SEPA
         United States
         Environmental Protection
         Agency
          Office of Research and
          Development
          Washington DC 20460
EPA/600/R-94/112
June 1994
Development and
Evaluation of a
Quantitative  Enzyme
Linked Immunosorbent
Assay (ELISA) for
Polychlorinated Biphenyls

-------

-------
                                                     EPA/600/R-94/112
                                                     June 1994
          Development and Evaluation of a Quantitative Enzyme
    Linked Immunosorbent Assay (ELISA) for Polychlorinated Biphenyls

                               by

              Jeffre C. Johnson and Jeanette M. Van Emon
             Environmental Monitoring Systems Laboratory
                     Las Vegas, Nevada 89119
ENVIRONMENTAL MONITORING SYSTEMS LABORATORY-LAS VEGAS
            OFFICE OF RESEARCH AND DEVELOPMENT
           U.S. ENVIRONMENTAL PROTECTION AGENCY
                 LAS VEGAS, NEVADA  89193-3478

                                                 d§) Printed on Recycled Paper

-------
                                   ACKNOWLEDGEMENTS

The authors would like to thank Virginia Kelliher, a Senior Environmental Employment (SEE)
Program enrollee assisting the Environmental Protection Agency under a Cooperative Agreement with
the National Association for Hispanic Elderly, for carrying out cross-reactivity studies, and Viorica
Lopez-Avila and Chatan Charan of Midwest Research Institute for carrying out the supercritical fluid
extractions. Thanks is also due to James Jajicek of the National Fisheries Contaminant Reseach Center
and Elliot Smith of the ASCI Corporation for timely and constructive reviews. Finally, thanks is due
to Kim Rogers of EMSL-LV for constructive criticism during the preparation of the report.

-------
                                          NOTICE

The U.S. Environmental Protection Agency (EPA), through its Office of Research and Development
(ORD), funded and performed the research described here.  It had been subjected to both the Agency's
peer and  administrative reviews and has been approved as an EPA publication.  The U.S. Government
has the right to retain a non-exclusive, royalty-free license in and to any copyright covering this report.
                                              in

-------
                                          ABSTRACT

       A 96-well, microplate-based enzyme linked imunosorbant assay (ELISA) for the quantitative
determination of PCBs (as Aroclors) in soil has been developed and evaluated. The assay detection
limit for Aroclor 1248 in soil is 8.95 ug/Kg. The detection limit for Aroclor 1242 in soil is 10.5
ug/Kg. The assay has a linear dynamic range of 8 to 200 ng/mL in assay solution, corresponding to
soil concentration ranges of 50 to 1330 pg/Kg after correction for dilution of soil extracts into the
ELISA. Extracts of soil samples containing more than 1330 ug/Kg must be further diluted to bring
them into the assay working range.

       The assay was characterized for potential cross-reactivity using 37 structurally related
compounds. Chlorinated benzenes, phenols and anisoles which might be present as co-pollutants in
environmental samples were found to exhibit no more than 3%  cross-reactivity relative to Aroclor
1248.  Cross-reactivities less than 0.1% were the norm.  In addition, insignificant cross-reactivity was
observed for a number of environmentally occurring chlorinated pesticides.

       The long term (6 month) relative  standard deviation (RSD) for determination of Aroclor 1248
in soils was found to range from 5 to 10%, dependant upon concentration. The long term (3 month)
RSD for  determination of Aroclor 1242 was found to range  from 5 to 30%, again dependant on
concentration.

       Sample preparation procedures for the quantitative PCB plate ELISA were developed by
performing extraction recovery studies on samples spiked with C-14 radiolabeled tetrachloro biphenyl
solutions. The samples were extracted with a simplified shake extraction procedure using methanol;
the extraction provided greater than 90% extraction efficiency for samples spiked with C-14
radiolabeled tetrachloro biphenyl. Similar extraction efficiencies (as quantitated by ELISA) were
obtained  for seven PCB soil standard reference materials (SRMs).

       Spike recovery (as quantitated by ELISA) for real-world samples ranged from greater than
80% to 143%, which complies with recovery requirements stated for EPA gas chromatography (GC)
methods  (e.g. SW-846 Method 8080/81 and the  Contract Laboratory Program method). Recovery of
spike solution  from clean extraction solution gave a mean recovery of 101.9%, illustrating ELISA
accuracy.

       Effects of the sample matrix upon assay performance were examined for real-world samples by
carrying out parallelism studies.  Sets of serially diluted samples were simultaneously analyzed with
the ELISA.  The assay response, corrected for dilution,  should provide equivalent results in the
absence of matrix effects. Ten percent of the real-world samples  were analyzed in this fashion.  After
arithmetic correction for dilution, the mean calculated PCB concentration for each group of sample
replicates had  RSD's no greater than 14% (n > 3).  Because this variability is not significantly greater
than assay variability, it did  not appear that the matrix interfered with the assay.

       Three  sets of real-world samples, obtained from EPA Superfund Innovative Technology
Evaluation (SITE) demonstrations, EMSL-LV Technical Support Center demonstrations and regulatory
activities, were analyzed employing the PCB ELISA. Splits from  110 Aroclor 1248 contaminated clay
samples were obtained from a SITE demonstration study at the abandoned Indian Creek outfall
(AICO) at a Department of Energy plant in Kansas City, MO.  The splits were analyzed by a
participant in the Contract Laboratory Program (CLP) and by ELISA. The sample splits analyzed by

                                               iv

-------
ELISA were extracted using the methanol shake procedure and a subset of randomly chosen splits was
extracted using supercritical fluid extraction (SFE) employing CO2.

       From these samples 3 data sets were generated (Shake/ELISA, SFE-ELISA, and CLP).
Shake/ELISA and the CLP data sets were judged to be non-equivalent by a paired t-test; the
Shake/ELISA results were biased high, with a mean relative difference of 1.5. The SFE-ELISA results
were found to be equivalent to the CLP results by paired t-test (t = 0.8729, p = 0.39). The SFE-
ELISA results appeared to be biased low relative to the CLP results, with a mean relative percent
difference (RPD) of -14%.

       Samples were obtained from a technical support study conducted at the Allied Paper/Portage
Creek/Kalamazoo River (APPC) Superfund site in Michigan consisting of Aroclor 1242 contaminated
soil, sediment and paper pulp waste.  Sample splits were  analyzed by EPA SW-846 Method 8081 and
by ELISA. The 39 sample splits analyzed by  ELISA  were extracted using the methanol shake
extraction and Soxhlet extraction using methanol.  Aroclor 1242 contamination ranged from non-detect
at 0.5 mg/Kg to 268 mg/Kg by CLP analysis.

       For samples below 30 mg/Kg, ELISA  results were generally within 2 standard deviations of
the  Method 8081 results,  whereas at the higher levels, results by ELISA were biased lower than the
Method 8081 results by up to a factor of 6.5.  ELISA analysis of the more vigorously extracted
Soxhlet extracts from the higher level  samples provided results which were within 2 standard
deviations of the Method 8081 results. This demonstrates the need to evaluate the quantitative PCB
plate ELISA as a separate determinative technique with bias from sample preparation removed.

