment with NaOH (to an
approximate pH of 7.3) and nutrient addition (nitrogen and phosphorous compounds) to enhance
bacterial growth inside the bioreactor. The samples collected at the second daily sampling point, "No.
5" in Figure 2 of Appendix A, consisted of bioreactor effluent before carbon filtration and removal of
solids. For this report, these samples are termed "influent" and "effluent" samples, respectively.
Many of the effluent samples were somewhat turbid, presumably as a result of the biological and
chemical processes and constituents inside the bioreactor. Centrifugation of effluent samples was
considered in the original design of the demonstration, but this option was abandoned in order to
avoid bias between the immunoassay and the GC/MS results. In addition, a weekly grab well water
sample, termed "raw influent," was collected (i.e., prior to pH and nutrient conditioning). The raw
influent sample was collected in bulk as a grab sample from a T-tap located forward of the
conditioning tank (sampling point "No. 1", Figure 2 of Appendix A). A total of 18 influent, 18
effluent, and 3 raw influent bioreactor samples were collected for the immunoassay demonstration. In
addition, 18 field blank samples were analyzed to evaluate the effectiveness of the decontamination
method for the sample collection apparatus and vessels.
Sample Collection Procedures
The influent and effluent samples were collected with separate automatic composite sampling devices
in 150-mL portions every 20 minutes over a 24-hour period. After the bulk sample (approximately
13 L) was .collected, it was split into homogenous subsamples for on-site analysis and shipment to the
EMSL-LV, WBAS, and SAIC laboratories. In addition, field blank samples were collected after daily
15
-------
Well
Holding
Tank
Bioreactor
Effluent
Samples
Collected
Weekly
Each Sample Split On Site
On Site
New Brighton, MN
Sample* Anal/zed by
Kit Immunoassay
WBAS
Rockvllle, MO
Samples Analyzed
by Kit and Plate
Immunoassays
EMSL
Las Vegas, NV
Samples Analyzed
by Kit and Plate
Immunoassays
and by GC/MS
SAIC
San Diego, CA
Sample* Analyzed
by GC/MS
• POINT OF SAMPLE COLLECTION
Figure 6. Sample flow and analysis scheme for the SITE demonstration of the WBAS
immunoassays for pentachlorophenol.
16
-------
decontamination of the composite sampler. They consisted of reaeent-erade water poured into a clean
(i.e.. decontaminated) composite jug, pumped through the sampler and collected directly into the
appropriate bottles. Field blanks were collected at influent and effluent sampling points on alternate
days.
Samples collected for immunoassay analyses were prepared in the same manner as the sarapics for
GC/MS analysis, except a smaller volume was collected for immunoassay. Sample splits were prepared
in amber glass bottles with Teflon-lined, screw caps in 30-mL, 250-mL, and 4-L volumes for the kit
immunoassay, plate immunoassay, and GC/MS analyses, respectively. All samples were stored at 4 "C
prior to on-site analysis or shipment. Field samples were packed with custody tape wrapped around
the neck and cap of each container, wrapped with insulation, and shipped in coolers maintained at
4 ° C. To maintain a record of sample collection, transfer, shipment, and receipt, a chain-of-custody
form was filled out for each shipment to each analysis location. In addition, EMSL-LV prepared and
shipped semiblind (audit) and known (QC) performance samples and matrix spiking solutions to the
Geld and the off-site laboratories.
Sample Tracking
Sample tracking was accomplished by assigning each sample a unique identification number as it was
collected. This number traced the sample day, time, and point of collection. Along with other
information on the label, the sample number provided the tracking information for samples analyzed
on site and off site. This numbering and documentation system proved invaluable when verifying data
after the field operations were completed. Because of the complexity of the study and QA designs of
the demonstration, a series of sample codes were also used to identify the samples for the purpose of
assessing data quality (Silverstein et al., 1991).
QUALITY ASSURANCE DESIGN
QA planning was a critical element of the study design. To ensure that data from this demonstration
were of known quality and representative of typical conditions, a QA project plan (QAPjP) was
prepared and followed. The QAPjP addressed the key elements required for Category II projects
(U.S. EPA, 1987) and enabled analysts to make data quality and performance estimates, conclusions,
and recommendations. Sections 4, 5, and 6 present these results.
Influent and effluent samples provided valuable QA data because they were homogenous subsampie
splits (field splits) that were sent to each analysis site. A variety of QA and QC samples were also
used in this demonstration. These are described below.
• Duplicate and method split samples-Duplicate samples are defined in this study as
. bioreactor samples that were divided into two separate subsamples at an analysis site prior
to being diluted into calibration range for analysis. Method split samples are defined as
bioreactor samples that are first diluted into calibration range and then split for analysis.
The purpose of these samples was to assess the variability associated with sample dilution
efficiency (with the duplicate samples) as compared to the variability associated with
analyzing an already diluted sample (with the method split samples) in two separate wells
i (for the kit immunoassay) or sets of wells (for the plate immunoassay). (NOTE: The
17
-------
above method splits should not be confused with the field splits, which were subsamples of
the original bulk sample.)
Matrix spikes-Matrix spike analysis consists of adding a known quantity of analyte to a
sample and determining the amount of analyte recovered from the spiked sample with
respect to that found in the original sample. The net measured change in concentration is
compared to the expected concentration change. Influent and effluent samples were
diluted into the calibration range and analyzed with and without the addition of a matrix
spike. For the kit immunoassay, the PCP spike level was 15 ppb; for the piate
immunoassay, the PCP spike level was 240 ppb. Spiking concentrations were chosen based
on the linear range of each method. The matrix spike recovery data for each immunoassay
was intended to provide information about matrix interferences in the samples after their
- dilution to the appropriate concentration range (Section 5). The acceptable recovery
window of ±25 percent about the spike amount for the plate immunoassay was taken from
the standard operating procedure for the method. The ±50 percent window for the kit
immunoassay was considered reasonable for a screening method and was the limit chosen
in the previous plate immunoassay evaluation (Van Emon and Gerlach, 1990).
QA audit samples-These were QA standard solutions that contained specified PCP
concentrations for analysis by kit immunoassay, plate immunoassay, and GC/MS for all
locations. These audit samples were considered semiblind because the analyst knew they
were audit samples and which dilution factors to use in preparing them for analysis, but did
not know the expected concentration of PCP. Two audit sample formulations were used.
One solution contained only PCP, noted as QAA; the other was a mixture of PCP and
other phenols, noted as QAB. The QAA audit sample had a nominal concentration of 25
ppm PCP and was prepared by dilution from EMSL-Cincinnati QA reference standard
EV-062-03-13 containing 4,950 ppm PCP in methanol (Personal Communication, 1991).
The QAB audit sample had a target concentration of 20 ppm PCP and was prepared by
dilution from EMSL-Cincinnati QA reference standard C-090-02, Acid Extractables II, in
methylene chloride. It contained a mixture of phenols, including: phenol, 3-methylphenol,
4-methylphenol, 2-nitrophenol, 2,4-dimethylphenol, 2,4-dichlorophenol, 4-chloro-3-
methylphenol, 2,4,6-trichlorophenol, 2,4,5-trichlorophenol, nitrophenol, and a nominal
PCP concentration of 1,950 ppm (Personal Communication, 1991). Both reference
standards were originally developed to provide a 10 percent accuracy window for
interlaboratory studies, with this window mostly due to method and laboratory variability.
Analysis of the QAB mixture was intended to provide information on the selectivity of the
anti-PCP antibodies. Both audit types were also intended to provide precision and
accuracy performance data for all methods at all sites. The target PCP concentrations of
25 and 20 ppm for QAA and QAB samples, respectively, were chosen because this was
expected to be the mid-range of the bioreactor sample concentrations.
Quality control performance samples—These standards were designed to provide
immediate information to analysts regarding the kit and plate immunoassay performance.
For this demonstration, the QC performance sample concentration of 20 ppm PCP was
known to the analyst and was prepared from the stock solution used for the QAA audit
sample (see above). Besides its utility to the analyst regarding daily performance, the data
18
-------
derived from this sample provided bias and precision estimates for the kit and plate
immunoassays (Section 5).
• Field blank samples-These samples consisted of a reagent-grade laboratory water rinse of
the decontaminated composite sampling apparatus.
• Negative control samples--For the kit and plate immunoassay methods, the sampie dilution
buffer was used as a negative control (NC) sample. These samples are useful in detecting
false positive responses or contamination in the analysis.
• Cross-calibration standards-These standards were serial dilutions of 2,4-dinitrophenol-
giycine (2,4-DNP-glycine), a yellow-colored compound. They were used to cross-calibrate
the microwell-strip and laboratory-plate spectrometers. The 2.4-DNP-glycine has an
absorption maximum near 405 nm. the wavelength used to measure the enzymatic product
of the kit and plate sample wells.
OA/OC and Bioreactor Sample Analysis
A QC strip was analyzed at the beginning of every analysis day at each site by the kit immunoassay.
The 8-well strip contained two field blank analyses (undiluted and 1:10 dilution), one NC sample, one
QC performance sample, and four calibration standards. (NOTE: The four calibration standards
were analyzed on every strip, not just the daily QC strip). For 3 days of each 6-day analysis week,
both influent and effluent samples were analyzed; on the other 3 days, either an influent or an effluent
was usually analyzed. Each influent and effluent sample was analyzed at several dilutions, both for
range finding purposes and to determine the optimal dilution (i.e., in the mid-range of the calibration
curve) for final sample analyses. Then, depending on the day, a series of duplicate, split, or matrix
spike samples were analyzed on multiple strips. Once per week, the raw influent sample was analyzed
at three different dilutions. On 3 days during each 6-day period, performance audit samples were
analyzed at three different "dilutions on a pair of strips. The QAA audit was analyzed on 2 of the 3
days, and the QAB was analyzed on the third day. A similar analysis scheme was followed for the
plate immunoassay. The GC/MS analyses of the influent, effluent, and field blank samples followed
the analysis scheme for the bioreactor demonstration (SAIC, 1989).
The QAPjP also specified the collection of field duplicate samples as part of the bioreactor
demonstration. However, no field duplicate sample splits were provided by the sampling team.
Data Quality Objectives
To adequately assess the utility of the kit immunoassay as a field screening method, data quality
objectives (DQOs) were proposed in the QAPjP as guidelines for evaluating the quality and validity of
the data obtained in this study. However, even if the data did not satisfy the stated objectives, useful
information regarding the limitations and applicability of this particular method for field testing could
still be obtained. The DQOs established for the kit immunoassay in this demonstration are listed
below.
1. For field (influent and effluent) samples, the test result should not differ from the GC/MS
result by more than a factor of two (i.e., within 50 to 200 percent of the GC/MS result).
19
-------
2. The maximum coefficient of variation (CV) for the QC performance samples, which are
diluted and run by different operators on different days, should not exceed ±50 percent of
the nominal concentration of 20 ppm (i.e.. within 10 to 30 ppm PCP).
3. The QA audit sample results should be i50 percent of the target PCP value (i.e.. 12.5 to
37.5 ppm for QAA and 10 to 30 ppm for QAB samples).
If data from the kit immunoassay analyses of QA audit or QC performance samples fell outside the
expected ranges or if blank samples exhibited contamination, then the kit immunoassay analysis for a
particular strip was usually repeated (if time allowed). In addition, remaining volumes of each field
sample were stored at 4 ° C at each analysis site in case the data review process (Silverstein et al., 1991)
indicated the need for reanalysis.
Ensuring QA Desien Conformity
WBAS prepared detailed SOPs containing QC protocols and acceptance criteria for the kit and the
plate immunoassays (Van Emon and Gerlach. in preparation). Coupled with the predemonstration
data (Section 3), the SOPs formed the basis for the QC procedures developed for the demonstration.
Additional checks were employed to ensure data conformity. For example, all micropipettors used in
the study were checked for accuracy and precision by a standard gravimetric testing procedure prior to
the start of formal sample analysis activities. Also, field data forms were designed so that kit
immunoassay data would be collected in a consistent format from all analysis sites.
During the initial stages of the data collection activities, an on-site systems audit was conducted by
EMSL-LV QA representatives familiar with the immunoassay technology. Auditors inspected on-site
activities such as sample collection, handling, tracking, storage, and analyses. The results of the
systems audit are summarized in Section 5.
DATA MANAGEMENT
Field data forms were used to document all pertinent information related to each sample analyzed by
the kit immunoassay at all analysis sites (Silverstein et al., 1991). The forms documented analytical
and field condition information, facilitated data tracking, and standardized the method by which the
data were reported. The format allowed the information to be entered easily into a data base. Data
generated from the plate immunoassay by EMSL-LV and WBAS were compiled in tabular form.
SAIC provided copies of the data generated from their GC/MS analysis in tabular form, including
sample number, PCP concentration, and data qualifiers (flags). The flags are those typically used in
EPA Contract Laboratory Program (CLP) data reporting for Superfund site analytical measurements.
Upon receipt at EMSL-LV, all data were subjected to a QA review for consistency in reporting,
reasonableness, transcription, and other data reporting errors. Suspicious data were flagged when
appropriate to indicate observations made during the data verification and validation process.
Definitions of the data qualifier flags used in this study are provided in Appendix C Personnel at all
analysis locations were contacted to verify or correct suspicious values.
After data review, the data were entered and stored in a Statistical Analysis System (SAS) data set
divided into separate files (members), depending on the source of the data (on site, EMSL-LV, SAIC,
20
-------
WBAS) and on the analytical method (kit immunoassay, plate immunoassay, GC/MS). The final data
base from this demonstration has been fully documented in a data .base dictionary. It is available for
use in future assessment of the WBAS immunoassavs or other immunoassav demonstrations.
21
-------
SECTION 4
METHOD RESULTS AND COMPARISONS
The method comparisons and results discussed in this section pertain specifically to the analysis of the
bioreactor influent and effluent samples. Field splits of the bulk influent and effluent samples were
analyzed on site and off site by the kit immunoassay, off site by the plate immunoassay, and off site by
GC/MS (Figure 6). Comparisons of (1) the on-site kit immunoassay to the GC/MS, (2) the on-site kit
immunoassay to the plate immunoassay, and (3) the plate immunoassay to the GC/MS are of special
interest and are discussed in detail below. Other method-to-method relationships are also addressed
briefly. Results and comparisons related to various QA and QC samples are presented in Section 5.
Before the data analysis began, formal criteria for identifying a typical field analysis result had to be
developed for the kit immunoassay. The protocols for this demonstration required repeated analyses
of each sample at a given site. Multiple dilution levels were run to determine the optimal dilution
level and replicate analyses at a given dilution level were performed to determine precision. In
addition, the protocols dictated that the same samples were analyzed from duplicate independent
dilutions, used in a matrix spike run, or reanalyzed at the discretion of the operator. Though results
of all analyses were tabulated in the data base, only one or a pair (duplicate) of results could be used
for comparison to the plate immunoassay or GC/MS results if one were to simulate a typical field
analysis. Appendix D contains the algorithms that were used to represent the data values of each
sample. By following these algorithms, an unbiased selection of results from the kit immunoassay
analysis protocols was intended. Hence, the kit immunoassay results selected for a given sample were
not necessarily the best in terms of the comparison. The sole purpose of the algorithms was to reject
results if they would naturally be rejected in the field and accept the first possible results with no
associated discrepancies.
Bioreactor influent and effluent samples generally represent different PCP concentration ranges;
therefore, each sample type is discussed separately. Figures 7 and 8 present the results for all analysis
methods at each analysis site for influent and effluent samples, respectively. Figure 7 depicts PCP
concentrations for conditioned influent and raw influent samples. Conditioned influent samples were
collected six times per week; raw influent samples were collected once per week. Each of the 3 weeks
of sampling, represented as "A", "B", and "C", produced different ranges of concentrations for the
conditioned influent samples. Conditioned influent samples collected during the first week had the
lowest PCP concentration range. Sample concentrations and ranges increased during each subsequent
week. This stepwise increase in concentration is related to the increased flow rate of ground water
through the bioreactor, which was increased from 1 gpm during week A to 3 gpm during week B, and
5 gpm during week C This marked difference in conditioned influent concentrations between flow
rates could be caused by a "backmixing" action in the bioreactor at the lower flow rates (Stinson et aL,
1991, and Appendix A). Concentration levels lower than 40 to 50 ppm were not expected for
conditioned influent samples. However, when found, they proved to be a useful concentration range
22
-------
70 -
Q.
£.
z
o
cc
h-
III
o
z
o
o
0.
o
o.
111
50 -
40
30-
20 -
10 -
0 -
A
CONDITIONED INFLUENT SAMPLES
D
D A
X
O
o x A
x + ft
O + X A
0 O A *
00 + + * * Y
x x x * J
0 D X A A. n *
A * Y •
* t A
* a R A i
Y « X *
6 x x
RAW
INFLUENT
SAMPLES
0
*
A
X
X Y
D X
O
Y
I I I I I I I I I I I 1
A1 A2 A3 A4 A5 A6 B1 B2 B3 B4 85 B6 C1
SAMPLE NUMBER
i i i i i i i
C2 C3 C4 C5 C6 A5 B4 C2
• LEGEND•
D EMSL-LV. Kit A WBAS.Kit O On-Site.KIt
+ EMSL-LV. Plate X WBAS, Plate * SAIC. GC/MS
Y EMSL-LV. GC/MS
A1 = Flow period and day sample collected
Figure 7. Pentachlorophenol concentrations for all influent samples by all methods and analysis sites over time.
-------
E
Q.
Q.