       An assortment of samples obtained from the EPA National Enforcement and Investigation
Center (NEIC) were extracted with the shake extraction and  analyzed by ELISA. Analysis by NEIC
using  gas chromatography indicated contamination with Aroclors 1242, 1254, and 1260. ELISA
results were on average biased high, with  a mean RPD of 37%.  The quantitative PCB plate ELISA is
approximately twice as  sensitive to Aroclor 1254 relative to  Aroclor 1242, thus the results are
consistent with analysis of 1242/54/60 mixtures with 1242 calibration.  Re-calibration with  Aroclor
1254 would provide accurate quantitative results.

       The quantitative PCB plate ELISA data tracked the confirmatory method data, although the
ELISA results appeared to have a slight statistical bias away from the standard GC based methods.
The quantitative PCB plate ELISA data for the three sets of real-world data contain more information
than results which may be generated employing commercially available semi-quantitative ELISA-based
methods, and thus, the quantitative PCB ELISA  fulfills the goal of providing an easily performed
method bridging the performance gap between GC methods and semi-quantitative ELISA.  The data
generated during the development, evaluation,  and application of the quantitative PCB ELISA strongly
suggest that the quantitative PCB ELISA can function in  a highly accurate and precise manner as a
"detection and quantitation device" when coupled to an efficient extraction procedure.

-------
                                         CONTENTS

Acknowledgements  	  ii
Abstract  	iv
Figures	vii
Tables  	viii

Section 1  Introduction  	   1

Section 2  Conclusions  	   3
   Assay Performance	   3
   Validation of ELISA Performance With Real-World Samples   	  4

Section 3  Recommendations  	   7

Section 4  Materials and Supplies	  9
   ELISA Procedure 	  9
   Extraction Procedures 	10
   Standards, Reference Materials and Spiking Solutions	11

Section 5  Experimental Procedures  	12
   ELISA Procedure 	12
   Extraction Procedures 	13
   Spiking Procedures	14
   Screening and Checkerboard Titration of anti-Sera and Coating Antigens  	14
   Development of Antibodies for PCBs 	14
   Synthesis of Coating Antigen	16
   Assay Format	16

Section 6  Results and  Discussion	19
   Dose Response Studies  	19
   Dose Response for Common Aroclors	19
   Extraction Solvent Effect on Assay Dose Response  	25
   Cross-Reactivity Studies  	25
   Extraction Efficiency Studies	25
   Extraction of C-14 Radiolabeled Tetrachloro Biphenyl Spiked Soils	30
   ELISA Performance Characteristics	34
   Effect of Sample Matrix on Assay Performance	38
   Results for Duplicate Analyses	51
   ELISA Analysis of Real-World Samples	52

References  	73
                                             VI

-------
                                          FIGURES

 Number

 Figure 1.   PCB ELISA immunization hapten	15
 Figure 2.   Inhibition ELISA schematic	17
 Figure 3.   Immunoassay dose response	20
 Figure 4.   ELISA dose response with four parameter fit	20
 Figure 5.   Dose response for Aroclor 1016  	21
 Figure 6.   Dose response for Aroclor 1221  	21
 Figure 7.   Dose response for Aroclor 1232  	22
 Figure 8.   Dose response for Aroclor 1242	22
 Figure 9.   Dose response for Aroclor 1248  	23
 Figure 10.  Dose response for Aroclor 1254  	23
 Figure 11.  Dose response for Aroclor 1260  	24
 Figure 12.  Dose response for Aroclor 1262  	24
 Figure 13.  Dose response, 5% methanol	27
 Figure 14.  Dose response, 10%;  methanol	27
 Figure 15.  Dose response, 15%?  methanol	28
 Figure 16.  Dose response, 20%  methanol	28
 Figure 17.  USATHEMA  C-14 extraction calibration curve	31
 Figure 18.  Reference Soil C-14  extraction calibration curve	32
 Figure 19.  Optical density versus log Aroclor 1248 concentration	37
 Figure 20.  Dose response for serially diluted
            Kansas City  samples, group 1	43
 Figure 21.  Dose response for serially diluted
            Kansas City  samples, group 2	43
 Figure 22.  Dose response for serially diluted
            Kansas City  samples, group 3	44
 Figure 23.  Dose response for serially diluted
            Allied Paper/Portage Creek samples, group 1  	45
 Figure 24.  Dose response for serially diluted
            Allied Paper/Portage Creek samples, group 2	46
 Figure 25.  Dose response for serially diluted
            Allied Paper/Portage Creek samples, group 3	46
 Figure 26.  Dose response for serially diluted
            Allied Paper/Portage Creek samples, group 4	47
 Figure 27.  Duplicate ELISA analyses of Kansas City samples 	52
 Figure 28.  ELISA and CLP results for Kansas City samples, group 1  	59
 Figure 29.  ELISA and CLP results for Kansas City samples, group 2	59
 Figure 30.  ELISA and CLP results for Kansas City samples, group 3	60
 Figure 31.  ELISA and CLP results for Kansas City samples, group 4	60
 Figure 32.  SFE-ELISA and CLP results for re-extracted Kansas City Samples	63
 Figure 33.  ELISA and SW-846 Method 8081 results for
            Allied Paper/Portage Creek/Kalamazoo River samples, group 1  	67
 Figure 34.  ELISA and SW-846 Method 8081 results for
            Allied Paper/Portage Creek/Kalamazoo River samples, group 2  	67
Figure 35.  ELISA and SW-846 Method  8081 results for Soxhlet
            extracts of high level  samples	69
Figure 36.  ELISA and NEIC results  for NEIC samples  	71
                                             vn

-------
                                          TABLES
Number
Table 1.   Fifty Percent Inhibition Level for Aroclors	19
Table 2.   Cross-Reactivity of Possible Co-Contaminants  	29
Table 3.   Recovery of Radiolabeled Tetrachloro Biphenyl From USATHEMA Soil	32
Table 4.   Recovery of Radiolabeled Tetrachloro Biphenyl From Soil SRMS  	33
Table 5.   Recovery of Radiolabeled Tetrachloro Biphenyl From
            Kansas City Soil/Clay Samples	34
Table 6.   ELISA Detection Limit for Aroclor 1248	35
Table 7.   ELISA Detection Limit for Aroclor 1242	36
Table 8.   ELISA Results For Aroclor 1248 Soil SRM Extracts	40
Table 9.   ELISA Results For Aroclor 1242 Soil SRMS 	41
Table 10.  ELISA Results For Serially Diluted Kansas City Soil Extracts 	42
Table 11.  ELISA Results For Serially Diluted Allied Paper Samples	44
Table 12.  ELISA Results For Serially Diluted NEIC Extracts	48
Table 13.  Spike Recovery of Aroclor 1248 From Kansas City Soil Samples, set One	48
Table 14.  Spike Recovery of Aroclor 1248 From Kansas City Soil Samples, set Two  	49
Table 15.  Spike Recovery of Aroclor 1248 From Kansas City Soil Samples, set Three  	49
Table 16.  Aroclor 1242  Spike Recovery From Allied  Paper/Portage Creek Samples	50
Table 17.  ELISA Results For Duplicate Analyses  	51
Table 18.  ELISA and CLP Results For Kansas City Soil Samples	54
Table 19.  ELISA, SFE-ELISA and CLP Results For Kansas City Soil Samples	62
Table 20.  ELISA and SW-846 Method 8081 Results for Allied Paper/Portage Creek/
            Kalamazoo River Soil Samples	65
Table 21.  ELISA Results For Soxhlet Extracts and Method 8081
            Results for Allied Paper/Portage Creek/Kalamazoo River Soil Samples	68
Table 22.  ELISA and NEIC Results for NEIC Soil Samples	70
                                            Vlll

-------
                                          SECTION 1

                                       INTRODUCTION

        This report details the development and evaluation of a 96-well microplate-based enzyme
 linked immunosorbent assay (ELISA) for the quantitative  determination of polychlorinated biphenyls
 (PCBs) in soil.  Procedures carried out during the developmental stage will be described, along with
 accompanying performance characteristic data.  In addition, the analysis of three sets of real-world soil
 samples, representing a wide variety of matrix challenges, is reported.

        After the initial detection of polychlorinated biphenyls (PCBs) in the environment in 1966
 (Jensen et al., 1966), mounting evidence led to the United States Environmental Protection Agency
 (EPA) to classify them as suspected human carcinogens, due in part to their low rate of degradation,
 their tendency to bioaccumulate, and their carcinogenic nature.  In 1976, the United States Congress
 banned PCB manufacture, processing, distribution and use, except for a handful of specific and limited
 uses. As a consequence, analytical method development resulted in the codification of a number of
 standard methods for PCBs.  One such method, Method 8080, described in the EPA Office of Solid
 Waste and Emergency Response Manual SW-846 (U.S. Environmental Protection Agency, 1986), is
 representative of the majority of PCB  analytical methods in that the method relies on rigorous
 overnight Soxhlet extraction, followed by gas  chromatography (GC) for quantitation of PCBs.

        Several factors have generated increasing interest in immunochemically  based analytical
 methods, such as ELISA, for PCBs.  In addition to the spiralling costs associated with regulatory
 compliance, current toxicological research has re-awakened the controversy surrounding the actual
 carcinogenicity and toxicity of PCBs.  In an effort to expand the array of tools available for PCB
 research, immunochemically based methods are  increasingly being relied upon to provide data under
 appropriate circumstances.

        As a result of the Supeifund Amendments and Re-authorization Act of 1986 (SARA), site
 monitoring and assessment will represent a large and possibly rising financial burden directly related to
 the collection, transport, and analysis of PCB containing samples.  ELISA based methods have been
 shown to offer the potential for data outputs which are complimentary and in some case comparable to
 established methods at a significantly lower cost on a per analysis basis.  ELISA methods also show a
 distinct advantage with regard to timeliness of data generation.

        Several ELISA based methods for PCBs have become commercially available over the past
 several years.  For example, Ohmicron Environmental Diagnostics, Inc., the Immunosystems division
 of Millipore, Inc., EM Science Inc., and Ensys, Inc., currently market tube-based immunoassays
 intended for field screening applications.  In general, these immunoassays are formatted to be used
 only for determining whether a given sample contains PCBs at a concentration above or below a set
 threshold value, although the Ohmicron assay  is intended for quantitative use. As such, these
 immunoassay kits are designed for rapid result generation, low cost, and ease of use by relatively
untrained personnel.  The benefits of such design criteria have become quite evident to the EPA,
which, under the Superfund Innovative Technology Evaluation (SITE) Program,  officially mandated
the use of low cost, rugged, field portable methods to ease the burden of using expensive, time
consuming, GC or GC/MS methods for the characterization of contaminated areas.

-------
       While the benefits of using commercial immunoassay kits are evident, the data which they
generate is complimentary but not comparable with current GC-based methods for PCBs.
Consequently, there exists a large gulf between these analytical methods.

        In the broadest of terms, this report describes work on an immunoassay aimed at bridging this
gap between the GC based instrumental methods and the commercial immunoassay  kits.  Such an
immunoassay procedure will provide data which are quantitative  and comparable to the GC based
methods for many applications, such as site characterization, mapping concentration isopliths, and
monitoring remedial activities, while providing high throughput analytical procedures which are to a
high degree as inexpensive, rapid, and simple  as the kit immunoassays.

-------
                                         SECTION 2

                                       CONCLUSIONS

ASSAY PERFORMANCE

       The quantitative PCB plate ELISA was characterized over the course of assay development
and subsequent analysis of three sets of real-world samples obtained from EPA SITE demonstrations
and regulatory activities.  Based on the Aroclors which contaminated these real-world samples the bulk
of data are focused on Aroclors 1242 and 1248. Initial characterization data show that this assay
could be used for Aroclors 1254 and 1260 equally well with similar performance characteristics.  The
remainder of the discussion centers around Aroclors 1242 and 1248.

       The assay described in the current study is intended for the analysis of PCB contamination in
solid matrices such as soil, sediment, clays and paper pulp, and thus the samples required extraction
prior to analysis.  The methanol based shake extraction procedure employed during the present study
was chosen for its simplicity, and was based on extraction procedures common to a number of field
methods (U.S. Environmental Protection Agency, 1992a, 1992b). It was not the goal of the current
study to develop extraction procedures per se,  as this represents  a potential research area in itself.
Preliminary extraction studies, with a wide variety of matrices, using a radio-labeled tetrachloro
biphenyl suggested that the extraction procedure would optimally provide an average extraction
efficiency of 92% with a relative standard deviation (RSD) of variation of about 4%.

        Extraction of commercially available "PCBs in soil" standard reference materials (SRMs),
followed by quantitation with the plate ELISA provided an indirect measure of extraction efficiency.
ELISA results for Aroclor 1248 SRMs suggest extraction efficiencies greater than 90%, while for
Aroclor 1242, employing 5 PCB levels, efficiencies ranging from 53% to 91% were observed. In all
cases for the 1242 SRMs, the reported value is within the EPA defined advisory range as specified by
SW-846 Method 8080/81.

        The assay had detection limits of 1.31  ng/mL for Aroclor 1248 with a a of 0.9 ng/mL and a
detection limit of 1.6 ng/mL for Aroclor 1242  with a o value of 0.61 ng/mL.  The detection limit  in
soil (based on a 5 gram sample) is 9.0 ng/g, a = 6.0 ng/g for Aroclor 1248 and 10.5  ng/g, o = 4.1
ng/g for Aroclor 1242 after correcting for the dilution factor imposed by adding soil extracts to assay
solution. The assay had a quantitation range of about 8 ng/mL to 200  ng/mL in assay solution,
corresponding to soil concentrations of about 50 ng/g to 1330 ng/g, or  0.05 mg/Kg to 1.33 mg/Kg.
Samples extracts containing greater than about  1.3 pg/mL PCBs require appropriate dilution to bring
the PCB concentration into the working range  of the assay.

        The assay provided  the long-term reproducibility required for use as a quantitative tool. Based
on repeated measures of Aroclor 1248 soil SRMs over a six month period, determinations were carried
out with RSD's for all SRMs of less than 10%.  Repeated measures of Aroclor 1242 soil SRMs over a
three month period provided similar performance.  Dependant on PCB  level, RSD's ranged from 30%
to 5%.

        The quantitative PCB plate ELISA was found to be highly selective for PCBs;  it exhibited
very little cross-reactivity with a large number of compounds which might potentially co-contaminate
environmental samples and interfere with accurate measurement of PCB concentrations. The  assay

-------
exhibited selectivity for PCBs which will allow the plate ELISA to be used in the presence of a wide
range of commonly occurring chlorinated anisoles, benzenes and phenol co-pollutants.  The 37
compounds which were studied in the preliminary development stage cross-react no more than about 3
percent relative to Aroclor 1248.