4
fe
tr
UJ
o
z
O
o
Q.
o
Q.
<
UJ
2 1
1
0 -
o
6 +
0 t *
f + + +
0 D D v x
9 * x * fi
* * * *
A1 A2 A3 A4 A5 A6
EFFLUENT SAMPLES
D
>
x Y
$
o
x x
+ ° 5
8- x o |
X v A
8 t + * a * ,
, t •
, « ' « 9
B1 B2 B3 B4 B5 B6 C1 C2 C3 C4 C5 C6
SAMPLE NUMBER
• LEGEND•
D EMSL-LV.Kit
+ EMSL-LV, Plate
Y EMSL-LV, GC/MS
A WBAS.Kit O On-Sile.Kit
X WBAS. Plate * SAIC. GC/MS
A1 = Flow period and day sample collected
Figure 8. Pentachlorophenol concentrations for all effluent samples by all methods and analysis sites over time.
-------
rbr the immunoassay demonstration. In fact, because of the quick turnaround for the immunoassay
results. RREL investigators used the immunoassay data to confirm GC/MS results. This fact
underscores the utility of immunoassay for on-site analysis and the advantages of conducting joint
demonstrations of treatment and monitoring technoloeies.
Figure 8 shows that the effluent sample PCP concentration levels stayed within a relatively constant.
low concentration range during the demonstration, regardless of flow rate. The ranges were slightly
different for each analytical method but easily distinguishable from the higher concentration range of
the influent samples. All effluent samples analyzed by all sites and methods fell below 3 ppm PCP
except one outlier result (see following subsection. "On-site Kit Immunoassay Comparison to Off-site
Kit Immunoassay").
Table 2 summarizes the comparison data for the most important intra- and intermethod comparisons.
In Table 2. each comparison was made by arbitrarily assigning the data from one analysis site and
method as Data Set 1 and the data from the comparison site and method as Data Set 2. One of the
DQOs for this demonstration stated that the kit immunoassay results should not differ from the
GOMS results by more than a factor of two. As Table 2 shows, the factor-of-two DQO for the kit
immunoassay influent sample results was met for only 50, 75, and 88 percent of the samples at WBAS.
EMSL-LV. and on site, respectively. However, this factor-of-two comparison was used for influent
samples only. Instead of using the factor-of-two comparison criterion for the effluent samples, it was
more practical to compare the concentration ranges (lowest to highest, in ppm) for each method.
There was no meaningful correlation between the results in the usual statistical sense for effluent
sample comparisons. If one method or site reported a low PCP concentration, the other method or
site also generally reported a low value within its corresponding range for effluent samples. This
result is consistent with the semiquantitative nature of the kit immunoassay. Key analytical method
comparisons are discussed below.
KIT IMMUNOASSAY COMPARISONS
In this section, the results obtained using the kit immunoassay are compared to the GC/MS results
and the plate immunoassay results. In addition, on-site kit immunoassay results are compared to off-
site kit immunoassay results.
Kit Immunoassav Comparison to the GC/MS
Kit immunoassay results were generated on site at the MacGillis & Gibbs Superfund Site and off site
at the EMSL-LV and WBAS laboratories. The on-site kit immunoassay results compare favorably to
the GC/MS results (Figure 9 and Table 2). A Spearman rank correlation of 0.93 (n = 16, 95 percent
confidence interval of 0.81 to 0.98) was calculated for the influent data. This correlation indicates good
relative (rank order) agreement between the kit immunoassay and the GC/MS results. Fourteen of the
16 on-site influent samples (88 percent) were within the factor-of-two objectives. Of the two samples
that fell outside the limit, one is a sample for which both methods detected less than 1 ppm PCP.
This sample and all effluent samples at lower concentrations are presented in an inset in Figure 9.
The inset shows that the two methods provided results that were in the same general range (0.2 to 2.3
ppm PCP for the kit immunoassay versus 0.008 to 0.9 ppm PCP for the GC/MS) at these low
concentration levels. (Note: other figures in this section do not provide insets as in Figure 9, but
exhibit similar behavior for the lower concentration samples.)
25
-------
TABLE 2. ANALYSIS SITE AND METHOD COMPARISON SUMMARY
Analysis Site and Method
Data Set 1
GCMSd
GC/MSd
GC/MSJ
On-site. Kit
On-site. Kit
GOMSd
GOMSd
EMSL-LV, Plate
EMSL-LV, Plate
EMSL-LV. Plate
WBAS. Plate
WBAS. Plate
Data Set 2
On-site, Kit
EMSL-LV, Kit
WBAS, Kit
EMSL-LV, Kit
WBAS. Kit
EMSL-LV, Plate
WBAS, Plate
WBAS, Plate
EMSL-LV, Kit
On-site, Kit
WBAS. Kit
On-site, Kit
Influent1
comparison
n"
16
12
10
12
11
18
18
21
14
19
12
19
n<"2c
14
Q
g
6
17
12
16
10
15
8
12
<•:
(%)
88
75
50
67
54
94
67
76
71
79
67
63
Effluent range
comparison
Data Set 1
(ppm PCP)
0.008-0.91
0.008-0.91
0.008-0.91
0.40-2.65
0.40-2.65
0.008-0.91
0.008-0.91
0.32-1.86
0.31-1.86
0.31-1.86
0.40-2.74
0.40-2.74 '
Data Set 2
(ppra PCP)
0.20-2.27
0.24-1.78
0.23-1.45
0.32-5.25
0.22-1.45
0.31-1.82
0.40-2.11
0.40-174
0.24-5.25
0.20-2.65
0.22-1.45
0.20-165
n
11
9
•J
Q
8
16
16
18
10
12
8
12
'Includes conditioned influen: and raw influent samples.
bn = number of samples compared.
c<*2 = samples with pentachlorophenol concentrations within a factor of two (50 to 200 percent)
of each other.
dGC/MS analyses from SAIC laboratory only.
As Figure 9 illustrates, the kit immunoassay results are systematically biased high compared to the
GC/MS results. Table 3 summarizes information on high bias in the kit results relative to the GC/MS
results. Due to the difference in size of the bias relative to the actual ppm values, the average bias for
influent samples is given as a percentage of the values. The average bias from each analysis site for
the kit immunoassay relative to the GC/MS results ranges from 65 percent to 119 percent high.
Though the percent bias is larger for the effluent samples, the average actual concentration differences
are not large. Hence, the average bias for the effluent samples is given as an actual concentration
difference. The bias is also evident in the range of values obtained in the effluent samples. Though
the net effect of the bias is marginal in terms of utility, it does minimize the potential for false
negative responses from the kit immunoassay, even for low concentration samples. A similar bias is
not seen when the kit immunoassay is compared to the plate immunoassay (see discussion in the next
subsection).
A possible source for this bias is the cross-reactivity of the anti-PCP antibodies to other compounds in
the sample. Since tetrachlorophenol had the greatest cross-reactivity, both penta- and
tetrachlorophenol were quantitated by GC/MS in 8 field samples selected to span the PCP
concentration range predicted from immunoassay analysis. Based on the levels of these compounds as
determined by GC/MS and the crossrreactivities from Table 1, it was predicted that the kit
immunoassay test results would be increased by 3 to 11 percent over the concentration expected on the
basis of pentachlorophenol alone. This is much less than the 65 to 119 percent bias found between
GC/MS and kit immunoassay results, indicating that other important sources of bias must be present.
26
-------
60 H
to
-J
II
H
tn
i
z
o
30 40
SAIC.GC/MS
I
50
1
60
LEGEND
O Ellluent Sample
A Influent Sample
O Raw Inlluent Sample
Figure 9. Comparison of results from bioreactor sample analyzed for PCP on-sile by the kit immimoassay and at SAIC by GC/MS.
(Units are in ppm PCP and represent mean concentrations from each analysis site.)
-------
TABLE 3. KIT AND PLATE IMMUNOASSAY BIAS VERSUS GG'MS RESULTS
Analysis
method
Kit
iramuaoassav
Plate
iramunoassav
Analysis
sice
On-site
EMSL-LV
WBAS
EMSL-LV
WBAS
Average bias to GG'MS
Influent
n
15
10
9
16
16
% Greater
65
SI
119
40
17
Effluent
n
11
9
8
16
16
ppm u relief
0.62
0.43
030
0.55
0.55
For influent samples, the EMSL-LV kit immunoassay results compared less favorably to GOMS (75
percent within a factor of 2; Table 2) than the on-site kit immunoassay analyses did (88 percent within
a factor of 2). but trends for bias were similar (Table 3). The WBAS kit immunoassay analyses
compared poorly to the GC/MS results (only 50 percent within a factor of 2 for influent samples).
This result is associated with generally poor analysis results from the WBAS facility during the third
week of sample analysis during which half of their analyses were performed. The laboratory problem
is underscored by inconsistencies in bioreactor and QA and QC sample results generated by WBAS
during this period (Section 5). These analytical problems and QA/QC inconsistencies should be kept
in mind when reviewing kit immunoassay results reported by WBAS.
Of the 76 kit immunoassay analyses of the field samples, only 2 false negative results (a 2.6 percent
false negative rate) were observed when compared to the GC/MS results. These two results were for
the same influent sample collected at the beginning of the study, when high (50 ppm PCP) results
were expected. The study design specified a minimum dilution of 1:1,000 for these samples, but
subsequent analysis showed these samples to have less than 1 ppm PCP. Thus, the only false
negatives reported for the field samples are associated with overdilution. It is likely that a more
flexible analytical protocol would have produced a positive estimate for these samples.
Kit Immunoassav Comparison to the Plate Immunoassav
The kit immunoassay results were compared to the plate immunoassay in order to detect differences
between the two methods. Figure 10 presents a comparison plot of the on-site bioreactor sample
analyses for PCP by the kit immunoassay versus the EMSL-LV analyses using the plate immunoassay.
This figure and Table 2 show reasonable agreement between the two immunoassay techniques (e.g., 15
of 19 [79 percent] of the influent samples analyzed on site with the kit were within a factor of two of
the EMSL-LV plate results). For the kit immunoassay, 12 of 19 (63 percent) influent samples
analyzed on site were within a factor of two of the WBAS plate immunoassay results. Based on plots
of kit versus plate immunoassay results (Figure 10) and the ranges for effluent samples (Table 2), no
significant bias was found between kit and plate immunoassay results, though they were based on
different antibodies and had different cross-reactivity profiles. In addition, when the off-site kit
immunoassay results were compared to the plate immunoassay results (analyzed at the same
laboratory), similar variability and ranges were observed.
28
-------
LU
H
CO
I
z
o
60 -
50 -
40 -
30 J
20 -
10 -
0 -
i
10
I
20
i
30
i
40
i
50
i
60
LEGEND
O Effluent Sample
A Influent Sample
O Raw Influent Sample
EMSL-LV, PLATE
Figure 10. Comparison of results from bioreactor samples analyzed for PCP on site by the kit
< immunoassay and at EMSL-LV by the plate immunoassay. (Units are in ppm PCP
and represent mean concentrations from each analysis site.)
29
-------
On-site Kit Tmmunoassav Comnnrison to Off-Site Kit Tmmunoassav
The comparisons of the on-site and off-site kit immunoassay results (Table 2) exhibit the same kind of
variability as is shown in the kit to plate analysis comparisons. In most cases there is similar scatter in
the data with no tendency for a bias in the influent samples. Effluent samples analyzed by the kit
immunoassay show ranges that are generally higher than the GC/MS and similar to the plate
immunoassay. The exception to this is one unusually high effluent sample result from a single
microtiter well analyzed by the kit immunoassay at EMSL-LV, which appears to be an outlier (see
Figure 8. sample Cl). A misreponed dilution factor is suspected in this case, though later
investigation could not substantiate this possibility.
Overall Kit Tmmunoassav Comparison
Based on the results of the method comparisons, the kit immunoassay performed well in terms of
providing a semiquantitative estimate of PCP concentrations. However, variability in the results was
higher than expected based on the DQOs chosen for this demonstration. The site-to-site variability
suggests significant operator-dependent or procedural contributions to the analytical error.
Some bias existed between the kit immunoassay and the GC/MS. Because the immunoassay estimate
was high, a conservative estimate of PCP was made. Several factors may have contributed to this high
bias. It could have been caused by the inefficiency of the extraction used in preparing a sample for
GC/MS analysis. The bias could also have been related to the tendency of the immunoassay to
overestimate PCP by cross-reactivity and matrix interferences.
PLATE IMMUNOASSAY COMPARISONS
In this section, the results of the plate immunoassay are compared to the GC/MS results. In addition.
EMSL-LV analyses of the plate immunoassay are compared to WBAS analyses of the plate
immunoassay.
Plate Immunoassav Comparison to the GC/MS
The plate immunoassay results compared more favorably to the GC/MS than did the kit immunoassay,
with the exception of the results generated at WBAS for the influent sample plate immunoassay.
Figure 11 presents the comparison plot of bioreactor influent and effluent samples analyzed at EMSL-
LV with the plate immunoassay versus the GC/MS results. Although not as pronounced, the high
bias exhibited by the kit immunoassay with respect to the GC/MS was also evident in the influent
samples for the plate immunoassay (see Table 3). The plate immunoassay also shows a tendency for a
higher result for the effluent sample range (0.31 to 1.82 ppm) when compared to the GC/MS (0.008 to
0.91 ppm) (see Table 2). For the EMSL-LV plate immunoassay, 17 of 18 influent samples (94
percent) were within a factor of two of the GC/MS results. WBAS reported only 12 of 18 influent
samples within a factor of two of the GC/MS results. Though better than the results for the kit
immunoassay, it is unexpectedly poor for the developer of the technology. Nevertheless, the data
indicate that the plate immunoassay results compare more favorably to the GC/MS results. None of
78 sample analyses using the plate immunoassay generated a false negative result.
30
-------
50
10
i
20
i
30
i
40
50
SAIC, GC/MS
LEGEND
D Effluent Sample
A Influent Sample
O Raw Influent Sample
Figure 11. Comparison of results from bioreactor samples analyzed for PCP at EMSL-LV by the
, plate immunoassay and at SAIC by GC/MS . (Units are in ppm PCP and represent
mean concentrations from each analysis site.)
31
-------
Evidence that the plate immunoassay provides a closer estimate to the GC/MS than the kit
immunoassay is also reflected in the high biases for the kit and plate results on influent samples (see
Table 3). For an individual laboratory the average plate bias ranges from 17 percent to 40 percent too
high, whereas the average kit bias ranges from 65 percent to 119 percent too high. Note, however.
that for effluent samples, the bias is about the same for both immunoassay methods in terms of actual
concentration (0.3 to 0.6 ppm greater than the GC/MS results).
As with the kit immunoassay, a comparison of predicted plate immunoassay to GC/MS responses was
carried out for the 8 field samples based on cross-reactivity of tetrachlorophenol. The comparison
predicted biases of 7 to 26 percent high for the plate immunoassay compared to the GC/MS results.
The actual bias of the plate immunoassay relative to the GC/MS results was from 17 to 40 percent
high. Thus, while the cross-reactivity may account for a larger portion of the bias from GC/MS results
for the plate immunoassay than for the kit immunoassay, an additional source of bias is indicated.
Comparison of EMSL-LV to WBAS Plate Immunoassav Analyses
The plot of the EMSL-LV to WBAS plate immunoassay analyses of the bioreactor influent and
effluent samples (Figure 12) shows poorer agreement than one would expect from a quantitative
method. Sixteen of the 21 influent samples (76 percent) are within the factor-of-two objective. Even
though the majority of the influent samples were within a factor of two and the effluent samples were
all very close in range, there is more scatter in the data than expected between laboratories. However,
the range of effluent PCP analysis results corresponded well; all 18 samples from each analysis site fell
between 0.32 and 2.74 ppm PCP (see Table 2).
Although a pronounced bias was not exhibited in this comparison, a grouping phenomenon was
evident in this intramethod comparison, especially in the mid-range concentrations (i.e., samples
collected during the second week of the demonstration). The results from six samples analyzed at
WBAS during the second week are grouped in the 15 to 20 ppm range. The EMSL-LV
concentrations, on the other hand, range approximately from 10 to 40 ppm for these same samples.
The grouping effect of WBAS results at the 15 to 20 ppm range by date of sample collection can also
be observed at the lower and higher ranges. None of the analyses conducted at EMSL-LV showed
this phenomenon. Though a definite explanation for this grouping effect could not be identified, it
may be related to plate-to-plate or operator-dependent factors.
Overall Plate Immunoassav Comparison
In general, the plate immunoassay performed reasonably well in terms of its comparison to the
GC/MS results and to the kit immunoassay, but the interlaboratory comparison was poorer than
expected for a quantitative method. The plate immunoassay performed better in terms of accuracy
and precision than the kit immunoassay (Section 5). In all cases, the effluent sample concentrations
compared well between methods, and the higher concentration influent samples were generally within
a factor of two of each other. Of particular note is that when the EMSL-LV plate results were
compared to the GC/MS. 17 of the 18 influent samples were within the desired window, for the
WBAS plate immunoassay results, 12 of 18 influent samples were within that window. The plate
immunoassay, like the kit immunoassay, appeared to be biased high when compared to the GC/MS.
Loss in extraction, cross-reactivity factors, or non-specific matrix effects could have contributed to this
bias.
32
-------
UJ
§
a.
en
<
CO
60
50 -
40
30 -
20
10
i
0
I
10
I
20
i
30
40
i
50
i
60
EMSL-LV, PLATE
LEGEND
D Effluent Sample
A Influent Sample
O Raw Influent Sample
Figure 12. Comparison of results from bioreactor samples analyzed for PCP at EMSL-LV and
WBAS by the plate immunoassay. (Units are in ppm PCP and represent
mean concentrations from each analysis site.)