VALIDATION OF ELISA PERFORMANCE WITH REAL-WORLD SAMPLES

       Validation of the ELISA using comparative results obtained by standard instrumental methods
is based on several important assumptions. Statistically, this procedure can become confounded by
sampling errors, sample preparation differences, and inter-lab variation, even before variability is
introduced by true inter-method differences. Comparison of the quantitative PCB plate ELISA soil
sample results with results generated using standard methods such as SW-846 Method 8080/81 is
made difficult by the fact that performance data for the standard methods is typically limited to either
solution phase measurement data or a limited number of soil matrices.  This lack of availability of
extensive soil analysis performance data for standard methods points out the difficulty of comparing
results across soil samples; each soil matrix may present new challenges a particular method cannot
meet. Extraction procedures which work well for sandy soils may provide irregular performance
characteristics when applied to oily clay samples or sediments.

ELISA Analysis of Kansas City Samples

       ELISA analysis of Aroclor 1248 contaminated clay samples obtained as sample splits from the
Kansas City, MO  Indian Creek Superfund site provided data which are interesting considering the
above discussion.  Using a paired t-test as the basis for decision, it was found that the PCB levels as
reported by the CLP laboratory were not equivalent to the PCB levels as determined by the
quantitative PCB plate ELISA. The average relative percent difference was found to be 46%.

       One problem with such an approach is the implicit assumption that both the ELISA  and CLP
reported values represent the true mean for each respective method. The large error bars for the
methods suggest that this is not likely to be the case.  One undesirable alternative would be  running
enough replicates of each sample to ensure the validity of this assumption. This degree of rigor is
possible with immunoassay, however, given the time and expense of the CLP analyses, this  is not
practical.

       Another problem is the distinction (or lack thereof) between preparative and determinative
steps in the data generation.  The ELISA and CLP methods employed very different extraction
procedures.  It is conceivable mat most of the measured difference between methods is due to sample
preparation alone.

       Four alternative hypotheses can explain the data. The first is: the quantitative PCB plate
ELISA is not accurate but the CLP method is.  The second is:  Some interferant or interferants in the
samples themselves causes assay performance degradation. The third hypothesis is:  the quantitative
PCB plate ELISA and CLP method are both accurate, but extraction performance varied greatly
between methods.  Finally, the fourth hypothesis is: there is a large quantity of some non-PCB cross-
reacting species present in the samples which elevates the apparent concentration of PCBs as measured
by the quantitative PCB plate ELISA.

-------
       The first hypothesis can be ruled out.  Data generated for the commercial Soil SRMs as well
as quantitation of spiked solutions shows that the ELISA can accurately measure PCB concentrations.
The second hypothesis can be ruled out as well.  Data collected during parallelism studies with the
Kansas City samples show that there are no significant non-specific matrix contributions for the
ELISA results.

       The third hypothesis cannot be readily discounted.  Based on earlier work (Spittler, 1986), it
might be suspected that extraction procedures employing methanol would work better than
hexane/acetone extractions as called for in the CLP method.  Hexane/acetone extraction as specified
in the CLP method may be optimal for a mixture containing all the chlorinated analytes covered by
the method, but not for the  specific, more focused use of PCB extraction exclusively. Of course, the
simple approach taken in a  methanol based shake extraction used for the ELISA might offset the gain
realized from substitution of methanol.  Results for the SFE extracts of the Kansas City samples
illustrate these extraction issues.  Using the identical ELISA procedure, it was found that the SFE
extracts gave results  which  converged toward the CLP results. The SFE-ELISA results were
equivalent  to the CLP results by a paired t-test (t = 0.8729, p = 0.39), whereas the ELISA results for
the same samples extracted by the methanol shake procedure were not (t = 2.118, p = 0.046).

       The fourth hypothesis cannot be ruled out easily either.  If there are cross-reacting compounds,
they are not commonly occuring chloro benzenes, phenols, or anisoles. One possibility is the presence
of PCB metabolites,  such as polychlorinated biphenylols, which would not be detected by standard
methods, but which may nevertheless be measured by ELISA.

       There are certain applications for which data provided by the quantitative ELISA  could prove
very useful.  The reported data for the Kansas City samples demonstrate the use of ELISA as a bridge
between GC methods and semi-quantitative immunoassay test kits.  The majority of the samples had
concentrations well below 5 mg/Kg.  At this level, relative percent differences of 100% may
correspond to as little as 0.033 mg/Kg (the detection limit of the CLP method).  For example, to easily
obtain a quantitative PCB result of 0.1 nig/Kg, ± 0.1 may have great value when the option is either
GC analysis or a semi-quantitative "less  than 5 mg/Kg" result obtained through use of a commercial,
semi-quantitative ELISA.  The quantitative PCB plate ELISA allows for the measurement of PCB
concentration in a way which provides more information than the semi-quantitative ELISAs currently
available commercially, while using an assay procedure of similar complexity.

ELISA Analysis of Allied Paper/Portage Creck/Kalamazoo River Samples

       ELISA analysis of Kalamazoo samples obtained from the Allied Paper/Portage
Creek/Kalamazoo River Superfund site in  Michigan provides further amplification upon the points
discussed above.  The data can be  thought of as consisting essentially of two subsets, the  low level
samples (PCB concentrations below approximately 30 mg/Kg) and the high level samples (PCB
concentrations greater than 30 mg/Kg).  For the low level samples,  the bulk of the ELISA and SW-
846 Method 8081 results overlap one another within the 95% confidence intervals of the respective
methods.  For the high level samples, ELISA and Method 8081 results overlapped in a  similar manner
after more vigorous methanolic Soxhlet extraction prior to ELISA analysis.  ELISA analysis of
extracts obtained using a 20 minute shake in methanol resulted in measured values of PCB up to a
factor of 6.5 lower than ELISA results for methanolic Soxhlet extracts.  Again, parallelism studies and
spike recovery data demonstrated that the quantitative PCB plate ELISA, as  applied to the Allied
Paper/Portage Creek/Kalamazoo River samples, showed high accuracy and no assay degradation due to

-------
 matrix artifacts.  Thus, the potential utility of the quantitative PCB plate ELISA as a determinative
 method for PCBs in sediment, soil and paper waste was demonstrated by the results.

        The results for the high concentration samples illustrate the need for differentiating the sample
 preparation from the determinative step itself.  The ELISA results for the simple "20 minute
 methanolic shake" extracts and the ELISA results for the methanolic  Soxhlet extracts are very
 different.  Clearly, only the extraction efficiency plays a significant role in altering the ELISA results.

        The fact that the ELISA results for the methanolic Soxhlet extracts are convergent with the
 CLP data (the ELISA results appear to be slightly lower than the CLP results with a calculated mean
 RPD of -17%) gives rise to the hypothesis that the quantitative PCB plate ELISA, as the determinative
 step, provided comparable data to the GC, for these environmental samples, provided that appropriate
 extraction  procedures were used.

 ELISA Analysis of NEIC Samples

        The results from ELISA analysis of samples obtained from the EPA Enforcement and
 Investigation Center (NEIC) illustrate a number of points.  Analysis of several serial dilutions of the
 sample extracts  demonstrated that the assay was  not subject to performance degradation due to matrix
 artifacts. The ELISA results were generally higher than the corresponding GC results,  with an average
 RPD of 37%.

        The results for this data set raise  the issue of calibration, an issue which is universal to any
 determinative method for ArocJors.  The  samples were characterized by NEIC as being mixtures of
 Aroclors 1242/1254/1260.  The apparent  bias high on the part of the ELISA may be due wholly to
 selection of the  calibration mixture. In addition, the PCB levels reported by NEIC most likely have a
 built in bias, due to analyst judgement calls on assigning peak patterns to the various Aroclors.

        Taken as a complete method, 20  minute methanolic shake extraction followed by ELISA
 determination of PCBs appears to have bias away from standard GC based methods, at least
 statistically speaking.  The qualitative PCB plate ELISA data for the three sets of real-world samples
 contain more information than results which may be generated employing commercially available
 semi-quantitative ELISA-based methods.  Thus, the quantitative PCB  plate ELISA fulfills the goal of
 providing an easily performed method bridging the performance gap between GC methods and semi-
 quantitative ELISA.

       The real-world data indicate that  extraction procedures play a major role in the ELISA results.
 Statistically speaking, in these studies,  it is improper to compare the  accuracy of the ELISA
 determinative step with the GC determinative step, because the extraction procedures confound
 matters.

       The data generated during development, evaluation, and application of the ELISA strongly
 suggest that the  quantitative PCB ELISA  can function in an accurate and highly precise manner when
considered  as a "detection and quantitation device" separate from the  sample preparation itself.

-------
                                          SECTION 3

                                    RECOMMENDATIONS

       The analyses conducted in the course of the current study strongly suggest that the PCB
ELISA has great potential for use as a determinative step in PCB analysis, in particular, when coupled
with an appropriate sample preparation procedure.  To ascertain the performance of the ELISA as a
"detector system,"  it will be important to remove the statistical ambiguity resulting when the data
being generated by two detectors (GC/electron capture detector and ELISA) are actually the result of
measurements on two distinct soil extracts.  The two soil extracts are very likely different in their PCB
concentrations, and thus, even in the best case  scenario of 100%  accuracy in the measurement step, the
results will of course differ.

       To remove the contribution  of errors in soil extraction, it is suggested that further experiments
be carried out in which samples are extracted and portions of the extracts are quantitated by both GC
and ELISA.  Any extract cleanup procedures should be carried out before splitting the extract.
Alternatively, a study design could be structured such that analyses could be carried out on extracts
which had been cleaned up as well as extracts  which had not been subjected to additional clean-up
steps, thereby allowing for checks on the possible effects of the cleanup procedure.

       Further work could be carried out to allow unambiguous comparison between the PCB ELISA
and GC/ECD as quantitation techniques.  One  possible  scenario would entail re-extraction of  the
Kansas City and/or Allied Paper  samples, followed by analysis using ELISA and GC  (in-house and/or
CLP laboratory) of splits derived from these extracts.

       The results of this study  suggest that the ELISA has  the capability of providing useful data for
certain applications.  The indirect inhibition  format was used because, generally, it is  one of the more
sensitive formats which can be chosen from  the myriad of ELISA formats.  In addition, this format
prevents exposure  of the enzyme to potential interferants present in the original sample.  The  assay can
be configured to allow even greater ease  of use.  For example, the equilibration times may be reduced
allowing a one day assay without a  substantial change in performance.

       The plate ELISA format can be easily  automated using any number of the readily available
robotic plate ELISA instruments.  This would permit screening of large numbers of samples,  and in
addition, it could allow for the carrying out of extensive parallelism studies  on a routine basis.
Extensive quality assurance of the ELISA data output could thus be ensured.

       A note of caution is raised with respect to simple or abbreviated extraction procedures.
Typically, most of the commercially available  semi-quantitative ELISAs  for soil screening rely upon
"quick shake" extraction procedures, enabling extreme streamlining of the entire ELISA based
procedure. The experiences noted in the current study reflect the possible dangers in  assuming that
these extractions always perform adequately.

       The quantitative PCB plate ELISA can be used to measure  PCBs with high accuracy  and
precision when coupled with appropriate  sample preparation  procedures.  Further utilization should
include coupling the quantitative PCB plate ELISA with efficient and potentially fieldable extraction
methods, such as supercritical fluid  extraction. Additionally, the quantitative PCB plate ELISA could
be coupled with rapid Soxtec™ type extraction procedures, potentially in mobile laboratories,  enabling

-------
rapid, relatively non labor-intensive measurement of PCBs. This would be an analytical scheme of
high utility, and acceptance, as use of Soxtec type extraction procedures for PCBs already has
precedence in such methods as EPA SW-846 Method 3541, Automated Soxhlet Extraction (U S  EPA
1993).

       Interest in the application of the quantitative PCB  plate ELISA has been generated within a
number diverse groups, such as the EPA Great Lakes National Program Office and the Fish and
Wildlife Service, National Fisheries Contaminant Research Center.  In order to facilitate the
application of the quantitative PCB plate ELISA, the ELISA procedure, coupled to a suitable
extraction procedure should be subject to peer verification through such avenues as the Association of
Official Analytical  Chemists (AOAC) Peer-Verified Methods program.

-------
                                         SECTION 4

                                MATERIALS AND SUPPLIES

ELISA PROCEDURE

Immunochemical Reagents

              Anti-PCB anti-serum, pool AC-3, produced under EPA Contract 68-03-3511.

              Anti-Rabbit IgG-alkaline phosphatase conjugate, Sigma Chemical Company, St Louis,
              MO, # A-8025.

              Coating Antigen; 4-(2,4,5-trichlorophenoxy)-butyric acid conjugated with bovine serum
              albumin.  Synthesized as described in Section 5.

              Sigma 104 Phosphate substrate tablets,  (Sigma Chemical Company, St. Louis, MO.).

Buffers

              Phosphate buffered saline with tween 20 (PBST)  pH 7.4; 8.0 g NaCl, 0.2 g KH2PO4,
              1.15 g Na2HPO4, 0.2g KC1,0.5 mL Tween  20, 0.2 g NaN3, dissolve and dilute to 1 liter
              with de-ionized water.  Adjust pH as needed with 3N NaOH.

              Carbonate coating buffer, pH 9.6; 1.59  g Na2CO3, 2.93 g NaHCO3, 0.2 g NaN3,
              dissolve and dilute to 1 liter. Maximum storage time 2 weeks.

              Substrate buffer; 97 mL diethanolamine, 800 mL  H2O, 0.100 g MgCl212H2O, dissolve
              and dilute to 1 Liter. Adjust to pH 9.8, store in closed container in dark.

Apparatus/supplies

              Maxisorb I 96-well microplate, or equivalent (A/S NUNC, Roskilde, Denmark).

              Automated plate washer, Skatron Model A/S or equivalent (Skatron, Inc., Lier,
              Norway).

              Mechanical plate shaker, Bellco Mini-Orbital Shaker, or equivalent (Bellco
              Biotechnology, Vineland, NJ).

              Microwell plate reader, Molecular Devices Model Vmax, or equivalent (Molecular
              Devices, Menlo Park, CA).

       •      Instrument control and data analysis software for  the plate reader, Molecular Devices
              SoftMax, or equivalent (Molecular Devices, Menlo Park, CA).

              Pipettes -  repeating pipette capable of delivering  1 mL, adjustable 20  - 200 uL single
              channel pipette; 0 - 25 uL adjustable positive displacement pipette; 8-channel 50 -

-------
              200 uL adjustable pipette; and assorted class A glass volumetric pipettes capable of
              delivering 15, 20, 25 or 50 mL liquid, dependant on scale of assay.

              Borosilicate glass tubes, 12x75 mm or 13x100 mm.

              Acetate plate sealing tape (Dynatech Laboratories, Chantilly, VA) or equivalent.