33
-------
SAIC GC/MS TO EMSL-LV GC/MS COMPARISON
Though the purpose of this demonstration was not to evaluate the GC/MS method, establishing
confidence in the GC/MS results was integral to the overall assessment of the utility of the
immunoassay methods.
One way to assess the performance of the SAIC GC/MS data was to have an independent set of
analyses of representative samples analyzed by another laboratory. To accomplish this, 1-L splits of
six bioreactor samples (three influent, two effluent, and one raw influent) and one field blank sample
were analyzed at EMSL-LV by the identical method (EPA Method 8270 analysis following Method
3510 extraction) (OSWER, 1986). These samples were selected at random from the complete set of
samples sent to SAIC for analysis. The EMSL-LV results represent a QC check on the GC/MS
results from SAIC for the entire study. Figure 13 shows the agreement between the two sets of values.
The field blank is not plotted, but both laboratories reported nondetectable results. The SAIC %'alues
are approximately 30 percent higher than the values obtained by EMSL-LV. For the range of 1 to 50
ppm, the relative standard deviation of the GC/MS is 30 percent (OSWER, 1986), which means a 95
percent confidence interval of approximately 60 percent. Hence, the two laboratories agree within the
accuracy limits of the GC/MS method. However, it should also be pointed out that the EMSL-LV
results were obtained after the prescribed holding time for the analysis of PCP in water. As a result.
effects such as analyte degradation or adsorption to sample container walls were also possibilities for
the consistently lower EMSL-LV results.
34
-------
40 -
30 -
CO
2
O
O
CO
LU
20 -
10 -
0 -
i
10
20
SAIC, GC/MS
30
r
40
> LEGEND
D Effluent Sample
A Influent Sample
O Raw Influent Sample
Figure 13. Comparison of results from bioreactor samples analyzed for PCP at EMSL-LV and SAIC
by GC/MS. (Units are in ppm PCP and represent mean concentrations from each
analysis site.)
35
-------
SECTIONS
QUALITY ASSURANCE AND QUALITY CONTROL RESULTS
Though the comparisons of immunoassay technologies to the GC/MS method in Section 4 represent
the most important facet of this SITE demonstration, the QA results are also important because they
provide estimates for parameters such as confidence limits, detection limits, and the quality of the data.
To accomplish the objectives of this SITE demonstration, the study design placed more emphasis (i.e..
more QA/QC) on the kit immunoassay technology. Similarly, this section stresses the interpretation
of the kit immunoassay results, but also includes information on the relative importance of plate
immunoassay and GC/MS results.
A number of different types of QA and QC samples were included in the kit and plate immunoassay
analyses. These included the QAA and QAB audit samples, the QC performance samples, duplicate
samples, method split samples, matrix spike samples, negative control samples, and field blank
samples (see Section 3 for their definitions). For comparison, only field blank samples and replicates
of the QAA and QAB audit samples were run by GC/MS (Method 8270) at the EMSL-LV and SAIC
laboratories. The QA and QC samples used in this study were intended to provide the following
information:
• Performance characteristics (accuracy, precision, bias) of both immunoassay methods with
standards of known concentration and matrix.
• Intra- and intermethod and interlaboratory comparisons.
• False negative and false positive rates.
• Data trends and correctable problems associated with the demonstration.
Acceptance criteria were established by the developer and used for screening raw data and monitoring
reagent integrity. In addition, field samples were assayed with and without matrix spike samples to
provide information about matrix effects and interferences. A set of serially diluted colorimetric
solutions were used to cross-calibrate the laboratory and field-portable spectrometers at all applicable
analysis sites.
QUALITY ASSURANCE AND QUALITY CONTROL SAMPLE RESULTS
The QA/QC sample types discussed in this section include: QAA and QAB audit samples, QC
performance samples, duplicate and method split samples, negative control (NC) samples, field blank
samples, and1 matrix spike samples.
36
-------
OA Audit Samples
Two different QA standard audit samples. QAA and QAB audit samples (see Section 3), were
analyzed at specified intervals at each site by each method. The QAA audit sample was prepared with
a nominal concentration of 25 ppm PCP. and the QAB audit sample was prepared with a nominal
concentration of 20 ppm PCP. QAA and QAB audit sample results are presented in Table 4 and
Table 5. respectively.
The QAA audit sample was semiblind: analysts knew it was an audit sample but did not know the
expected concentration of PCP (nominal concentration of 25 ppm). Table 4 summarizes the kit
immunoassay results for the QAA audit samples in terms of the lowest dilution (1:1.000) and the
mean of all dilutions (1:1.000. 1:2.000. and 1:4.000). Figure 14 provides a visual representation of this
information. It is apparent that using all dilutions provides better precision only for the on-site QAA
results. On-site results are both above and below the nominal concentration. Results from the
EMSL-LV and WBAS laboratories indicate a systematic bias (Figure 14) above the nominal
concentration. The last several QAA samples from WBAS reflect a trend that may be associated with
other problems of the WBAS laboratory during the final stages of this study. The 1:4,000 diluted
samples often gave the highest calculated concentrations, a problem that indicates a possible
systematic error associated with dilution level. This effect could also be associated with a systematic
bias at the lowest concentration level of the calibration curve. The QAA and QAB audit samples
were each assayed at three dilutions, and a significant systematic error was observed only for the QAA
samples.
One of the DQOs for this demonstration was that the kit immunoassay results should not differ from
the nominal results for the QA samples by more than ±50 percent. Considering the QAA (1:1,000)
and the QAB (mean) results for the kit immunoassay method. 25 of 34 (74 percent) were within the
±50 percent window around the nominal values. Thus, 26 percent of the QA sample values were not
estimated very well with respect to the DQO. Overall, a higher percentage (90 percent) of QAB
(mean) sample results were in the ±50 percent window than QAA (mean) sample results (63 percent).
For both the QAA (1:1.000) and QAB (mean) results, the on-site location had the best mean
accuracy, but the most variability (Figure 14). For the QAA (1:1,000) results, the WBAS laboratory
had excellent accuracy, while on average the EMSL-LV results were biased high by about 40 percent
(see Section 6 for a discussion of bias). For the QAB (mean) results, on-site results were biased
slightly low relative to the target concentration of 20 ppm, while both WBAS and EMSL-LV
produced similar results biased slightly high (Table 5). In summary, the QA results from the kit
immunoassay displayed more variability than was expected, and biases were generally high compared
to the nominal value.
Only two of 112 (1.8%) audit sample analyses (1 well/analysis; each dilution considered separately)
generated no detectable PCP. However, each of these two false negative results had used a dilution of
1:4,000, which would have an expected concentration of 5 ppb, near the lower limit of quantitation.
For the plate immunoassay, both the EMSL-LV and WBAS laboratories generated consistent results
(16 to 24 percent CV) near the target concentration for both QAA (25 ppm) and QAB (20 ppm)
samples (tables 4 and 5). The only exception was one obvious outlier QAA result (giving 70 ppm)
from one of three dilution levels run for one sample at EMSL-LV. The plate immunoassay QA
results suggest good behavior in terms of both accuracy and reproducibility for this method. Not one
of 54 plate immunoassay analyses of the QA samples produced a false negative result.
37
-------
TABLE 4. PERFORMANCE AUDIT SAMPLE RESULTS FOR OAA SAMPLES
Analysis
method
Kit
immunoassav
Plate
immunoassav
GGMS
Analysis site
On-site
EMSL-LV
WBAS
EMSL-LV
WBAS
EMSL-LV
SAIC
Dilution
factors
1:1,000
All dilutions'
1:1,000
All dilutions'
1:1,000
All dilutions'
All dilutions*
All dilutions
N/A*
N/A
na
11
11
5
5
8
8
16
16
4
3
Mean cone.
PCPb
(ppm)c
23
30
39
47
25
42
21
19
22
0.088h
CV1
(*)
39
27
24
30
29
79
16
16
19
13
Samples with mean cone outside
±50% of nominal cone
Below
115 ppm PCP
2
0
0
0
0
1
0
0
N/A
N/A
Above
37.5 ppm PCP
1
3
1
2
4
3
0
0
N/A
N/A
a n = number of QAA samples analyzed.
b PCP = pentachlorophenol.
c Nominal PCP concentration = 25 ppm.
d CV = coefficient of variation.
e Average of 1:1,000, 1:2,000, and 1:4,000 dilution results.
One outlier (at 70 ppm) not included.
8 N/A = criterion not applicable to analysis method.
h Incorrect dilution scheme used in preparing the samples for analysis yielding a 0.125 ppm target concentration.
Only a few QAA and QAB samples were analyzed by the GC/MS method (tables 4 and 5). QAA and
QAB results from EMSL-LV were between 14 and 25 percent low, respectively. The discrepancy
from the nominal concentrations may be due to error introduced by dilution or preparation of the
standard solutions or to random error in the GC/MS results. Due to miscommunication between
EMSL-LV and SAIC project management, personnel at SAIC prepared QA audit samples by diluting
them over 100 times more than desired (with respect to analytical protocols) prior to analysis. The
results from SAIC, therefore, are not directly comparable to those from EMSL-LV. For the dilutions
actually used, the expected concentrations are 0.125 ppm for the QAA and 0.040 ppm for the QAB
audit samples. Thus, the average values reported are low for each type of QA audit sample.
OC Performance Samples
The QC performance samples were run once per analysis day by kit immunoassay at all sites (Section
3). Table 6 summarizes the kit immunoassay results at all sites in terms of the mean and CV for
replicate analyses. The QC performance sample results are summarized in terms of the number of
test results that fell within ±50 percent of the 20 ppm PCP target value. On-site QC results were
38
-------
TABLE 5. PERFORMANCE AUDIT SAMPLE RESULTS FOR OAB SAMPLES
Analysis
method
Kit
iramunoassay
Plate
iraraunoassay
GOMS
Analysis
•sice
On-site
EMSL-LV
WBAS
EMSL-LV
WBAS
EMSL-LV
SAIC
Qa
3
3
4
8
8
4
2
Mean cone.
PCP6 (ppmjc
16
26
28
24
21
15
0.016f
CV*(%)
53
9
14
24
16
10
47
Samples with mean cone. Outside
the ±50% of nominal cone.
Below 10 ppm
o
0
0
0
0
N/Ae
N/A
Above 30 ppra
0
0
1
0
0
N/A
N/A
" n = number of QAB samples analyzed.
b PCP = pentachlorophenol.
c Nominal PCP concentration = 20 ppra.
d CV = coefficient of variation.
e N/A = criterion not applicable to analysis method.
f Incorrect dilution scheme used in preparing the samples for analysis yielding a 0.040 ppm target concentration.
within this window 67 percent of the time. This is relatively poor performance and may be due to lack
of training or poor field conditions. EMSL-LV had a much better rate with 89 percent within these
bounds, while WBAS had a much poorer rate of only 42 percent. It should be noted, however, that all
the WBAS values that were unacceptable were consecutive runs at the end of the study, a period when
problems were affecting laboratory performance at WBAS.
EMSL-LV was the only site to meet the CV DQO of no more than 50 percent, with a CV of 43
percent. On-site QC values tended to be biased low (14 ppm, compared to the target concentration of
20) and had a CV of 61 percent. The WBAS QC values were biased high on average (23 ppm) and
had a CV of 57 percent
There was a relatively high rate of false negative responses for the QC performance samples (3 of 49,
or 6 percent). This high rate can be partly attributed to two false negatives at the WBAS laboratory
near the end of the study, when various other samples (such as audits and NCs) were also out of
control. There were no false negative QC results from the on-site kit analysis.
During the early part of the demonstration, numerous plate immunoassay false positives were
observed at EMSL-LV. The EMSL-LV plate immunoassay results may have been affected by (1)
stability and liter of the antibody or coating antigen supplied by WBAS, (2) lyophilization of
immunologic reagents, (3) contamination, or (4) some other unexplained phenomenon. This
problem, which also resulted in offset standard curves and higher than expected results for the QC
performance samples, was corrected by changing the antibody dilution factor (see the subsection
below, "QA Problems and Resolutions.") For the plate immunoassay method, EMSL-LV results from
39
-------
Q.
z
O
UJ
O
o
o
a.
o
a.
150
125
100 •
75 •
50
25
0
D
A
O
ON-SITE KIT
O
A
0
O A
A a o
A A A
61 n ° ° D n
D a ^ a A a v
0 A
D
9 Q
I I i I I I I I I I I
EMSL-LV KIT
O
O O
A
A
a
E D
a D
A
§
i ii i i
WBAS KIT 0
A
0
A
0
O
o a A a
" Q
D u
D
O
A §
I 1 I I I 1 I I
»- CM CM »-
< < < ffl
m m
»- »- CM CM
O O O O
CM T- «- eg
< O O O
CM
O
CM •- CM
»-»-CMCM»-»~CMCM
<<<
-------
TABLE 6. KIT IMMUNOASSAY QC PERFORMANCE SAMPLE RESULTS
Anaivsis
site
On site
EMSL-LV
WBAS
All
n1
19 (18)e
18 (17)e
12 ( 9)e
49 C44)
Mean
PCP° cone.
(ppm):
14
20
23
19
CV*(%)
61
43
57
--
Window (10
n within
12
16
5
33
to 30 ppm)
n outside
7
2
7f
16
n false
negative
0
1
2
3
a n = number of quality control performance samples analyzed.
PCP = pemachlorophenol.
c Nominal PCP concentration = 20 ppm.
d CV = coefficient of variation.
c (n) Indicates number of sample results used in mean and CV calculation (deleted samples were either false
negative or had diluted concentrations above the linear range of calibration standards).
All six samples analyzed in the second half of the demonstration fell outside the window.
17 QC sample analyses generated an average 34 ppm (CV = 62 percent), and WBAS results from 12
QC sample analyses generated an average 13 ppm (CV = 8 percent). Results from WBAS were
consistently low, but EMSL-LV had four results greater than 50 ppm while the rest were only slightly
higher than the target value. Only three of the QC performance samples were run after the problem
with the antibody was solved. The results for these analyses were within the lower portion of the
performance window, with a mean PCP concentration of 14 and a standard deviation of 2 ppm. None
of the 29 QC sample analyses produced a false negative result. Field samples analyzed during the
period of unacceptable QC performance were reanalyzed after the reagent problem was corrected.
The results from the sample reanalysis were used in the method comparison assessments (Section 4).
Duplicate and Method Split Samples
A general overview of the within-strip well-to-well variability associated with kit immunoassay analysis
is shown in figures 15 and 16. Duplicate and method split samples (see Section 3 for definitions) were
used in this comparison. AJ1 paired results from on site, EMSL-LV, and WBAS for both influent and
effluent samples were used in these figures.
Figure 15 is a plot of the mean response for each pair versus the absolute difference in ppb units.
This plot uses the "original" estimates from the standard curve for the diluted samples. Most of the
differences are within 6 ppb, and this suggests that future work might require duplicate analyses (i.e.,
the same sample on the same strip) to be at least this close in agreement. Also, the averages of most
pairs are below 25 ppb. The error appears to grow as a function of average response, which would
tend to counteract dilution errors.
Figure 16 is a plot of the mean response of duplicate or split sample pairs versus their difference in
ppm units. This plot represents the error after scaling up the sample concentrations by the
appropriate dilution factor. The difference between most pairs is less than 50 percent of their average.
This plot may be useful in establishing acceptance criteria (e.g., acceptable variability for the kit
iramunoassay method).
41
-------
60
50
Ill 40
O
LU
DC
HI
LL
U.
Q
30
20
10
0
— LEGEND —
O EMSL-LV
A On-Slte
O WBAS
O
O
O
A O
10 20 30 40 50
ESTIMATED SAMPLE MEAN (ppb PCP)
i
60
Figure 15. Mean versus difference in results from pairs of analyses for kit immunoassuy duplicate and split field samples diluted into
calibration range (ppb PCP).
-------
70
60 -
50 -
HI
O
UJ 40 H
CC
LJJ
Q 30
Q.
CX
20 -
10 -
0 -
— LEGEND —i
D EMSL-LV
A On-Slte
O WBAS
O O
O A
A A
A
A A
a
D
10
I
40
I
50
20 30
ESTIMATED SAMPLE MEAN (ppm PCP)
60
70
Figure 16. Mean versus.difference in results from pairs of analyses for kit immimoassay duplicate and split field samples will) dilution factors
applied (ppm PCP).
-------
Negative Control Samples
Negative control samples were analyzed in one well in the initial strip of each daily sample analysis
series. This was one of a number of performance checks used to monitor for problems such as
contamination of reagents and equipment. The overall rate of false positives for the kit immunoassay
from all three laboratories was 19 percent (Table 7). The on-site rate of 22 percent was greater than
the EMSL-LV rate of 13 percent. WBAS had a false positive rate of 23 percent, unexpectedly high
considering WBAS was the developer. Further investigation showed that all false positives for WBAS
are for samples analyzed on August 28 and 31. This time frame coincides with that of other data from
WBAS suggesting poor performance characteristics during this period. The higher false positive rate
seen in the on-site results is probably due to higher levels of contamination at the field
laboratory and perhaps to the fact that field analysts were less familiar with the technical aspects of
the immunoassay.
The overall false positive rate in the NC samples appears relatively high, but the average level is about
6 ppb. This level of contamination would not be a very significant problem if the detection limits
were raised slightly. A protocol specifying a minimum value of 7 ppb would have substantially
decreased the relative influence of low level contamination.