EXTRACTION PROCEDURES

Methanol Shake Procedure

              Pesticide grade  Methanol

              High density polypropylene wide-mouth screw top bottle, 30-mL, Nalgene 2104-0001
              or equivalent (Nalge, Inc., Rochester, NY).

              Stainless steel balls, 1/4" diameter, 5/extraction bottle,  (Small Parts, Inc., Miami, FL,
              # J-BX-4 or equivalent).

              Self-contained syringe filtration unit, 0.45  uM PTFE frit, Whatman Uni-Prep
              UN113UORG or equivalent (Whatman Laboratories, Clifton, NJ).

              Centrifuge tubes, 15 mL polypropylene.

       •      Anhydrous sodium sulfate.

              Repeater pipette, capable of delivering 5 mL.

              Mechanical wrist-action shaker, Burrell Model 75 or equivalent (Burrell, Pittsburg,
              PA).

              IEC Centra-8R  centrifuge or equivalent (International Equipment Company, Needham
              Heights, MA).

Supercritical Fluid Extraction

              Isco Model SFX2-10 Extractor, or equivalent, equipped with a Model 260D Syringe
              Pump (ISCO, Inc., Lincoln, NE).

              Stainless steel capillary restrictors, 34.5 cm x 50 urn i.d., (ISCO, INC., Lincoln, NE).

              SEE Grade CO2.

              Anhydrous magnesium sulfate.
                                             10

-------
Carbon-14 Labeled Tetrachloro Biphenyl Extraction Recovery

       •      Beckman LS6000IC liquid scintillation counter or equivalent (Beckman Instruments,
              Inc., Fullerton, CA).

       •      Ecoscint A scintillation solution or equivalent (National Diagnostics, Manville, NJ).

STANDARDS, REFERENCE MATERIALS AND SPIKING SOLUTIONS

Aroclor Standards

              Aroclors 1016, 1221, 1232, 1242, 1248, 1254, 1260, 1262 and 1268 were purchased as
              IxlO^g/mL solutions in methanol. (Ultra Scientific, North Kingstown, RI).  These
              primary standards were diluted serially 1:2 into pesticide grade methanol using 10-mL
              Class A volumetric flasks and 5-mL class A volumetric pipettes, to obtain ELISA
              standards.

Radiolabeled Tetrachloro Biphenyl

              C-14 Ring labeled 2,2',5,5'-tetrachloro biphenyl,  50 uCi in 55 uL toluene (14.2
              Ci/mmol), (Sigma Chemical Company, St. Louis, MO). Diluted to 100 mL in pesticide
              grade methanol, corresponding to 0.5 uCi/mL or  1.07 x 10"5 g/mL of PCB.

Reference Materials

       PCBs in soil standard reference materials (SRMs) were obtained from Environmental Resource
Associates, Arvada, Colorado.
              Aroclor 1248
              Two "PCBs in Soil Quality Control Standards" with certified Aroclor 1248 levels of
              282 and 33.9 mg/Kg were obtained as mentioned above. The SRMs were prepared by
              spiking with fresh (unweathered) Aroclor 1248.

              Aroclor 1242
              Five Aroclor 1242 in soil SRMs were obtained as mentioned above. The reference
              samples were prepared by spiking  with unweathered Aroclor 1242 at five concentration
              levels. The SRMs were spiked at  0.500, 1.50, 8.00, 25, and 100 mg/Kg.
                                              11

-------
                                          SECTION 5

                               EXPERIMENTAL PROCEDURES

ELISA PROCEDURE

       This protocol is written for one 96-well microplate and should be adjusted accordingly if more
samples than can be accommodated on one plate are run.

Day One

       Pipette 20 mL carbonate coating buffer into a glass vessel using a 20-mL volumetric pipette.
Add 5uL coating antigen (freshly thawed) and mix gently.  Transfer 150 uL into each well of the
microplate using an 8-channel pipette.  Cover plate with acetate plate sealing tape and store overnight
at 4 °C.

       Pipette 35 mL PBST into a glass flask. Add 11 uL AC-3 anti-PCB serum (freshly thawed)
and mix gently. Add 1 mL of this solution to appropriately labeled glass test tubes (12x75 mm or
13x100 mm work well), one tube for each sample, calibration standard or quality control sample.  Add
175 uL of the methanol solution containing sample, standard or quality control sample to each tube.
Mix gently on a vortex mixer, cap and store overnight at room temperature.

Day Two

       Remove sealing tape, place microplate in washer and wash the coating antigen solution from
the wells (three times with PBST).  Remove microplate from washer, rotate it 180 degrees, and place
back into washer. Repeat above wash step.  Remove microplate  from washer and remove any residual
wash solution  by inverting plate and rapping it several times on an absorbant paper towel.

       Add 100 pL of solution from each of the above tubes prepared  during day one to the
microplate  in triplicate (100 uL/well x three).  Cover microplate with plate sealing tape,  place it on the
mechanical shaker, and shake for three hours at room  temperature.

       Prepare a solution of the goat anti-rabbit IgG-alkaline phosphatase (GAR-AP) conjugate in
PBST to give  15 mL of solution. The dilution factor is lot dependant, but generally ranges between
1:1000 and 1:2500.  (The dilution must be experimentally determined prior to assay by running a zero
standard using this assay protocol such that an optical density of approximately 1 to 1.5  develops after
about 0.5 hr. color development time).

       Remove microplate from shaker. Wash and dry as above.  Add 100 uL/well of the GAR-AP
solution using the 8-channel pipette.  Cover microplate with sealing tape and shake gently on the
shaker  for two hours at room temperature.

       Prepare 15 mL of  a 1 mg/ml solution of substrate in substrate buffer by placing three substrate
tablets  into a flask and adding 15 mL substrate buffer with a volumetric pipette.

       Remove microplate from shaker, and wash and dry as above.  Pay particular care to ensure all
the GAR-AP solution and  wash  solution are removed.

                                              12

-------
       Add 100 uL/well of the substrate solution using an 8-channel pipette. Ensure that the elapsed
time between adding solution to the first row of wells and the last row of wells is as short as possible.

       Place microplate in the reader and begin reading plate at 405-650 nm at about 15 minutes
elapsed time.  Continue to make readings until the optical density of the zero standard is about 1.

       Perform data reduction as  outlined in the SoftMax Documentation.  For additional guidance,
please refer to "A User's Guide To Environmental Immunochemical Analysis" (Gee et al., 1994).

EXTRACTION PROCEDURES

Methanol Shake Procedure

       Five stainless steel balls (cleaned with methanol) are placed into each 30 ml bottle.  Five
grams of sample is weighed into each bottle.  One scoopula tip (approx. two grams) of sodium sulfate
is added  to each bottle.  Five  mL  methanol is added with the repeater pipette. If the methanol is
largely soaked up by a highly absorptive sample, 10 mL methanol can be added and an additional
dilution factor is used.

       Clamp the bottles  into the mechanical shaker, and shake for 20 minutes at maximum
amplitude.  Remove bottle, and allow sample to settle.  Decant  the methanol into the  base container of
the filtration device, filling the container 2/3 full.  Insert the capped plunger into  the base and push
steadily until sufficient filtrate is obtained inside the hollow plunger.  Remove cap from the plunger
and decant filtrate into appropriately labeled borosilicate screw-top storage vial.

       Modification: Some samples, in particular clays, were difficult  to filter due to formation of
suspensions of very fine particles.   The following addition was  inserted immediately after removal of
bottles from the shaker:

       Decant approximately 10 mL of filtrate/suspension into  15-mL centrifuge tube.  Cap tube
tightly and centrifuge at 2100 g (3900 RPM on the IEC Model  Centra 8R) for 35 minutes at room
temperature.  Remove tubes from  centrifuge and decant methanol into the lower container of filtration
device. Proceed as above.

Supercritical Fluid Extraction

       Add 1 gram anhydrous magnesium sulfate to 4 grams sample  in weighing dish. Mix with
spatula until all clumps are broken up to obtain a free flowing solid. Transfer samples to extraction
vessel, placing two plugs of silanized glass wool over the vessel frits to prevent plugging.  Extract at
350 ATM and 100°C for 20 minutes in the dynamic mode. The flow rate of CO2 was maintained at
approximately 1.25 to 1.5  mL/min. using the capillary restrictor heated to 100°C.  Collect extract in
methanol at ambient temperature.   After extraction, adjust volume to  5 mL with methanol.

Soxhlet Extraction Procedure

       Follow the extraction procedures in EPA SW-846 Method 3540A (U.S. Environmental
Protection Agency, 1990a), substituting pesticide grade methanol for the hexane/acetone solvent
                                              13

-------
mixture.  Remove extract from Soxhlet apparatus. The extracts are ready for ELISA analysis. Do
NOT follow the remaining Method 3540A procedures for concentration, cleanup or solvent exchange.

SPIKING PROCEDURES

Method A

       Weigh 5 grams of sample into a 25 mL scintillation vial and slowly add pesticide grade
methanol until the soil is wetted with a thin film of methanol throughout the sample.  Drip the
methanol spiking solution containing either the radiolabeled tetrachlorinated biphenyl or Aroclor
standard onto the surface of the sample near the center of the container away from the glass walls.
Cap the vial and allow to sit overnight.  Remove the  cap and cover the top of the vial  with a kimwipe
to prevent contamination of the sample by airborne particulates.   Allow the samples to evaporate to
dryness at room temperature over several days.

Method B

       Weigh 5 grams of sample into a plastic weighing dish and spread out.  Slowly drip the spike
solutions onto the surface of the samples, taking  care to disperse the liquid as  widely as possible
without allowing the solution to come in contact with the dish. Allow the samples to evaporate to
dryness at room temperature.


SCREENING AND CHECKERBOARD TITRATION OF ANTI-SERA AND COATING ANTIGENS

       Several objectives were targeted during this phase of development.  First, anti-serum which
exhibited a high degree of selectivity toward PCBs was identified. Second, the optimal concentrations
of the two primary immunochemical reagents (antibody and coating antigen) were determined.
Optimal concentrations and activity were determined  utilizing a matrix dilution scheme (checkerboard
titration) as outlined in "A User's Guide to Environmental Immunochemical Analysis." (Gee, et al.,
1994)

Carbon-14 Extract Counting

       Methanol soil extracts were added to scintillation solution (400 uL extract/10 mL scintillation
solution) and counted using the auto DPM mode according to procedures outlined in the liquid
scintillation counter documentation (Beckman Instruments, Fullerton, CA), with a user specified error
value of 0.2%.
DEVELOPMENT OF ANTIBODIES FOR PCBs

       Immunoassay development requires the production of an antibody specific to the analyte of
interest, in this case PCBs.  Environmentally significant PCBs are derived from commercially available
preparations (Aroclors in the United States) which are mixtures theoretically containing up to 209
congeners.  The first consideration which must be addressed in PCB immunoassay development is that
of the identity of the immunogen.  As a result of trends in production, Aroclors 1242, 1248, 1254, and
                                              14

-------
1260 are most prevalent in the environment (Hutzinger et al., 1974) and consequently the current assay
was targeted to these Aroclors.

       PCBs themselves are relatively small immunologically speaking, and are incapable of eliciting
the immune response necessary for the generation of anti-PCB antibodies.  A general solution to this
problem involving small organic molecules entails the synthesis of an analogue molecule possessing a
relatively long side chain which itself contains a reactive moiety amenable to further chemical
manipulation.  This analogue molecule (termed the hapten) is then covalently linked to a large globular
protein via the reactive side  chain moiety, which by synthetic design, is at or near the end of the side
chain.  The end result is essentially a large globular molecule with the physico-chemical features
unique to the analyte of interest, and large enough to function as an immunogen, that is, to elicit an
immune response.

       It has been determined  that pentachloro biphenyls make up a major fraction of the Aroclors of
interest, in fact up to as high as 48% in Aroclor 1254 (Hutzinger et al.,  1974).  It is expected from
chemical principles that the  most abundant pentachloro biphenyls in Aroclors are those with one di-
and one tri-chlorophenyl ring making up the biphenyl.  Nuclear magnetic resonance (NMR) analysis of
chromatographically resolved congeners shows this to be the case (Hutzinger et al., 1974). A
commonly  observed pattern  for this pentachloro substitution is the 2,2',4,5,5'-pentachloro biphenyl
congener.  With this information in mind a hapten based on a 4-hydroxy analog (2,2',4',5,5'-
pentachloro-4-biphenylol)  was synthesized under EPA Contract 68-03-3511. (U.S.  Environmental
Protection  Agency, 1989). From this analog was prepared a derivative with a functionalized side
chain, as shown in Figure 1.

       The terminal carboxyl group was then covalently linked with free amino  groups on the protein
keyhole limpet hemocyanin, to form  an amide linked immunogen. Multiple rabbits were immunized
with this immunogen, and blood serum collected, resulting in the production of several hundred
milliliters of PCB antiserum. Further details can be found in the relevant documents for EPA Contract
68-03-3511.
                                                                            COOH
Figure 1. PCB ELISA immunization hapten.

                                                15

-------
SYNTHESIS OF COATING ANTIGEN

       The current PCB immunoassay is based on the "inhibition" ELISA format, as will be described
in the next section, and as such, utilizes a PCB analog, or hapten, linked covalently to a carrier
protein. The resulting hapten-protein conjugate is termed the coating antigen.  For the present assay,
4-(2,4,5-trichlorophenoxy)-butyric acid was chosen as the hapten for the coating antigen due to the
analogy with the distal ring of the immunization hapten described in the previous  section.

       The coating antigen was prepared according to the method of Schmidt (Schmidt, et al., 1990).
The hapten was conjugated with bovine serum albumin  (BSA) and chicken egg derived  conalbumin.
Several protein/hapten loading (defined as  number of hapten molecules/protein molecule) combinations
were synthesized.  For the coating antigen incorporated  into the quantitative PCB  plate ELISA, the
hapten loading was estimated spectrophotometrically at 280 nm, and found to be approximately 15.
The solution of this coating antigen contained 4.4 mg BSA/mL, based on preparative data.
ASSAY FORMAT

       The format employed in the current PCB ELISA is termed an "Inhibition" ELISA.  As is the
case with all immunoassays, this format relies on the binding of an antibody (Ab) which is specific for
a particular analyte (referred to  as the antigen in the generic sense).  Thus, in the current assay, the
antibody for PCBs (the "anti-PCB" antibody) binds specifically with  PCB molecules.  The details of
the indirect inhibition format are depicted in Figure 2 (modified from Gee, et al., 1994).