For the plate immunoassay method, none of the 9 EMSL-LV or 12 WBAS NC sample results
represented a false positive. The difference between the kit and plate immunoassay results of NC false
positive rates may be due to the higher detection limit (30 ppb) for the plate method.
Field Blank Samples
Daily field blank samples consisted of reagent-grade laboratory water (no detectable PCP) used as a
rinse of the sample collector after it had been cleaned (SAIC, 1989). Field blanks were run once per
daily sample collection series, with each immunoassay analysis performed on an undiluted sample and
after a ten-fold dilution. The GC/MS field blank runs were only carried out on the undiluted sample.
The percent of positive responses for field blank samples was 13 percent for the kit immunoassay and
9 percent for the plate immunoassay (Table 8). The mean response was 5.0 ppb for the raw kit
immunoassay analysis results and 34.5 ppb for the raw plate immunoassay analysis results (i.e., without
the 10-fold dilution factors included). Positive field blank results were felt to represent false positive
results rather than estimates of contamination of the field blanks from the sample collection
decontamination rinse procedures. The pattern of false positives for the undiluted and 10-fold dilution
field blank results is not statistically different (at the 90% confidence level) from what one might
expect from random false positive generation. In addition, there was an absence of common false
positive results for field blank samples split between sites. Hence, it is concluded that false positive
field blanks are most likely due to contamination or procedural errors at the analytical level and not to
contamination or carryover between samples from the sample collector. It should be noted here that
the low kit immunoassay standard was 3.0 ppb, the low plate immunoassay standard was 20 ppb, and
the low GC/MS standard was 20 ppb. Also, the plate immunoassay method has a lower limit of
detection of about 30 ppb, and GC/MS Method 8270 has a lower limit of detection of 50 ppb. Trace
levels of PCP would thus be more easily measured by the kit immunoassay than either the plate
immunoassavor GC/MS.
44
-------
TABLE 7. KIT IMMUNOASSAY NEGATIVE CONTROL SAMPLE RESULTS
Analysis
site
On she
EMSL-LV
WBAS
All
a1
37
30
31
98
n with no PCPb
detected
29
26
24
79
n with PCP
detected'
8 (22%)
4 (13%)
T* (23%)
19 (19%)
Mean cone of n where PCP
detected (ppb)
4.6
6.2
6.9
5.8
a n = oumber of negative control samples analyzed.
b PCP = peniachlorophenol.
c Lowest calibration standard = 3 ppb, but detectable samples extrapolated down to 2 ppb PCP.
d Six of the 15 samples analyzed in the second half of the demonstration had PCP detected.
TABLE 8. FIELD BLANK ANALYSES RESULTS
Analysis
method
Kit
immunoassay
Plate
immunoassay
GC/MS
Analysis
site
Oa-site
EMSL-LV
...»«...«. .««»»....
WBAS
All
EMSL-LV
WBAS
All
SAIC
....«.».». „*,»«...
EMSL-LV
All
Dilution
factor
Undiluted
1:10
Undiluted
1:10
«—•—•«"—••
Undiluted
1:10
—
Undiluted
1:10
Undiluted
1:10
_
N/Af
.— . .«. ».«. ««. «~
N/A
N/A
na
19
20
18
18
»— — —~
22
12
109
18
18
18
18
72
18
•*™11"1"111
1
19
n with
PCPb not
detected
17
18
15
17
.«».« .™*.™~ .
18
11
96
18
15
16
18
67
17
«T T.II1M11 lltlr 111!
1
18
n with
PCP
present
2
2
3d
1
—— «-——«—••
4
1
13
0
3
2
0
5
0
"*•""• ••«— • ••••—
0
0
% n with
PCP
present
—
~~
..»*« ........ .......
12.7
—
9J
_
0
Mean cone n
with PCP
present' (ppb)
5.6
3.5
.„.„. «— «- -..™.
5.6
5.0
35.4
33.6
34.5
N/A
Detection
limit
(ppb)
3e
30°
50«
* n = numbers of field blank samples analyzed.
b PCP = pentachlorophenol.
c 1:10 dilution factor not applied in estimating the mean.
d One sample with a concentration above linear range of calibration.
e Stated detection limit (from WBAS).
' N/A = not applicable to analysis method.
8 From EPA Method 8270 (OSWER, 1986).
45
-------
Though there was a slightly higher rate of positive field blank responses for the kit as opposed to the
plate method (13 percent versus 9 percent), analysis location, method, or dilution level did not
significantly affect the percentage of field blanks giving detectable PCP. One exception was a lack of
field blank samples giving positive responses for the GC/MS method; however, the GC/MS method is
less sensitive. It is possible that some field blank samples were contaminated, despite the lack of
statistical evidence. The collection and analysis of field blank samples entails many steps, allowing for
the higher possibility of contamination for field blanks than for NC samples. However, analysis
results did not corroborate this possibility. A comparison of field blank false positive rates to NC
false positive rates shows that the on-site NC rate (8 of 37) is about twice as high as the field blank
rate (4 of 39). For EMSL-LV, the false positive rates for the NC and field blank samples were about
the same. For WBAS. the false positive rate was higher for NCs than for field blank samples.
Matrix Spike Samples
Matrix spike samples were analyzed in duplicate (i.e., two pairs per daily run) several times per week
using selected influent and effluent samples. The rationale for running these samples was to obtain
information about matrix effects and interferences in influent or effluent samples. Pre-spike dilution
factors were selected to generate PCP concentrations in the lower pan of the linear dynamic range for
the kit immunoassay. Matrix spikes of 15 ppb were used at all sites, except on site for the first week
when a 30 ppb spike was inadvertently used.
In general, matrix spike recoveries were highly variable for the kit immunoassay analyses. Table 9
shows that for on site and for WBAS, only 50 percent of the matrix spike recoveries were within a ±50
percent range around the expected spike recovery. EMSL-LV was somewhat better, with 75 percent
within this range. The on-site recoveries were generally low but were also highly erratic. Similarly, 24
of 26 WBAS results were less than the expected result. EMSL-LV generated a slightly high spike
recovery average (118 percent), but these results were also erratic.
In general, there was better performance from the matrix spiking procedure using the plate method
than using the kit method (Table 10). EMSL-LV had a much wider range of percent recoveries than
did WBAS, indicating that WBAS had better precision when performing the analysis of these samples.
The overall percent recovery, regardless of laboratory, was slightly below 100 percent, but was still
within the 75 to 125 percent window.
ASSESSMENT OF DATA QUALITY
The kit immunoassay has the potential to be used to rapidly screen samples in the field and to
determine which samples should be sent to a laboratory for quantitative analysis. Consequently, the
kit immunoassay must be capable of producing data which is sufficiently accurate and precise to assist
the field analyst in making judgments about the source, distribution, and approximate concentration of
the target analyte, PCP.
Data Quality in Terms of Five Data Quality Elements
The following sections contain a summary of data pertaining to the five basic data quality elements:
accuracy, precision, representativeness, completeness, and comparability. Conclusions drawn from the
data quality elements can be found in Section 6.
46
-------
TABLE 9. KIT IMMUNOASSAY MATRIX SPIKE RESULTS
Analysis sice
On-site
EMSL-LV
WBAS
All
n1
42
12
26
80
n with recovery6 in
50 to 150% window
21
9
13
43
a in
Window
(%)
50
75
50
54
Median
recovery
(%)
67
118
72
N/AC
Range or'
recoveriesb
(%)
-166 to +256
+ 57 to 1-313
- 146 to + 186
-166 to +313
numbers of matrix spike % recoveries calculated/
N/A = not applicable.
c % Recovery calculated as:
100
[sample * spike] - [sample]
[spike]
TABLE 10. PLATE IMMUNOASSAY MATRIX SPIKE RESULTS
Analysis site
EMSL-LV
WBAS
All
na
20
15
35
a with recovery in
75 to 125% Window
10
7
17
n in window
(%)
50
47
49
w
Mean recovery0
(%)
87
79
84
Range of recoveries
(%)
41 to 169
68 to 114
41 to 169
1 n = number of spikes analyzed.
' % recovery calculated as:
[sample + spike] - [sample]
. [spike]
Accuracy--
The accuracy of the plate and kit immunoassays was assessed in a number of different ways.
Immunoassay results for influents, effluents, field blanks, and QA audit samples were compared
directly to those obtained by GC/MS. Accuracy was also assessed by comparison of the immunoassay
results with those expected for the QA and QC samples (QA audit samples, QC performance
standards, NC samples). Sections 4 and 5 provide discussions of the immunoassay results on field and
QA samples, respectively. Information on the preparation and composition of QA and QC samples
is discussed in Section 3. Summary statements pertaining to accuracy data for the immunoassay
methods are listed below.
47
-------
For the kit immunoassay method:
• Seventy-four percent of all the QA audit sample analyses (n = 34) were within ±50 percent
of the expected value.
• The false negative rate for replicate analysis of the nominal 20 ppm QC performance
standards from all sites was 6 percent (n = 49), the rate for the influent and effluent field
samples was 2.6 percent (n = 76), and the rate from the QA audit samples was 1.8 percent
(n = 112). It should be noted that several instances of false negatives are associated with
overdilution of the sample.
• The mean false positive rate for NC samples was 19 percent and ranged from 13 percent at
one site to 23 percent at another.
'• For influent samples, 74 percent of the results were within a factor of two (50 to 200
percent) of the respective GC/MS results. Percentages at individual sites ranged from 50
percent to 88 percent.
• The influent sample results were biased high by an average of 84 percent relative to the
GC/MS results.
• The bias of the mean results of the QA and QC samples was site dependent. Bias ranged
from near 0 to 40 percent.
• Measurable concentrations of tetrachlorophenol were found in selected bioreactor samples.
The tetrachlorophenol, a cross-reactant in the kit immunoassay, could have contributed 5 to
10 percent to the high bias relative to the GC/MS analyses (Section 6).
• For all analysis sites combined, lowest dilution results were closer to the expected
concentration than higher dilution results for 19 of 24 of the QAA samples and for 7 of 10
of the QAB samples.
• For the QC performance samples, 69 percent of the results were within a factor of two (50
to 200 percent) of the expected results. The percentages ranged from 42 to 89 percent,
depending on analysis site.
Effluent samples with GC/MS concentrations in the range from 0.008 to 0.91 ppm PCP gave
immunoassay results in the range from 0.20 to 2.27 ppm. Although the factor-of-two criterion was
not always met, the data substantiate that the kit immunoassay can provide useful information about
the approximate levels of PCP in environmental samples.
For the plate immunoassay method:
• Eighty-one percent of the influent sample results (n = 36) were within a factor of two (50
to 200 percent) of the GC/MS results.
48
-------
• For both laboratories combined, plate immunoassay results for induent samples were biased
high by 28 percent relative to the GGTvlS results.
• The false negative rate was 0 percent for the bioreactor samples (n = 66), the QA samples
(n = 48), and the QC performance samples (n = 29).
• The false positive rate for the NC samples was 0 percent (n = 21).
• The mean PCP result was 20.1 (n = 32), for the QAA audit samples and 22.5 (n = 16) for
the QAB audit samples. These results are close to the expected concentrations for these
samples (25 ppm for the QAA and 20 ppm for the QAB). The results for both sites were
within ±50 percent of the expected values.
The results indicate that the plate immunoassay is quantitatively more accurate than the kit
immunoassay. It is more comparable to the GC/MS results for field samples and to the expected
results for the QA samples. Both the kit and plate immunoassays had low false negative rates, an
essential criterion for screening methods.
For the GC/MS method, one measure of accuracy was available from the Biotrol SITE demonstration
(U.S. EPA. n.d.). Eight samples were analyzed in duplicate with matrix spikes of 200 ppb PCP. Six
of the eight samples had no detectable pre-spike PCP levels, and two samples had pre-spike PCP
concentrations of approximately 200 ppb. The mean spike recovery was 96 percent (median recovery
of 83 percent). There was considerable variability in determining percent recoveries on an individual
sample basis. The range of recoveries was from 65 to 204 percent. The estimate of the standard
deviation calculated from this data set was 58 percent. Based on the above, the average GC/MS
recovery (based on the mean) is expected to be 0 to 10 percent low, and the typical GC/MS recovery
(based on the median) is expected to be 10 to 25 percent low.
Precision--
The precision of both immunoassay methods was assessed by analyzing results obtained from
replicates (duplicates and splits) of field samples, QA audit samples, and QC performance standards.
Summary statements pertaining to precision data for the immunoassay methods are listed below. See
sections 4 and 5 for a comprehensive discussion of the immunoassay results on field. QA, and QC
samples. Information on the preparation and composition of QA/QC samples is discussed in Section
3.
For the kit immunoassay method:
• The CV for the replicates of the QC performance samples exceeded 50 percent for the
combined results from all sites.
• Five out of six of the CV values for the QAA (1:1000) and QAB (mean) audit sample
results were below 40 percent. The other value was 53 percent.
• The plot of mean response versus difference for field samples (Figure 15) indicated that
samples analyzed in duplicate on one microtiter strip usually differed by 6 ppb or less.
49
-------
• A plot (not shown) of mean response (in ppm) versus relative difference for field samples
indicated that duplicates with mean concentrations above 30 ppm generally differed by 15
percent or less, while those with mean concentrations of 5 ppm or below differed by 50
percent or less.
The variability in the kit immunoassay results was higher than expected. The DQO requiring that the
CV not exceed 50 percent for replicates of the QC performance samples was only met at one of the
three laboratories (EMSL-LV).
The mean versus difference plots for field samples indicated that the difference between most pairs is
less than 50 percent of their average and that the error appears to grow as a function of average
response. The data indicate that dilution error and operator or protocol-dependent error are
important factors in interlaboratory variance.
For the plate immunoassay method:
• The CV for replicates of the QA audit samples at all sites ranged from 16 to 24 percent.
• The CV for replicates of the QC performance standards was 62 percent for EMSL-LV and
8 percent for WBAS.
Overall, the plate immunoassay precision was better than the kit, which was not unexpected. Most of
the CVs were in the expected range of 10 to 16 percent. The abnormally high CV for the QC
performance results at EMSL-LV resulted from plate reagent performance and stability problems that
were encountered during the early part of the demonstration.
For the GC/MS method, two sets of data, were available from the Biotrol SITE demonstration (U.S.
EPA, n.d.) from which to analyze the precision of the GC/MS results. One set of data consisted of
eight field duplicate pairs. The range of CVs from these pairs was 0 to 101 percent, and the pooled
CV was 33 percent
The second set of data used to estimate precision consisted of effluent sample splits, where one split
was analyzed after filtration to test whether PCP was retained on any filterable solids, such as dead
cells exiting the bioreactor. Of 58 samples split for analysis, 32 splits yielded filtered results less than
the unfiltered results, 22 splits yielded unfiltered results less than the filtered results, and 4 splits were
identical. The ratio of higher to lower results of sample splits for testing the effects of filtration is not
statistically different from 1 at the 95 percent confidence level. Therefore, the differences between
pairs of results were used to generate a second estimate of precision. Since these pairs are not true
duplicates, the estimate of variability derived from them will be conservative. That is, the variability
estimated from them may be biased high.
The range of CV estimates for the before-and-after filtration pairs was from 0 to 115 percent.
However, 80 percent of the CVs were below 27 percent, and 90 percent of the CVs were below 43
percent. The pooled CV was 21 percent, and the median CV was 13 percent Based on both sets of
precision analyses from both sets of paired samples, our overall estimate of the standard deviation for
the GC/MS results is 20 to 30 percent
50
-------
Representativeness--
The data obtained for the kit and plate immunoassay methods in this demonstration are representative
in two respects. Because of the extensive set of QA/QC and intermethod comparison samples
analyzed in the demonstration, the results are representative of the capabilities and variability of the
immunoassay methods. Also, the influent and effluent samples analyzed by the immunoassays are
representative of the Biotrol bioreactor process. However, because the demonstration involved only a
single site and the sample selection was limited, the data are not necessarily representative of what
would be obtained at other sites or with other sample matrices. Further evaluation of the '
immunoassay methods may need to incorporate analysis of a wider variety of aqueous samples and
matrices.
CompletenesS"
The completeness objective of analyzing over 90 percent of the planned number of field samples by
the immunoassay methods was met (see Table 11). Another aspect of assessing completeness involved
determining the number of kit immunoassay runs that gave usable results compared to the total
number of runs. For the kit immunoassay, a run was defined as the results from a single 8-well
microtiter strip. Table 12 gives a summary of the total number of strips run versus the number of
strips yielding usable data. Approximately 7 percent of the strips run gave unusable data because no
straight line could be constructed using at least 3 of the 4 standards. Thus, the 90 percent
completeness objective was met for this category. For the strips with one inconsistent standard, useful
data were obtained by drawing a straight line through 3 of the 4 points. Overall, 10 percent of the 256
strips providing quantifiable data fell in this category (see Table 12).
The only location that had a substantial number of strips (16) with poor calibration curves was the
field site. This was probably due (in part) to a substrate contamination problem associated with the
pipetting procedure. The problem was alleviated during the second week of the demonstration by a
procedural modification. Out of a total of 16 unusable runs, 11 were from the first analysis week, 4
from the second week, and 1 from the third week. By percentage, this corresponds to 21 percent, 11
percent, and 3 percent of the total number of strips run in those respective weeks. These percentages
substantiate the improvement in technique and performance over time.
Fewer QA audit samples were analyzed by GC/MS than the number stated in the QA plan. Due to
logistical and time constraints, only 3 of 6 QAA samples were analyzed at SAIC by GC/MS, and only
4 of 6 QAA samples were analyzed at EMSL-LV by GC/MS.
Comparability--
Data pertaining to intermethod and interlaboratory comparisons are presented and discussed in
sections 4 and 5. Summary statements about the immunoassay comparability data are given below.
• For influent samples, approximately 74 percent of the kit immunoassay results were within
a factor of two (50 to 200 percent) of the GC/MS results. The number of results within a
factor of two ranged from 50 percent at one site to 88 percent at another. The average high
bias was 84 percent
51
-------
TABLE 11. TYPES AND NUMBERS OF FIELD SAMPLES ANALYZED BY ANALYSIS SITE
Sample type
Influent
Effluent
Raw
influent
Field
blank
AJ1
Target
18
12
3"
18
51
Kit immunoassay
On-Site
17
12
3
18
50
EMSL-LV
12
10
2
14
38C
WBAS
11
8
1
12
32C
Plate immunoassay
EMSL-LV
18
18
3
18
57
WBAS
18
18
3
18
57
G
SAIC
16
16
•\
18
52
OMS
EMSL-LV1
3
2
t.
1
8
a The target values do not apply to this column.
b Although three samples were collected, only two were common to immunoassay and GC/MS analysis.
c The target value for this total is 34 because one week of analyses were not performed.
TABLE 12. SUMMARY OF QUANTIFIABLE DATA FOR THE KIT IMMUNOASSAY
Analysis site
On-site
EMSL-LV
WBAS
All
All strips used
n
127
78
69
274
Percent with no acceptable
calibration curve
16 (13%)
1 (1%)
1 (1%)
18 (7%)
Strips with quantifiable data
0
111
77
68
256
Percent with inconsistent
standard
11 (10%)
11 (14%)
4 (6%)
25 (10%)
• For influent samples, 81 percent of the plate immunoassay results were within a factor of two
(50 to 200 percent) of the GC/MS results. The number of results within a factor of two
ranged from 67 percent at one site to 94 percent at another. The average high bias was 28
percent.
• For influent samples, the percentage of kit immunoassay results within a factor of two (50 to
200 percent) of the laboratory plate results ranged from 63 percent to 79 percent.
• For the influent samples, both the kit and the plate immunoassay results were biased high
relative to the GC/MS. On the.average, kit results were biased 65 to 119 percent too high and
plate results were biased 17 to 40 percent too high.
Although the percentages of kit and plate influent sample results within a factor of two of the GC/MS
results were similar (i.e., 74 percent and 81 percent, respectively), the plate immunoassay results were
more comparable, as indicated by the relative percentages of high bias. The broad percentage ranges
for the factor-of-two comparisons (kit versus plate, kit versus GC/MS, and plate versus GC/MS)
indicate higher than expected, location-dependent variability.
52
-------
OA Problems and Resolutions
One of the problems encountered in analyzing the data was that the validity of the raw data from
some of the strip and plate runs was questionable because of outliers, nonlinearity of the standard
curve, or other problems. In some cases, different dilutions of field samples gave quite different
analyte concentrations. The primary difficulty was in selecting which immunoassay data were suspect
and which would be used for intermethod and interlaboratory comparisons. The QC acceptance
criteria for the kit and plate immunoassay methods served as a basis for eliminating inconsistent
results. Appendix D contains a discussion of the approaches taken in selecting plate and kit
immunoassay data and an explanation of the rationale used in selecting data for intermethod and
interlaboratory comparisons.
Another QA problem involved the GC/MS analysis of the QAA and QAB audit samples analyzed at
SAIC. These ampulated samples were provided, semi-blind (with the approximate concentration
ranges given), to the SAIC GC/MS laboratory in San Diego. EMSL-LV did not provide written
procedures regarding sample preparation, and the concentration range given by EMSL-LV to SAIC
was too broad (1 to 100 ppm PCP). As a result, the QA audit samples sent to SAIC were not diluted
properly. Although unfortunate, this problem did not seriously affect project QA because: (1) the
EMSL-LV GC/MS laboratory analyzed the QA samples at the proper dilution, and the measurements
were within acceptable accuracy and precision limits for the method, and (2) the correlation between
the EMSL-LV and SAIC GC/MS results for a number of field samples substantiated the validity of
the SAIC GC/MS results (Section 4).
Difficulties were encountered in analyzing the immunoassay data because of the magnitude and
complexity of the variances affecting the methods. Due to the difficulty in determining confidence
intervals, it was not feasible to use overlapping confidence intervals to assess accuracy. The alternative
approach was to plot the immunoassay and GC/MS results on X-Y plots and determine the scatter
from the 1:1 equivalence line. As stated in the QAPjP (Silverstein et al., 1991), it was possible to
determine the number of immunoassay test results within a factor of two of the GC/MS results. This
was done for the influent samples that were high in PCP. However, for effluents that were lower in
concentration the factor-of-two criterion was not useful. Instead, the ranges of the immunoassay and
GC/MS methods were compared. The rates of false positives on the NC samples and false negatives
on the QC performance standards were also used to assess accuracy. For similar reasons, the planned
analysis of variance (ANOVA) could not be used to analyze sources of variance. The assumptions of
the ANOVA were not met, which caused too much uncertainty to be associated with those results.
Alternatively, plots of difference versus mean concentration (figures 15 and 16) were prepared for
duplicates and method splits of field samples assayed by immunoassay at each location. These plots
allowed easy determination of variability as a function of analyte concentration.
During early stages of the demonstration, the EMSL-LV laboratory experienced difficulty generating
standard curves for the PCP plate immunoassay that were comparable to those reported by WBAS.
This problem caused EMSL-LV to obtain unsatisfactory influent and effluent range-finding and QC
performance sample results. After a trial and error period to isolate the reason for poor standard
curve generation, the EMSL-LV laboratory obtained acceptable standard curves by adjusting the
antibody concentration used in the immunoassay.
53
-------
Another problem involved differences in the sampling designs of the two SITE demonstrations.
Protocols for the bioreactor demonstration did not require the collection of raw influent samples. The
bioreactor sampling scheme was modified to accommodate the sampling needs of the immunoassay
demonstration, for which one raw influent sample was collected weekly.
Changes to the QA Plan
The following list identifies substantive changes to the QA plan:
1. The statement that all effluent samples were to be run (in the kit immunoassay) with and
without a 15 ppb internal standard spike was not correct. This was in conflict with the
sampling and analysis plan and should have been corrected in the original QA plan draft.
2. Contrary to the QA plan, no results for immunoassay analysis of QA audit samples were
reported for the predemonstration testing phase (Section 3). The QAA and QAB samples
were quantitated by GC/MS during this time.
3. A short summary report was not written at the conclusion of the preliminary evaluation phase
because there was not sufficient time to analyze and interpret all the data.
Results of the On-Site Svstems Audit
A questionnaire was prepared and used for the on-site systems audit that was conducted by an EMSL-
LV QA representative during the first week of the demonstration. Table 13 shows the results of the
audit.
The on-site auditor observed that the work area where the immunoassays were run was not kept
sufficiently clean to eliminate the possibility of contamination of the highly sensitive kit immunoassay.
The auditor noted the possibility of leakage from liquid waste containers and the fact that bulk sample
preparation of PCP-contaminated ground water was carried out in the same area as the kit
immunoassay. Though these problems were inherent to space limitations in the on-site trailer, on-site
personnel should have been instructed in more detail about the precautions necessary to minimize the
potential of contamination. The problems were corrected, and performance improved.
The sampling and analysis plan was followed closely, except in one instance when the raw influent
sample sent to EMSL-LV and WBAS was not the same as the one that was sent to SAIC for GC/MS
analysis. However, this error did not seriously affect project results and interpretations.
Sample handling, tracking, and labeling was managed well, except for one instance in which influent
and effluent sample labels were switched. After a review of the results verified the problem, a
correction was made by relabeling the samples in the data base.
54
-------
ADDITIONAL QA/QC OBSERVATIONS AND CONCLUSIONS
Changes in Optical Density Levels of Kit Immunoassav Standards
Analysis of the OD values for the low and high (3 ppb and 40 ppb) kit immunoassay calibration
standards showed two types of temporal trends associated with the standard curves. Figure 17 shows
that the on-site kit immunoassay calibration standard OD levels dropped during the first week of the
study and then remained generally lower than expected. Although the exact cause for this drop is
unknown, it is probably associated with reagent handling, storage, and/or stability. This lowering of
the OD values for the standard curve was not correlated with any change in performance. The
standard curve OD values for EMSL-LV and WBAS did not show this type of trend.
The second type of change in OD readings that was seen at all three sites was a change in the
difference between the OD readings for the 3 ppb and 40 ppb standards. Oo-site differences
diminished over time, starting from a difference of about 0.4 OD on the first day and ending with
differences of about 0.2 OD for the last week. An example of a standard curve generated on site
during each week of the analysis is given in Appendix B. EMSL-LV differences dropped from about
0.3 OD in the first days of analysis to about 0.2 OD for samples analyzed in the third week. WBAS.
differences increased over the course of the first week of analysis from approximately 0.25 OD to 0.4
OD, but were generally lower in the third week (averaging about 0.2 OD, but highly variable).
Whether these changes represent the effects of aging of reagents or are associated with other site-
specific trends is unknown.
One of the QC acceptance criteria for raw data from the strips was that the OD for the NC samples
must exceed 0.5. It was no longer possible to meet this criteria because many of the results for NC
samples in the later part of the demonstration fell below 0.5. This problem was probably associated
with reagent storage, handling, or stability.
Kit Immunoassav Results-Hand-Drawn Versus Computer-Calculated
All kit immunoassay results were calculated from calibration lines that were drawn by hand with a
straight edge. The lines were drawn with respect to the three or four standard concentrations used. In
order to determine whether this procedure was causing any systematic bias or whether noticeably
improved results could be obtained using a more formal technique, the standard curves were
recalculated by least-squares methods. On-site sample results were then predicted with the least-
squares calibration curves. Figure 18 is a plot of the least-squares results versus the hand estimated
results. This plot shows a random scatter about the line of complete agreement, with no systematic
bias due to hand-drawn estimation. Plots of the least-squares results versus GC/MS results or versus
plate values (not shown) were not noticeably different from those using hand-drawn calibration curves.
It appears that least-squares fitting of the data would offer only marginal benefit.
Instrument Cross Calibration
Replicate analyses of the N-2,4-DNP-glycine, cross-calibration solutions on the three strip readers
used in the study were evaluated in terms of relative bias and variability. In general, the mean of the
three OD values obtained on different readers differed by 5 percent or less. Standard deviations for
55
-------
TABLE 13. ON-SITE SYSTEMS AUDIT CHECKLIST
Field forms filled out
completely and accurately
Kit imraunoassay SOP
followed
QC acceptance criteria
observed
Sampling and Analysis
Plan followed
Sample preparation and
dilution performed as per
SOP
Sample handling, labeling,
tracking and archiving
performed according to
instructions
Sample packaging and
shipment handled properly
Safety observed
Cleanliness, adherence to
GLP observed
Yes
X
X
X
X
• Yes, with numerous
exceptions
X
X
X
Yes, with few
exceptions
X
X
No
Comment
Reagent instability used caused
OD of NC to drift below 0.5
Incorrect raw influent samples
sent
One switched sample, rare 1- to
2-day lapse between collection
and analysis
One switched sample, rare 1- to
2-day lapse between collection
and analysis
Except for cleanliness of work
area
Work area not kept clean
enough to eliminate possibility
of contamination
SOP = Standard Operating Procedure
QC = quality control
OD = optical density
NC = negative control
GLP = good laboratory practice
triplicates ranged from 1 to 2 percent of the means, which is within the instrument manufacturer's
specifications of ±2 percent
56
-------
1.0 •
0.9
." 0.8
15
8 0.71
c
to
n
o 0.6 -\
.a
—• 0.5 H
ill
Q
O
0.4
0.3-
0.2 H
0.1 -
0.0 -
• •
" •" +
+ • + . .+ + + + ++ +
+ +*+ + * +v. ++ +«• .+ ++ * + ++ \
+ . ++ + . +* +++. + * *
+ ***+%+ + + + +*
11 I I I III I I I III I 11 III I I I I I III III I I III 11 I I III I I I I III I I I I I III I I I I III I 11 III I III I I I I I I III I I I I 11 III 11 I I I
*- I (M I co I •* Lnl co I <- I CM I co |*rlin|u>| »- I CM I co
< I < I < I < Rl < I m I m I CQ I m I CD I m I u I u I u
o
in co
O I O
CHRONOLOGICAL ORDER
• 3 ppb Standard
+ 40 ppb Standard
A1 = Flow period and day sample collected
Figure 17. Optical densities of the low and high standards for the on-sitc kit inimunoassay analyses.
-------
100 -
u_
co
LU
CC
<
O
CO
I—
CO
<
LU
_J
Q
LU
§
<
9
CC
LU
I-
Z5
CL
O
O
75 -
50 -
25 -
i
0
i
25
i
50
i
75
100
HAND DRAWN
Figure 18. Comparison of the results of sample concentrations determined on-site by graph paper and
straight-edge ruler plotting versus a least squares fit of the same samples. (Units are in
ppm of PCP.)
58
-------
SECTION 6
CONCLUSIONS AND RECOMMENDATIONS
Results of the SITE demonstration indicated that the WBAS kit and plate immunoassay technologies
provide effective screening capabilities for the analysis of PCP in aqueous samples. Rapid, portable,
and cost effective, the immunoassays measured approximate concentrations of target analytes,
exhibited little tendency toward false negatives, and provided real-time data. The demonstration
exhibited the utility of immunoassays as analytical tools that can be used in the field to complement
conventional laboratory methods.
In spite of the positive results, the demonstration also reflected the need for continued development of
QA/QC guidelines and protocols to improve the quality of the data. For the kit immunoassay,
defining new QC acceptance criteria, raising the stated detection limit, and incorporating more
procedural precautions in the SOPs should significantly improve performance. For both the kit and
plate immunoassays, incorporation of stricter QC guidelines in the development of reagents could
improve immunoassay reagent stability and performance.
The evaluation of the plate immunoassay was a secondary objective of this demonstration. The plate
immunoassay, which was not evaluated on site, exhibited better precision and accuracy than the kit
immunoassay, with quantitative results closer to those generated by the GC/MS. The plate
immunoassay is field portable. Although it requires somewhat longer processing time to operate, it
has a higher sample throughput than the kit immunoassay. The plate immunoassay may require more
training to operate than the kit immunoassay. Like the kit immunoassay, the plate immunoassay
requires additional development of QA/QC guidelines.
KIT IMMUNOASSAY CONCLUSIONS AND RECOMMENDATIONS
Kit Immunoassav Conclusions
The kit immunoassay performed well, providing a semiquantitative estimate of the approximate PCP
concentrations. However, the variability of the results was higher than expected The variability by
analysis location of the accuracy and precision suggests a significant operator-dependent or procedural
component in the error. In addition, data on QA and field samples run at several dilutions indicate
significant systematic error associated with sample dilution.
The false negative rate was low (2 to 6 percent) and partly due to over-dilution in the sample analysis.
A low false negative rate is critical for a screening method. The high false positive rate (19 percent )
on NC sample analyses apparently resulted, in part, from the developer laboratory setting a method
detection limit that was too low compared to the lowest detection limit of the standard curve.
59
-------
The kit immunoassay results were systematically biased high compared to the results from the GC/MS
analysis. A similar bias is not seen when the kit immunoassay is compared to the plate immunoassay.
The GC/MS results may be biased low due to incomplete extraction efficiency during sample
preparation using Method 3510 or to factors inherent to the GC/MS procedure itself. In addition, the
immunoassay results may be biased high due to the reported cross-reactivity with the substantial levels
of tetrachlorophenol found in the GC/MS analysis of randomly selected influent samples. However.
these levels of tetrachlorophenol would not account for all the high bias based on the reported cross-
reactivity (Section 5). Table 1 provides a list of cross-reactive compounds for the kit and plate
immunoassays. In addition, other single-laboratory effects may contribute to the bias. The kit
immunoassay replicate results on PCP standards during the preliminary evaluation phase were biased
high by 30 to 47 percent. Thus, it appears that a substantial component of the bias for the kit
immunoassay is inherent in the method. This error may be associated with the effects of curve fitting
and linearity on method quantitation. The bias is also evident in the range of values obtained in the
effluent samples. Though the net effect of the bias is marginal, it does minimize the potential for false
negative responses from the kit immunoassay.
Other conclusions about the kit immunoassay are presented below:
• The factor-oi-rwo accuracy DQO was met in most cases (88 percent for on-site analyses).
• The concentrations of the kit immunoassay and GC/MS comparison samples were in
relatively close agreement. In fact, the kit immunoassay values were considered good
enough to be used as a validation tool for the GC/MS results in the BioTrol bioreactor
demonstration. When the GC/MS concentrations for influent samples were unexpectedly
low, EMSL-LV was contacted by the RREL data interpretation and QA staff to find out if
the immunoassay results were also low. This confirmation allowed RREL to investigate
other factors as to why the influent samples were so low.
• The kit immunoassay was able to detect the same basic trends in the samples collected from
the bioreactor as the GC/MS. These findings included the high concentrations and wide
ranges of influent samples (-0.1 to 50 ppm PCP) and the relatively low and constant
concentrations of PCP (-0.01 to 3 ppm) detected in the effluent samples.
• The kit immunoassay method is quicker and requires less technical skill than the GC/MS.
The kit immunoassay results were obtained on site and within hours of sample collection,
whereas the GC/MS requires a minimum of several days for sample shipment and off-site
sample extraction and analysis steps. Field personnel were trained in 4 hours to use the kit
immunoassay. However, the kit immunoassay responds only to PCP and to a lesser extent
to structurally similar compounds, while GC/MS can identify and analyze a wide spectrum
of organic compounds.
• The variability of the kit immunoassay was higher than desired, based on the results from
QA and QC performance samples that fell within ±50 percent of the nominal
concentrations. Precision ranged from 25 to 60 percent, depending on the performance
(QA/QC) sample type and the analysis location. The variability observed in this study
would categorize the kit immunoassay as a semiquantitative method.
60
-------
In addition, the kit immunoassay method generates far less hazardous waste than the GC/MS (<10
mL aqueous wash versus 1 L of methylene chloride per sample analysis). Plotting calibration curves
by hand and estimating concentrations using graph paper did not produce significantly different
results from sample concentrations calculated with computer-based, least-squares methods.
Table 14 presents a comparison of method performances and other critical comparison parameters
related to methods for analyzing PCP in aqueous media (modified from Van Emon and
Gerlach.1990). The percentages given for accuracy and precision, for the immunoassays and EPA
Method 8270. were those found in this demonstration.
Kit Immunoassav Recommendations
The kit immunoassay shows potential as a technique that can be used as a semiquantitative field
screening method for site characterization and remediation activities. The method should be used in
conjunction with initial confirmatory analyses to assess possible site-specific or matrix interferences.
In addition, the kit immunoassay could be used to check for contamination of field blank samples and
sampling equipment.
The usefulness of the kit immunoassay can increase with refinement in various procedural.
documentation, and QA/'QC limits and confidences. Method improvement recommendations include:
• Defining new QC acceptance criteria for raw data, such as:
a) maximum (±) differences between duplicate samples at high and low concentrations.
b) tests for linearity of the calibration curve.
c) minimum OD value for NC samples.
• Raising the stated level of detection to lower the false positive rate (e.g., mean response for
NCs + 2 standard deviations).
• Rewriting the kit immunoassay SOP to emphasize stricter adherence to critical procedural
steps to improve on precision and accuracy of the method (e.g., strict adherence to pipetting
protocols to limit substrate contamination).
• Improving QC protocols in the formulation of the immunoassay reagents by:
a) documenting the shelf life and stability claims (e.g., temperature affects) for all kit
reagents.
b) reformulating reagents to improve stability, if necessary.
• Attempting to expand the relatively narrow linear dynamic range (3 to 40 ppb) of the
calibration curve by plotting data on log-log or log-logit plots or by adjusting the levels of
antibody or enzyme-labeled conjugate.
61
-------
TABLE 14. COMPARISON OF METHOD PERFORMANCES FOR PCP ANALYSIS
IN AQUEOUS SAMPLES
Performance
parameters
Detection limit (ppb)
Linear dynamic range
(Ppb)
Precisionc
Accuracy'
Analysis time based
on sample load
(detection only)
Extraction required
Cost/sample
Key interferents
% matrix spike
recovery
Rapid on-site
analysis capability
Total analysis tiraed
WBAS, kit
iramunoassay
3-5
3-40
±30-40%
±50%
0.5 hour/10 samples
No
S7JO
2J.5,6-tetra-
chlorophenol
75-125%
Yes
1.5 hours/10 samples
WBAS, plate
immuDoassay
30-40
30^00
±20-30%
±40-50%
2.5 hours/40 samples
No
$150
2,3,5,6-tetra-
chlorophenoi
75-125%
Yes
5 hours/40 samples
EPA Method 8270.
GOMS*
30-50
30-200
±20-30%
-10 to -25%
1 hour/1 sample
Yes
J300-S750
Various
10-95%
No
5 hours/1 sample
EPA Method 604.
GC"
1-15
1-200
±20%
±30%
0.5 hour/1 sample
Yes
J100-S300
Polyaromatic
hydrocarbons.
matrix dependent
20-80%
No
4.5 hours/1 sample
a From Test Methods for Evaluating Solid Waste, SW-846, (OSWER 1986).
b From Methods for Organic Chemical Analysis of Municipal and Industrial Wastewater, (EPA 1982).
c Results from this field and earlier laboratory studies.
Includes extraction, cleanup, detection, quantification, data package assembly, and associated quality assurance.
• Replacing the 8-well microtiter strip format with 12-well, microtiter strips (or 2 by 8 well
strips) so that negative and positive control samples and performance samples, along with
unknowns (i.e., environmental samples), can be included in each run.
• Using precision pipettors as standard equipment in the field kit.
• Investigating the cause(s) of the bias between the immunoassay and the GC/MS results.
62
-------
PLATE IMMUNOASSAY CONCLUSIONS AND RECOMMENDATIONS
Plate Immunoassav Conclusions
The accuracy and precision of the plate immunoassay were generally better than for the kit
immimoassay. The plate immunoassay data were also more comparable to the GC/MS data. In
addition, quantitation of the QA audit samples was more accurate and precise, and the false positive
and negative rates were lower. However, the initial problems with the reagents forced the reanalysis of
a large quantity of samples at one analysis site.
The plate immunoassay performed reasonably well in terms of its comparison to the GC/MS results.
the kit immunoassay results, and the interlaboratory results. In all cases, the effluent sample
concentrations compared well across all methods, and the higher concentration influent sample results
were generally within a factor of two of each other. In fact, the results for 17 of the 18 influent
samples analyzed at EMSL-LV by the plate immunoassay were within a factor of two of the GC/MS
results. The plate immunoassay, like the kit immunoassay, appeared to be biased high when
compared to the GC/MS. The loss in extraction and cross-reactivity factors could have caused much
of this high bias. However, when compared to each other, the immunoassay technologies exhibit no
bias.
Summary conclusions about the plate immunoassay are presented below.
• There was good relative agreement between the plate immunoassay and the GC/MS results.
• The plate immunoassay proved to be more quantitative than the kit immunoassay; however,
the plate immunoassay results were more variable than desired. This is evident particularly
when the EMSL-LV and WBAS laboratory plate immunoassay results are compared.
• No false negative responses were generated by the plate immunoassay based on the influent
and effluent field samples and the QA audit and QC performance samples.
• No false positive responses were generated by the plate immunoassay based on the NC
samples.
• The information obtained on the performance of the plate immunoassay provided important
supplementary data for the previous study, which dealt primarily with surface, drinking, and
ground water spiked with PCP. The SITE demonstration data added the aspect of
environmental water samples contaminated with PCP to the assessment of the plate
immunoassay results presented in Van Emon and Gerlach (1990).
Plate Immunoassay Recommendations
The plate immunoassay can be useful for the analysis of PCP in water samples. Although it is field
portable, the plate immunoassay is more complex to perform than the kit immunoassay. However,
since it is more quantitative than the kit immunoassay and has a larger sample throughput per run
(i.e., 96-well', microtiter plate versus 8-well strips), the plate immunoassay presents some advantages
over the kit immunoassay in both fixed laboratory and field laboratory environments.
63
-------
The technology could be improved if more thorough QC protocols were developed in the formulation
of the immunoassay reagents. These protocols are needed to document the shelf life and stability
claims for all reagents, especially the anti-PCP antibody, and to reformulate the reagents to improve
their stability and uniformity. Other conclusions and recommendations concerning the plate
immunoassay can be found in Van Emon and Gerlach (1990).
JOINT SITE DEMONSTRATION CONCLUSIONS
The WBAS immunoassay demonstration reflected the advantages of joint SITE demonstrations. The
bifocal nature of this SITE demonstration proved timely and cost effective. It also showed that one
set of confirmatory methods can be used to assess multiple technologies if analytical controls on these
methods are understood by all demonstration participants. This joint demonstration underscored the
importance of careful planning, organization, coordination of effort, and communication among
participants.
64
-------
REFERENCES
OSWER (Office of Solid Waste and Emergency Response), 1986. 'Test Methods for Evaluating Solid
Waste, Vol. 1 B: Laboratory Manual Physical/Chemical Methods." Third Edition, EPA/SW-
846. U.S. Government Printing Office, Washington. D.C
Personal Communication. 1991. Paulette Yongue, Mantech Environmental Technology, Inc..
Research Triangle Park, North Carolina.
SAIC (Science Applications International Corporation). 1989. "Demonstration Test Plan for BioTrol.
Inc. Biological Treatment of Contaminated Ground Water at New Brighton. Minnesota.
Science Applications International Corporation, 1 Sears Drive. Paramus. New Jersey.
Silverstein. M. E.. R. J. White, R. W. Gerlach, and J.M. Van Emon. 1991. "Superfund Innovative
Technology Evaluation Program Demonstration Plan for Westinghouse Bio-Analytic Systems
Pentachlorophenol Immunoassays." EPA/600/4-91/028. Environmental Monitoring Systems
Laboratory, U.S. Environmental Protection Agency, Las Vegas, Nevada.
Stinson, M. K, H. S. Skovronek, and T. J. Chresand. 1991. "EPA SITE Demonstration of BioTrol
Aqueous Treatment System." J. Air Waste Manage. Assoc., 41(2): 228-233.
Superfund Amendments and Reauthorization Act of 1986. 99th Congress. SARA, Public
Law 99-499. Section 311 (b): 130-136.
U.S. Environmental Protection Agency. 1982. "Methods for Organic Chemical Analysis of Municipal
and Industrial Wastewater," J. E. Longbottom and J. J. Lichtenberg, (eds.), EPA-600/4-82/057,
U.S. Environmental Protection Agency, Cincinnati, Ohio.
U.S. Environmental Protection Agency. 1987. "Quality Assurance Program Plan." EPA/600/X-87/241.
Environmental Monitoring Systems Laboratory, U.S. Environmental Protection Agency, Las
Vegas, Nevada.
U.S. Environmental Protection Agency. n.d. Technology Evaluation Report: SITE Program
Demonstration Test. Biological Aqueous Treatment System for Wood Preserving Site
Groundwater.by BioTrol, Inc., (Vols. I and II), Draft Report, U.S. Environmental Protection
Agency, Cincinnati, Ohio.
Van Emon, J. M. 1989. "EPA Evaluation of Immunoassay Methods." In: Immunochemical Methods
for Environmental Analysis, Symposium Series, J. M. Van Emon and R. O. Mumma (eds.).
American Chemical Society, Miami, Florida.
65
-------
Van Emon. J. M. and R. W. Gerlach. 1990. "EPA Evaluation of the Westinghpuse Bio-Analytic
Systems Pentachlorophenol Immunoassays." EPA/600/X-90/146. Environmental Monitoring
Systems Laboratory, U.S. Environmental Protection Agency. Las Vegas, Nevada.
Van Emon. J. M. and R. W. Gerlach (eds.). In preparation. Protocols Used in the SITE
Demonstration of the Westinghouse Bio-Analytic Systems Pentachlorophenoi Immunoassays.
U.S. Environmental Protection Agency, Las Vegas.
66
-------
APPENDIX A
SITE DEMONSTRATION OF BIOLOGICAL
TREATMENT OF GROUND WATER BY BIOTROL INC.
AT A WOOD PRESERVING SITE IN NEW BRIGHTON, MN
by
Mary K. Stinson, ORD/RREL-USEPA, Edison. NJ
William Hahn. SAIC, Paramus, NJ
Herbert S. Skovronek, SAIC, Paramus. NJ
67
-------
SITE DEMONSTRATION OF BIOLOGICAL
TREATMENT OF GROUNDWATER BY BIOTROL. INC.
AT A WOOD PRESERVING SITE IN NEW BRIGHTON. ."_*'
by
Mary K. Scinson. ORD/RREL-USEPA, £dison. NJ
'Jilliaa Hahn. SAIC. Paramus. N'J
Herbert S. Skovronek. SAIC. Paraaus. .VJ
ABSTRACT
A wood preserving sice in New Brighcon. .U.N on EPA's National Priorities
List was selected for evaluation of a groundwater treatment for pentachloropnenoi
with a fixed-fila biological system. The system employs indigenous microorganisms
but is also amended with a specific pentachlorophenol-degrading bacterium. The
mobile, pilot-scale unit used for the demonstration houses a 540 gallon, three-
stage bioreactor filled with structured PVC packing for biomass support. After
an initial acclimation period, groundvater from a well on the site was fed to
the system at 1, 3. and 5 gpo with no pretreatment other than pH adjustment.
nucrienc addition, and temperature control. Each flow regime was maintained for
about two weeks while samples.were collected for extensive analyses.
AC 5 g?n. the system was capable of achieving about 96* removal of the
pentachlorophenol in the incoming groundvater and producing effluent
pentachlorophenol concentrations of about 1 ppra, which met the local POTV
requirement for discharge. At the lower flows (1 and 3 gpm) removal was higher
(about 99%) and effluent pentachlorophenol concentrations were well below 0.5
ppm.
Operating costs, including power (pumping of liquids and heating),
nutrients and caustic, and operator labor, are reported. This system appears to
be a compact and cost-effective treatment for pentachlorophenol-contaiainated
wastewaters. Pre- and post-treatment such as for oil or solids removal, may be
required on a site- and wastewater-specific basis.
The results reported in this paper are preliminary and a full report is
in preparation. This paper has been reviewed in accordance with the U.S.
Environmental Protection Agency's peer and administrative review policies and
approved for presentation and publication.
68
-------
INTRODUCTION
Soil ana erounawacer cincaninacion bv chemicals conmoniy resuicir.s f"Q
vood preserving operacions has freauencly been found at Suoerf'-nd sices on -the
N'acionai Priorities List. I'.-.der che auperrur.a Araenomencs ana Reaucnorizacicn Ace
of 1936 (SARA). ;.-.e U.S. Environmental Protection Agencv --as erapouerea ;o
initiate a S.uper£und inr.ovacive Technology ^valuation iSITE) program co deveioo.
denonscrace. ana evaluate new and innovacive cechnologies chat could be used ac
Sup«rfund sices. A method for the destruction or removal of hazaroous
chemical species sucn as pencacniorophenoi (PC?) ana creosoce-derived poiynuciear
aromacic hydrocaroons (PAHs) found ac vood preserving sices was deereea co be
suitable for ir.vescigacion under chis program.
SioTroi. :.-.c. of Chaska, MN offered a biochemical descruccion cecnnoiogy
and encouraging claims fr=a earlier, small-scale scudies chac indicated :hac
efficient removal of such pollutants from contaminated soil ana groundwacer could
be achieved, '-"Mle btotreatrsent has a long history as a cost-effective
destructive method for organic chemicals in both industrial and municipal
wa*tewacers. :.~ was uncertain wnetner such cecnnoiogy would be effective at
Superfuno sices for the recalcitrant chemicals that might be encountereo as a
result of long term wooo preserving operations, specifically pentacnlorophenoi
and poiynuciear aromatic hydrocarbons.
Subsequently, che BioTrol. inc. Aqueous Treatment System (ATS) was selected
for investigation under che SITE program. After considering aicernate sices, a
facility recently added to che National Priorities List was chosen for a pilot-
scale evaluation of the technology. The selected sice, in New Brighton. XX. a
suburb of Minneapolis, ha* been used for wood treatment with various
preservatives, including creosoce. pencachiorophenol. and chrooaced copper
arsenace since the 1920s. Teacs ac the site as part of a RI/FS Indicated chat
both che soil and cha underlying groundwater were contaminated with
pencachlorophenoi and poiynuciear aromacic hydrocarbons, even though these
chemicals were no longer being used in wood creacment. The owner and operator
of the site, the MacGlllls and Clbbs Company, agreed co host che pilot scale
testing of Che BioTrol system.
PROCESS DESCRIPTION
Th« BioTrol Aqueous Treatment Syacem (ATS) shown in Figure I consists ot
a conditioning or temper tank, a heater and heat exchanger, and a three-stage
fixed-film biological reactor. Incoming vascevacer is first brought to the
conditioning tank where the pH Is adjusted (if necessary) to jusc above 7.0 wich
causcic and inorganic nitrogen and phosphorus nucriencs are added. After passing
through the in-line heacer and heat exchanger co assure a more constant
temperature In the vicinity of 70 F, che wascewacer is introduced to the base
of the firsc of the chree bioreacclon chambers (Figure 2). each chamber is
filled wich an inert support for bacterial growth; in che study corrugated
polyvinyi chloride sheets were the support medium used (Figure 3). The influent
Is passed up through each chamber while air is injected ac che base of each
chamber through a sparger cube system, as shown in Figure 2.
69
-------
INFLUENT
PUMP
NUTRIENTS
HEAT EXCHANGER
EFFLUENT PUMP
FABLE
J i
»
\
HEATER
'CAUSTIC \ ^-SLOWER
-TEMPER TANK
v
REACTOR ^-CONTROL
PA N E L^
Figure I. BloTrol. Inc. Mobile Aquuo.is TiU.UIIUMII System (ATS) System.
-------
Air
Nuiittnta. S«mpl»
Camnc .
P)\
llf|ttfd3unpU / — ' ^\ /
<»«™« / i,o.,,d X
®X" "^S / '1«mnU
Y V\ / 3«mpi»
lit/ ©-
/ |
Ma i 1
»-^vl | |
™«" cJSJZn. IM /
Pump ^ 1 gSS I T"* E'«-*"/
/ n r /
s\
X / licjirfd
3«mpl»
llnHutnl)
(«)
[\
Well .
~\
\
1
1
1
^
\ -x
\
>»
»*
Al / Al
•Actlv«-
Blomiii
Simple
l^\
I I
^ Curton .|p Alt
| 1 ^V *'• 3«mpU
" 0
-\
)
•N
*^
-'*-
Ail
i
.
' liquid Samplt
^ ©
CUnlmr or
FUIH
I I CiiU
r-\n-~
llq.ildStnipU J1"""
"__ " SufTlflltt
(Elliu*nl| aunipw
© ©
to n>iw
1— or
••^~ M'lOGiUik
T and
U'l \ Oil 41*
liquid UainpU
(Efliutnlj
/^i
(10)
NoU: CbdMl nuirt*<> •! ••o^U point I»!M 10
Klgure 2. AIJIIUOIIS TrealincnL System (ATS) ulili S.nii|il inj; I'oiiu.s Slmun.
-------
POmnm. CHLCRIDE HIDIA
MJ/jfe^;
AJCPi/
caken
growcn
The systea is accliaaced by introducing an indigenous bacterial population
iron tne local soil. After allowing about one week for acciination of this
--a.cew.cer. ch. syttem is 'seeded" with an inoeuiua of a
specific :o rhe carget contaainant. in :his case
co che
anan. case
pentacnloropnenoi. The w«SCeweter containing che contaainant is Chen recycled
chrougn che system to allow che bacterial population to readjust, '."hen che system
is fully aaaptea co the wastevater. once-chrough processing is ready ca begin.
At the MacCillis and Ctbbs site ic vas decerained chat the quality of the
groundvater did not varrant any pretreatment . even chough ic contained a
significant level of oil (about 50-60 ppm). '."hile pretreacaenc such as oil/water
separation or solids removal may be needed in other cases, such decisions must
be site and wastewater specific. Slailariy. post- treatment decisions also depend
on che specific site. At this facility, a decision was made co install a bag
filter to collect the snail aaount of sloughed biooass that vas anticipated.
priaarily so that pollutants in the sludge could be measured as part of che EPA
investigation.
SITE 7IS7INC PROGRAM
'.'orking in collaboration with che developer of che process, it va«
decerained that operation of the system at three Increasing flow rates. 1. 3.
12
-------
ana : 5-=. ; = rresoonai.r.£ :s restcence cises o: 9. '• sna 1.3 hours, resoeccive iv.
eacn :or :vo weeics. vouid allow cne effectiveness of the process cs oe aecerniinea
at low concaainanc loaainzs ana at :.-.e oesign Level. lr. face, vnile cne screemnz
^" "rnr"a '~ L98i " "" 3:' ~he RI/FS hao s"?gescea him concentrations
,-iOO-tGO ppa; o: pencacnioropnenoi sight be present ;- c.-.e eraunowacer. vhen
-•jo wail* w.ra arilled in preparation for cr.e proiecc. i raaxintlm of ahouc were founo co be presane.
The graunowacer cbcatnea frsa c.-.e seiecceo well. :r.e ir.fluenc :a. cr.e
erfluenc iron, ana che cvo incerneaiace scazes of che bioreaccor ware oonxeoreo
cor pencacnioropnenoi. scner seaivolacile organics, chloride, and TOC. Chloride
and TOC ware aonicoreo co provide suoporcing evidence for che venaor's claim chac
pancacnloropnenoi removal occurreo by mineralisation ca uacer. carbon cioxide
and sale by cr.e following equation.
OH
!
r
S ~ N
Cl-C C-CI
I I - excess 02 ..... .-...> /, C02 * 0 . 5 H20 * : Cl"
Cl
Other paraaecers also monitored co provide a. complece hiscorv of che
jroundwacer as it passed through che system Included coeal and volatile suspended
solids, oil and grease, r.icrogen and phosphorus, volatile organics. and heavy
necals. Secause chere is always concern when creating vascewacers concaining
chlorinated aromacics. testing wa> also dona for chlorodloxins and furans.
Because this investigation was part of che SITE program and careful and
complete analytical history (and safety) was desirable, carbon adsorption units
were installed on both the aqueous discharge and on che air leaving che covered
reactor chamber. Samplings and analyses were carried out before and after chese
units co determine whether significant quantities of che conta.-ainancs were lost
by any route other Chan biodegradation.
Finally, static bibassays were carried out on che incoming groundwater.
che influent ts the reactor, and che effluent to learn whether the grounowacer
was coxic co aquatic species and whether creacment removed che chemical source
of any coxicity.
RESULTS
From comparison of cha pentachlorophenol concentrations for the
groundwacer aa removed from the wail and che effluent from che bloreaccor. i:
is clear that che BloTrol system is capable of achieving afaouc 96» removal of
pentachlorophenol ac the design flow rate, 5'gpm. And. at that flowrate. final
effluent coneentracions. before carbon polishing, are approximately 1 ppm. Table
suamarizes che pencachlorophenol removals ac che chree different flow rates.
73
-------
7A3LE 1. AVERAGE ?E>rTACHLCRO PHENOL REMVAL a"
THE 5IOTROL ACUEOUS TREATMENT SYSTEM.
"low
Race
•>S?BJ
!_
3
5
-a car
•?pm;
-2.0*
24.5 +
27.5*
Effluent
(ppmi
0.13
0.36
0.99
' c cc crtc
•;*;
Average
99.8 **
98.7 *»
97.6 «•
1
Range
£7.4-99.9-
?5.8-99.8
-9.3-99.4
* decrease wich ciae may reflecc dravaown of aquifer
'» baseo on average or daily effluents
How«ver. .; ruse be noted that ss the analytical results '-ere obtained, i:
aecaae aoparent :hac an unexpected dilution phenomenon was cccurnr.z ir. c.-.e
influent chaaner ---nere che composice influent saapies were caken. The effect vas
a significant reauceian in :he apparent "'.r.f luenc" .-.sncencrscisns ::r
rencacr.ioroonenoi (ana ocner parameters) • »na. presunaoiv. ir. t.-.e values a: "e
:vo ir.termeaiace sanoiir.z points as well, '."nere c.-.ese values snouid have oeen
•essentially t.-.e same as t.-.e values cor tr.e grounowater. ir vas ooservea tr.at c.-.ey
•-are conaideraoiy Lower. Crab samoies oocaineo bv the venaor betveen che
conairisning car.x ana che bioreactor ana anaiyiea f;r pencachioropnenoi using
anotner necnoa also contiraeo the oiscrepancy. C In t.-.is alternate mecnod. high
pressure liquid chroraacography (HPLCJ. the aqueous saapie is injected directly
onto a column at aaoienc temperature and che levels of pencacnloropnenoi neasured
wich a UV daceccor ac 2J4 na and 220 nm. Although the method is not "SPA-
approved' and vas noc subjected Co che excensive quaiicy assurance used for che
GC/MS method, an abbreviacea evaluation has demonstrated tnac the results are
reliable and comparable co those obcained by GC/MS.) Ic is believed :hac che
differences in concentrations, which were particularly significant at the lower
flov races, are the resuic of backaixing from each of che reaction chambers ir.co
che preceding mixing chaaoers. Consequently, che results being presented are
basea prinariiy on che incoming groundwacer as ic vas analyzed ac che well heaa
and che final effluent from che bioreactor. using EPA Mecnod 3510/8270. for
which che Methoa Detection Limic for pencacnloropnenoi is 50 ug/L.
AC Che icwer flow races scudied. 1 and 3 gpm. pencachloroohenoi removals
(based on che cnange from che grounauater co che effluenc) increase co 99** and
final pencacnloropnenoi concencracions of 0.1 ppm and even Less are achievaoie.
These results an sumaarized in Table 1.
The changes in chloride and TOC results (weekly) parallel che decrease in
pencacnloropnenoi ac ail flows (Table 25: however. :hey are noc sufficiencly
precise co provide more than supportive evidence for mineralisation of
pentachlorophenoi co sodium chloride, vacer. and carbon dioxide.
74
-------
7.\3LE 2. COMPARISON OF AVERAGE CHLORIDE. "C. A.ND PCP RESULTS
—
•'. «u
Rate !
-
gpm> 1
1 1
2 1
5 1
?C?
-41.
•34.
•26.
• cicn ::
9 I
1 1
5 1
i. ?
I ^4.2
1 -40.5
I -22.0
pmi
-27
-22
*17
coi iccrn TCC
.9
.7
.6
1 -24
i -32
1 -21
•11
. 3
• *
(C)
.3
.2
.0
i
i
'.
(£) - found: (c) calculated
As pare es che effort ca eonfir= tnat pentachloroohenoi was being reaovea
by biochemical =inerall:ation ano r.oc by aosorption on che biosoiids or oy
stripping due co the air in cha bioreactors. boch solids ano air emissions vere
also .-.onitorea. Although the sludge crappea in che oag filter -as :ouno ca
contain pentacnloropnenoi (24 ana 170 ?pm fcuno in cvo samoies). :r.e amount oc
sludza was so small chat iisorntvan of pentacnlorotmenoi on che biosotids sno
removal vicr. c.-.e suspenaeo solids coes not represent a significant alternate
removal r.ecnanism. Thus, even :£ ail cne susoenaed solids . effluent •
jrounawaterj procuced by iha system curing cr.e cveive oays of che i. ;?o run were
Trapped In cr.e filter. :hls would aaount co only aoout 7 Ibs of sludge. :.ven
•-rich' a oentachioropnenol content as high as 170 ppm (which was measured in a
later saapie). chis would only account for about 0.0012 Ibs of PC? or aoout 0.02*
of che total pentachlorophenot Input of about 6.05 Ibs. Similarly.
pentacnlorophenol was not present above the detection llnic in any oc the air
saaples obtained over the reactor chamber, using a modified Method 5 collection
system wich an XAD resin trap and an analytical method with a detection limit
of 1.7 ug/cubtc nmcer or 0.2 ppb. Therefore, it do«s appear chat biological
degraoatlon is. by far. che primary means of eliminating the pentachlorophenoi
from che grounowater.
Concentrations of the various polynuclear aromatic hydrocarbons as part ot
che senivoiatile fraction were below detection limits in the samples ot incoming
grounowater usad in the demonstration program. Two analyses during che pre-
demonstration testing indicated total PAHs of 145 and 295 ppo: Consequently, it
is not possible to draw any conclusions as to removal efficiency or mecnanism.
However, several PAHs. including naphthalene and methyl naphthalene at maximum
levels of 34.6 ppb and 47.9 ppb. respectively, and others at consideraoly lower
levels, were found during the modified Method 5 testing of the air emissions
from che reactor. This suggests chat some air stripping of these constituents
may be occurring.
Saall aaounts of various chlorinated dloxins were found in the effluent
(040 ng/L. using method SU8280) and, particularly, the sloughed biomass sludge.
where one saapie did exhibit 1900 ng/g of OCDD isomer. This value is currently
being re-exaained. With on. exception, an effluent sample found to contain oZ
nz/L che 2.3 .7 , 8-tetrachlorodtoxin of primary concern wa« not detected in any
of the influent, effluent, or sludge saaples using high resolution CC couplea
with low resolution MS.
The incoming groundwater was found to contain low concentrations of several
75
-------
or cr.e heavy cecais. i.-.ciudinz nickel «91 uz/LK cir.c «32 uz/L) . c=pper «25
ug/U. lead (<11 ug/L) . ana arsenic : <6.5 uz/L) from che c.nromacea cooper
arsenace wooa preservative currenciy useo ir. wooa creacaenc sc c.-.e sice, -'ten
che execution of one saaoie wnich is beiieveo to be an anomaly. :.-.ere was no
c.-.ange ir. cr.e concentrations of cr.e metals across c.-.e syscem.
Acute bioaonitonng vich fresn water minnows (96 hr scacic case) ana daohnia
.-sagna (i.3 hr scactc ceso demonstrated chat che coxicicy in the incoming
jrounowacer or cr.e influenc was essentially cacaily removed" by che creacaenc.
LC20's ir.creasea frsn an escisacea low of 0.2% (grounowacer/concrol water) for
che groundwacer :s more Chan 100» (as calculated from resuics) in che creaced
effluenc.
COSTS
Preliair^ary cost esclaaces were carried ouc by che vendor far operation
of che piloc plane at MacClllis ana Clbbs excluding che ancillary equipoenc such
as carbon units ana bag filter but including cost far nutrients, electricity.
he*c. labor ana causclc. In aadlclon. cases ware excrapoiacea by che venaor ca
a full scale syscen caoaoie of creatir.z 30 zpa of a siaiiariy contaminaceo (--0
ppn centacnloroonenoi) grounawacer Based on cr.e demonstration scuay ana ocner
information ac cheir disposal. On chese bases, operating cost at che 5 gpn and
che 30 gpm rate would be $4.24/1000 gallons ana S2.62/1000 gallons, respeccively
(Table 3). As shown in che cable, certain costs do not increase at the expected
rate. Tor exaaple. unit nutrient cost would decrease because of bulk purchase:
electricity cost/gallon created decreases because it is assumed chat with deeper
bioreaccor bed* in the 30 gpm unit (3 fc instead of 4 fc) che energy for che
compressor supplying che air would be used more efficiently: operator labor cost
also are noc expected to increase in direct proportion to che size of the unit.
TABLE 3. OPERATING COST 07 TREATME.VT (S/1000 gal)
Cose Icem at S gpm at 30 gpn
nutrients 0.042 0.017
electricity 0.416 0.216
heat 1.46 • 1.46
labor 2.08 0.69
caustic 0.24 0.24
TOTAL 4.24 2.62
These costs do not include leasing or amortization of che capital
equipment, which are approximately $3,200/montn (5 gpn mobile). $30.000 (5 gpm
skid mounted) and $80.000 (30 gpm skid mounted), respectively.
Clearly labor and heat (electrical) requirements are the major factors to
consider when treating waters at a specific site. And, of course, any site-
specific pre- or pose-treatment requirements, such as oil/water separation.
solids removal, polishing, air emissions control, etc.. would have to be factored
into che cost calculation for chat sice.
76
-------
CONCLUSIONS
On cha basis of the piloc pUne scuay carried ouc ac the HaculUis and
Cibbs sice in Minnesoca. che BioTrol process would be successful in creacir.g
groundvacer or ocher pencachlorophenol-concaainaced uascevacers (ac ~0 ppm
pencachlorophenoi) Co levels suicable for discharge co a POTV or reuse wichin
a plane. Cn« unforeseen benefit of cha creacmenc va« chac bioeoxicicy in Che
incoming groundvacer ««« eliminaced by che creacmenc.
Concaainaced wacers of differenc concencracions can be accomaodaced by
increasing or decreasing che chroughpuc rac«. recycling a porcion of che scream
or by sizing cha syscem dlfferencly. Sice-specific faccors such as groundwacer
cemperacure. aobienc cemperacure. excenc of concaoinacion vich oil and/or solids.
ece.. can all play a role in che cosc-effecciveness of overall creacmenc.
Al Chough a secondary objeccive of che scudy •-«* co evaluace chi
effeccivenesa of che syscem for removal of polyaromacic nuclear hydrocarbons chac
nighc be preaanc ac che sice as a resulc of che use of creosoce. che levels of
chese conscicuencs in che groundvacer used for che scudy were coo Low co reach
anv conclusions as co removal.
10
77
-------
APPENDIX B
EXAMPLES OF KIT AND PLATE IMMUNOASSAY STANDARD CURVES AND SAMPLE
PLACEMENT LAYOUT FOR THE PLATE IMMUNOASSAY
78
-------
SAMPLE ANALYSIS INFORMATION BY STRIP
Soffit rtumb«r - STI-A-0I-0S-/ I? J-OL8^/oM-5»re- Strip ID: 6- J
Sample Well
Code * (
5T&03 / 0
STPCJl X >- fl
STD
<~ X * &
fitfr\^~l r~. f rt
OD
ABS)
(.1*
>..&
..JL
.-3<
J.3<
./
?i
L*
1.7
,7
hse-i ~ t {Lair
RgirjS-2j^ T ^.3Hi
M4E-1 ^ i
/«
>.t
B ~
S
^.^
^.x
r
^j
- ^^
^
^
Plotted
Cone (ppb
JUt
-**
tl A
if
V.
•v.
..a.
f..2i
.QL
.&.
fn
<
^
to
ppm
X 0.001 X
X 0.001 X
X 0.001 X
X 0.001 X
X 0.001 X
X 0.001 X
X 0.001 X
X 0.001 X
^
Ami (ppb) %
Dilution Sample Cone Spike Spike
Factor (ppm) Added Roc
—
— . IT
^.o- '
JS-p /33. —
Comments:
i •> -\ y 5?
2 3 4 5 6 7 8 9 10 20 30 40 50 60
[PCP] ppb
Figure B-l. Example of Field Data Form documentation for on-site kit immunoassay analysis during period A of the demonstration.
-------
SAMPLE ANALYSIS INFORMATION BY STRIP
Sample, number- 5TJ- B~/> 1-0.2. / 07X1UG»8^ / DN-SlT£ Strip 10: fo-j
Sample Well OD Plotted
Code * (ABS) Cone (ppb
STD030 1 0.30Z 3_C5
STD lt» X -
5TDHGfr* <
Kt ;;
R. IT i
. ._ .K Jh ., /^i
NC 2 f
j#-30
(^.35"
A
S
pur
V-X
N
L ^
i- *
? 4
9 ff.
\\
\
^3i
i!/O4
{._O.Z
Lm
ST jV0
T £3 V
"S.
X
_JZ.L_
_ife.a.
-$£.$-
-3LL.OL
=^1
^
X
f
s
to
ppm
X 0.001 X
X 0.001 X
X 0.001 X
X 0.001 X
X 0.001 X
X 0.001 X
X 0.001 X
X 0.001 X
\
Amt(ppb) %
Dilution Sample Cone Spike Spiko
Factor (PP'n) Added Roc
—
—
£00 O 2
^ 000
Vod
5
_1
u..
Li..
.8..
\
X.
X
V
0
«(
B
Comments:
2 3 4 5 6 7 S 9 10 20 30 40 50 60
[POP] ppb
Figure B-2. Example of Field Data Form documentation for on-site kit immunoassjiy analysis during period B of the demonstration.
-------
SAMPLE ANALYSIS INFORMATION BY STRIP
CO
Sa_mDle. number = QAA- C-Ct>&>-tf>\ / -23Al)(}a7X •»
S1~D I F> o ~
_QfiA:01 o ^
QArt-pM Q 1
PAA-|*IO -
NC X £
A
B
s
<^>
1
all
f (
I '
i i
'- i
: s
'- f
^
OD
ABS)
£ I(>V
i.-ULtf
\I3J
5, /tb
^
Plotted
Cone (ppb]
^1
—
—
^
to
ppin
X 0.001 X
X 0.001 X
X 0.001 X
X 0.001 X
X 0.001 X
X 0.001 X
X 0.001 X
X 0.001 X
^
Amt(ppb) %
Dilution Sample Cone Spike Spiko
Factor (PP>") Added floe
—
—
• —
__/OQO _ . -2fc._0_ _.
•-/ ooo
- —
>
_3
—
^
^.
At
"
003
Comments:
2 3 4 5 6 7 8 9 10 20 30 40 50 60
(PCP) ppb
Figure B-3. Example of Field Datu Form documentation for on-silc kil immnnoussay iinulysis dnrint; Period C of ilic demonstration.
-------
0:P STANDARDS
SECOND ANTIBODY CONTROL*
• Contains No Anclyto and No PCP-Speclllc Antibody
Figure B-4. Example of the sample placement layout on a typical plate immunoassay analyzed in (he
SITE demonstration.
-------
1 I I I I I
20 30 50 70 100
I I I I I II I I
zee 300 see 1000 2000
Optical Density vi. Concentration, logarithmic (ppfa)
Fitting Method:
Four-Parameter
Regression Parameters:
a - -0.0031
b - -1.4085
c - 4.9438
d - 0.930)
R-sqr - 0.9913
Sterr - 0.0240
Calculated Concentrations:
Concentration at mid-point of 0.0. range • 140.3052
Concentration at I0X of O.D. range - 667.G631
Concentration at 901 of 0.0. range • 29.4842
Figure B-5. Example of a typical standard curve generated for the plate immune-assay.
83
-------
APPENDIX C
DATA QUALIFIER FLAGS AND DEFINITIONS APPLIED TO KIT IMMUNOASSAY
DATA DURING DATA VERIFICATION
84
-------
Flag
A = Definition: data point (sample concentration) inconsistent with other dilutions on the same
strip.
Application: This flag only applies to influent and effluent sample range-finding data or to
audit sample (QAA. QAB) data. If optical densities of range-finding strips of one or all
serial dilutions analyzed on a particular strip did not follow in a logical order (i.e., the greater
the dilution the higher the absorbance units), then an A flag was applied to the dilution(s)
(11 of 104 range-finding strips had this occurrence across all analysis sites). Field sample
(influent, effluent, raw influent) concentrations with A flags were not used when determining
the concentration of the sample for intra- and intermethod comparisons (see Appendix D).
On the other hand. QAA and QAB audit samples that had A flags applied to them (5
occurrences out of 36 strips) were included in the statistical analyses of such performance
parameters as precision, accuracy, and false negatives.
B = Definition: inconsistent calibration standard, data point not used in calculation of sample
concentration.
Application: This flag was applied to one of the four PCP calibration standards (3.0. 7.1,
16.9. or 40.0 ppb) when a straight line could only be drawn through the other three. Of the
256 strips used, 25 had calibration curves with one inconsistent standard (see Section 5 for
details, and for computer versus hand-calculated results). There was never more than one B
flag per strip. If a straight line could not be drawn using at least three calibration standards,
no B flag was applied. See the description for D flags for a discussion on when more than
one calibration standard was inconsistent
C = Definition: illegible or omitted number entry or value reported - number critical in
calculating sample concentration.
Application: The C flag was only applied to one calibration standard (out of more than
1,000) that was illegible and to one sample for the dilution factors of a set of analyses.
D = Definition: all calibration curve data suspect, no confidence in any sample data generated
from this strip.
Application: When a straight line could not be drawn between at least three of the four
calibration standards, or when the optical densities of the standards were not acting in an
expected fashion (i.e., the lower the ppb standard, the higher the optical density), the
calibration curve was considered erratic As a result, these results were not used for any
further analysis. Of the 274 calibrations used in the demonstration, 18 had this occurrence
(see Section 5 for a discussion).
E = Definition: operator-noted analytical problem with the analysis in the particular well.
t
Application: This flag was reserved for standards and samples for which comments on the
field forms indicated analytical problems such as reagents omitted from analyses or bubbles
85
-------
observed in the well. These data points were not used in any sample or statistical
calculations. There were only 6 of these occurrences, representing less than 0.2% of the kit
immunoassay data.
F = Definition: inconsistent matrix spike-related result.
Application: These flags were applied to matrix spike samples exhibiting grossly poor (and
negative) spike recovery results. Of the 82 spiked samples. 7 have F flags applied to them.
See Section 5 for the discussion of matrix spike performance.
G = Definition: deviation from standard protocol, strip results reported twice.
Application: For only one strip, the sample results were reported twice by the analyst. This
is a deviation from standard protocol, and the second set of results were not used in any
statistical analyses because results would be improperly weighted.
ZZ = Definition: sample result used in the determination of PCP concentration to represent the
sample in intra- and intermethod comparisons.
Application: Sample results with the ZZ flag for each sample were used in the comparisons
of kit to plate immunoassay and immunoassay to GC/MS methods. For each sample, if there
was more than one result with a ZZ flag applied to it, the average was taken to determine the
PCP concentration to be used in the comparisons (see Appendix D for a detailed
explanation). NOTE: These flags were also applied to plate immunoassay and GC/MS
samples for the method comparisons.
86
-------
APPENDIX D
ALGORITHMS USED TO DETERMINE PENTACHLOROPHENOL CONCENTRATIONS
IN SAMPLES USED IN METHOD COMPARISONS OF THE
KIT AND PLATE IMMUNOASSAYS
87
-------
Kit Immunoassav
Samples were analyzed by the kit immunoassay method according to specific procedures in order to
yield a variety of performance and concentration data for each sample. Five field kit strips per
sample, resulting in ten or more "valid" sample concentration results, were not unusual. As a result, it
was difficult to determine which sample concentration would "represent" each sample, an issue that
was not adequately considered in the demonstration or quality assurance project plans. In an effort to
select fair and unbiased sample concentrations for intra- and intermethod comparisons, sample
concentrations were chosen based on an approach that might be used by a field analyst who had little
or no prior knowledge of the amount of analyte expected.
Specifically, an analyst using the kit immunoassay would first perform a range-finding (screening) step
in order to ascertain whether there was any detectable pentachlorophenol (PC?) and, if so, which
dilution would bring the analysis within the linear (calibration) range of the method. After the proper
range was estimated, replicate analyses could be conducted at that optimum dilution. With this
background, the following logic was used to select the PCP concentration for each kit immunoassay
sample used in the method comparison analyses:
1) If a duplicate or method split strip was run. the average from the first pair was used.
(For influent samples the mean of the RI (routine influent) and SRI (split of routine
influent) samples was used; for effluent samples the mean of RE (routine effluent) and
SRE (split of routine effluent) samples was used.
2) If no duplicate or method split analyses were available for the sample and a matrix
spike strip was run for the sample, then the average from the pair of "unspiked"
samples (i.e., mean of the RIMS-1 and RIMS-2 and the mean of effluent samples
coded REMS-1 and REMS-2) was used. (NOTE: RIMS and REMS are matrix spike
"pre-spike" samples for influent and effluent samples; the numbers 1 and 2 refer to the
two spikes per strip).
3) Samples were reanalyzed after the original analysis day because of either a request
resulting from the EMSL-LV QA data review or an indication that the results seemed
suspicious for various documented reasons (e.g., bad calibration curves, presence of
bubbles in the wells, or a preconceived expectation by the operator of the sample
concentration or the optical density). It is standard QA practice to report reanalysis
results instead of original analysis results if the original results are suspicious and the
reanalyzed results are considered sound and valid. In these cases, the same rationale
for steps 1 and 2 were used to determine sample concentration, as appropriate.
4) If, after performing the previous 3 procedures, no appropriate sample concentration
could be determined, then the average of the two lowest dilutions from the range-
finding strip was used. If only one dilution was in range, then only that concentration
was used. The less diluted samples were used to minimize possible errors and
variability associated with sample dilution, which can increase with the number of
dilution steps. (A separate statistical analysis of dilution variability was performed
using a variety of sample types and dilution factors; see Section 5.) In addition,
whenever steps 1, 2, and 3 yielded sample values that were diluted below detectable
88
-------
limits, a quantified concentration was chosen for the sample from another source (e.g.,
the range-finding strip).
5) If. when selecting a PCP concentration based on range-finding strip results, there was
a large discrepancy in the two lowest dilutions (i.e.. by a factor of 2 or more), the two
most "consistent" numbers were picked if they existed on the same or another strip.
For example, when there was 4.2. 42.0. and 36.0 ppm for 1:100. 1:1,000. and 1:2.000
concentrations, respectively, the 42.0 and 36.0 were averaged even though the 4.2 was
the lowest dilution.
Plate Immunoassav
The selection of the PCP concentrations for the plate immunoassay in the intra- and intermethod
comparisons was based on a priori knowledge of immunoassay performance and data quality
components (e.g., calibration curve linearity). The "best possible" plate concentration was determined
by averaging all the available laboratory data for each sample. This procedure differs markedly from
those used to determine the PCP data analyzed by the kit immunoassay, which was based on the
probable approach a field technician would normally take during a site investigation. The most
accurate concentration for the plate immunoassay is desirable for the rigorous comparison of the strip
method. Hence, the plate immunoassay results are based on more QA. QC, and range-finding sample
results than would be typical in a normal, production-oriented, sample analysis mode by this method.
The following logic was used for sample concentration selection for the plate technique:
1) Based on the fact that the most accurate sample concentration for a particular analysis
can be obtained from the linear portion of the standard calibration curve, samples
diluted into this range (approximately 50 to 550 ppb) for the plate immunoassay were
included as data to be pooled and averaged for the concentration of that sample. In
other words, any range-finding, duplicate, split, or matrix spike (unspiked portion)
analysis generating sample data within this 50 to 550 ppb range was considered a
reliable and defensible value to be used in determining the best estimate of the sample
concentration. After these analyses were selected, all of the concentrations were
averaged to estimate the concentration for that sample for each analysis site.
2) If none of the plate concentration analyses fell within the 50 to 550 ppb range, the
average of all analyses for the sample outside of the range was used. Although these
results may not be considered as accurate or reliable as those within this range, they
were chosen because they were still the best possible values available for the particular
sample.
3) The EMSL-LV laboratory had difficulty with the initial range-finding analyses by the
plate immunoassay method. The set of reagents used in these analyses generated
standard curves that were markedly off-set from the ranges expected, indicating that
the immunoassay conditions were not optimized. After completing the range-finding
analyses on all the field samples (on September 17,1989), another set of reagents was
supplied to EMSL-LV by WBAS. These reagents were titered to determine optimal
89
-------
levels. EMSL-LV plate immunoassay analyses that were performed with the new set
of reagents had standard curves in the expected concentration range. Subsequently,
the duplicate, split, and spike sample analyses were performed and almost all of the
samples originally analyzed on range-finding strips were reanalyzed. Data generated
after September 17. 1989 were selected, when available, instead of initial results
obtained from the first set of reagents. .
90
-------
APPENDIX E
PARTICIPATING PERSONNEL
Many individuals from various organizations contributed to the success of the WBAS immunoassay
demonstration and participated in the preparation of this report Their names, organizations, and
contributions are provided below.
Responsibility
Authors
SITE Matrix
Manager
Project Management
Quality Assurance
Experimental Design
i
Health and Safety
Name
M.E. Silverstein
RJ. White
R.W. Gerlach
J.M. Van Emon
E.N. Koglin
J.M. Van Emon
M.E. Silverstein
W.D. Munslow
M.K Stinson
W. Hahn
H.S. Skovronek
S.D. Soileau .
TJ. Chresand
R J. White
VA. Ecker
M. E. Silverstein
D.G. Easterly
R. Schmon-Stasik
R.W. Gerlach
G.T. Flatman
F. Padilla
Affiliation
Lockheed Engineering & Sciences Co., (Las Vegas, NV)
Lockheed Engineering & Sciences Co., (Las Vegas, NV)
Lockheed Engineering & Sciences Co., (Las Vegas. NV)
U.S. Environmental Protection Agency, (Las Vegas, NV)
U.S. Environmental Protection Agency, (Las Vegas, NV)
U.S. Environmental Protection Agency, (Las Vegas, NV)
Lockheed Engineering & Sciences Co., (Las Vegas, NV)
Lockheed Engineering & Sciences Co., (Las Vegas, NV)
U.S. Environmental Protection Agency, (Edison, NJ)
Science Applications International Corporation, (Paramus, NJ)
Science Applications International Corporation, (Paramus, NJ)
Westinghousc Bio-Analytic Systems, (Rockville, MD)
BioTrol, Inc., (Chaska, Minnesota)
Lockheed Engineering & Sciences Co., (Las Vegas, NV)
Lockheed Engineering & Sciences Co., (Las Vegas, NV)
Lockheed Engineering & Sciences Co., (Las Vegas, NV)
U.S. Environmental Protection Agency, (Las Vegas, NV)
Science Applications International Corporation, (Paramus, NJ)
Lockheed Engineering & Sciences Co., (Las Vegas, NV)
U.S. Environmental Protection Agency, (Las Vegas, NV)
Science Applications International Corporation, (Paramus, NJ)
91
-------
Responsibility
(«•
Logistical Support
Analytical Support
Data Base
Management
Statistical Analysis
Graphics
Technical Editor
Word Processing
Reviewers
Name
RJ. White
H.S. Skovronek
S. Stavrou
RJ. White
S.W. Ward
D. Youngman
J.F. Armour
S. Stavrou
K. Eckert
P.Lin
N. Rottunda
J. Arlaukus
S.D. Soileau
A.C. Neale
R.W. Gerlach
L-A, Gurzinski
S.O. Garcia
D. Baumwoll
P. Killinsworth
J.M. Nicholson
D. W. Sutton
J. Y. Aoyama
P.F. Showers
C McPherson
P.S. O'Bremski
KL. Fuller
A. Viray
L.N. Steele
S-A. Riesselmann
DJ. Chaloud
J.K. Rosenfeld
Affiliation
Lockheed Engineering & Sciences Co., (Las Vegas, NV)
Science Applications International Corporation, (Paramus, NJ)
Science Applications International Corporation. (Paramus. NJ)
Lockheed Engineering & Sciences Co., (Las Vegas, NV)
Lockheed Engineering & Sciences Co., (Las Vegas, NV)
Lockheed Engineering & Sciences Co., (Las Vegas, NV)
Lockheed Engineering & Sciences Co., (Las Vegas, NV)
Science Applications International Corp., (Paramus, NJ)
Science Applications International Corp., (Paramus. NJ)
Science Applications International Corp., (San Diego, CA)
Science Applications International Corp., (San Diego, CA)
Science Applications International Corp., (San Diego, CA)
Westinghouse Bio-Analytic Systems, (Rockville, MD)
Lockheed Engineering & Sciences Co., (Las Vegas, NV)
Lockheed Engineering & Sciences Co., (Las Vegas, NV)
Lockheed Engineering & Sciences Co., (Las Vegas, NV)
Lockheed Engineering & Sciences Co., (Las Vegas, NV)
Science Applications International Corp., (McLean, VA)
Computer Sciences Corporation, (La* Vegas, NV)
Lockheed Engineering & Sciences Co., (Las Vegas, NV)
Lockheed Engineering & Sciences Co., (Las Vegas, NV)
Lockheed Engineering & Sciences Co., (Las Vegas, NV)
Lockheed Engineering & Sciences Co., (Las Vegas, NV)
Lockheed Engineering & Sciences Co., (Las Vegas, NV)
Lockheed Engineering & Sciences Co., (Las Vegas, NV)
Lockheed Engineering & Sciences Co., (Las Vegas, NV)
Lockheed Engineering & Sciences Co., (Las Vegas, NV)
Lockheed Engineering & Sciences Co., (Las Vegas, NV)
Lockheed Engineering & Sciences Co., (Las Vegas, NV)
Lockheed Engineering & Sciences Co., (Las Vegas, NV)
Lockheed Engineering & Sciences Co., (Las Vegas, NV)
92
-------
Responsibility Name Affiliation
M J. Miah Lockheed Engineering & Sciences Co., (Las Vegas, NV)
VA. Ecker Lockheed Engineering & Sciences Co., (Las Vegas. NV)
M.L. Pomes U.S. Department of Interior (Lawrence, KS)
J.F. Brady Ciba-Geigy (Greensboro, NC)
M. Bernick Roy F. Weston, IncJREAC (Edison, NJ)
93
------- |