       In the first step, the coating antigen described in section 5 is dissolved in a buffer  of basic pH
and dispensed into the wells of  a polystyrene 96-well microplate. The  coating antigen binds with the
polystyrene surface by adsorptive forces.

       In step 2, samples containing PCBs or PCB calibration standards are added to tubes containing
a buffer solution of anti-PCB antibodies. The mixture is allowed to  come to equilibrium, at which
point the PCB molecules are bound by  the antibodies. The quantity  of unbound or "free"  antibody
remaining in solution is dependant  on the quantity of PCBs in the sample or standard.  Columns
designated "A", "B", "C", and "D"  in Figure 2 depict varying concentrations of PCBs.

       In step 3, the solution from each of the tubes is pipetted into wells on the antigen-coated
microplate prepared in step 1 (usually in replicate, for example, 3 wells/tube) and reaction is allowed
to proceed to equilibrium, at which point the remaining free anti-PCB antibody binds with the hapten
of the coating antigen. The  antibodies  which  bound with PCB in step  2 cannot bind with  the coating
antigen due  to the fact that the binding  sites of these antibodies are occupied by PCB  molecules.
These antibody-PCB bound complexes  in solution are then washed away, leaving only antibodies
which were  able to bind with the coating antigen.  Based on this relationship, it is apparent that the
quantity of antibodies bound to  the coating antigen is inversely proportional to the concentration of
PCBs in the sample or standard.
                                               16

-------
                                     B
           Stepl
           Step 2
           StepS
           Step 4
                        E  E  E
           Steps
                       o «o «o •   o •  o •
                       yyy    y   y
                        E  E  E     E 0  E
    Polystyrene well
    Coating Antigen
    Anti-PCB Antibody
    (polyclonal rabbit IgG)
 ?  Enzyme Linked to
 p^  Secondary Antibody
 i  PCB Molecule
of Substrate to
 
-------
       In step 4, a second antibody (termed the "secondary antibody") to which an enzyme has been
covalently linked is added to each well of the microplate as a solution in buffer. The salient property
of this antibody is that it binds with the species specific protein comprising the anti-PCB antibody. For
the PCB ELISA, the anti-PCB antibodies were raised in rabbit and are thus composed of rabbit
immunoglobulins (rabbit IgG). The secondary antibody, purchased commercially, was raised in goats
immunized with rabbit IgG, and is thus an anti-rabbit IgG antibody.  The binding reaction is allowed
to come to equilibrium, and then the solution is washed away.

       Finally, in step 5,  a solution of the substrate on which the enzyme acts is added to each well.
For the PCB ELISA, the enzyme is  alkaline phosphatase, and the substrate is p-nitrophenyl phosphate.
The enzyme converts the substrate to a yellow colored product; the more enzyme-antibody conjugate
present, the more intense the color development.  Due to the inverse relation between PCB
concentration and bound primary antibody mentioned in step 2, it also follows that the color formation
is inversely proportional to PCB concentration. The assay  is completed by measuring the optical
density in each well and then mathematically relating concentration of PCB  to optical density.  This
relation then allows for calculation of PCB concentration in the samples based upon their measured
optical density values.  The response of the ELISA to varying PCB concentrations is plotted in the
lower right corner in Figure 2.

       Several advantages are gained by using this format.  This format typically provides great
sensitivity to small changes in concentration, that is, the assay is able to easily discern relatively small
differences in  concentration between samples.  In addition, this is advantageous due to the fact that the
enzyme is never exposed to materials that might be present in the sample which might prevent or alter
the enzyme-mediated color producing reaction.
                                               18

-------
                                         SECTION 6

                                RESULTS AND DISCUSSION
DOSE RESPONSE STUDIES

       A characteristic of paramount importance in immunochemically based analytical systems is the
response  of the system as function of analyte concentration.  In the case of ELISAs, this response  is
physically characterized by measurement of optical density.  A typical  response function is shown in
Figure 3.  This response function can be mathematically transformed using a 4 parameter logistic fit of
the form:
                                  y = (A - D)/(l + (x/C)B) + D
Where:
       x = Analyte Concentration
       y = Optical Density
       A = Upper asymptote

This function is plotted in Figure 4.
B = Slope
C = Midpoint of Curve
D = Lower asymptote
       This curve takes on a characteristic sigmoidal shape. At "low" analyte concentrations, no
significant binding has occurred and the response of the system  is close to that of a zero concentration
standard or sample. As concentration increases, binding proportional to analyte concentration occurs,
and the optical density decreases, until finally at "high" analyte concentrations, all of the binding sites
on the antibodies are occupied and therefore the system can no longer respond.  Of particular
significance analytically is the C value; this represents the value at which the system  response is at
50%.  In immunochemical terminology, it is said that the response is inhibited by 50%, and the  C term
is thus termed the 50% inhibition level, or the I50.  The I50 value is a direct indicator of the general
concentration level at which the system is capable of functioning.

DOSE RESPONSE FOR COMMON AROCLORS

       The current system was  characterized for Aroclors 1016, 1221, 1232, 1242, 1248, 1254, 1260
and 1268 using the standard ELISA  protocol.  Dose response functions for these Aroclors appear in
Figures 5-12, and the I50 values  for each of these Aroclors are summarized in Table 1.  Aroclor  1268
showed little response, and hence is  not graphically represented.
              TABLE 1.  FIFTY PERCENT INHIBITION LEVEL FOR AROCLORS
Arolcor
1016
1221
1232
I50,ng/mL
71
585
77
Aroclor
1242
1248
1254
I50,ng/mL
25
22
10
Aroclor
1260
1262
1268
I50,ng/mL
15
31
>500
       Interestingly, Aroclor 1254 exhibits the greatest response (lowest I50)) of the Aroclors tested,
and correspondingly, it is composed of 48% pentachloro biphenyls, the largest percentage of any
Aroclor. (Hutzinger et al., 1974).
                                              19

-------
               cn
               c

               CD

              O



               O
               O

              -t—'
               Q_

              O
-  Q
                    0      100    200    300    400


                         Concentration  (arbitrary)


Figure 3.  Immunoassay dose response.
                                       500
           ^  2-°

            en
            c
            (D
           Q  1.5

            O
            y

           "Q-  1.0
           o
               0.5
b
                  0.1         1         10       100


                         Concentration  (arbitrary)



Figure 4.  ELISA dose response with four parameter fit.
                                       1000
                                    20

-------
                1.2 -
             c
             
-------
                0.2 L-
                          10           100         1000



                            Concentration  (ng/mL)




Figure 7.  Dose response for Aroclor 1232.
                                     10      100     1000



                           Concentration (ng/mL)




Figure 8. Dose response for Aroclor 1242.
                                   22

-------
             cn
             c
             CD
             Q
             D
             U
             CL
             O
                 1.8
                 1.6
                 1.4
                 1.2
                 1.0
                 0.8
                 0.6
                 0.4
                 0.2
                      0.1
                               1       10      100     1000
                            Concentration  (ng/mL)
Figure 9.  Dose response for Aroclor 1248.
             (D
             D
             O
             CL
             O
                 1.2
                 1.0
                 0.8
                 0.6
                 0.4
                 0.2
                 0.0
                   0.1        1        10       100      1000
                             Concentration (ng/mL)
Figure 10. Dose response for Aroclor 1254.
                                    23

-------
                      0,1       1       10      100     1000

                             Concentration  (ng/mL)

Figure 11.  Dose response for Aroclor 1260.
             cu
             o
             o
             -+->
             CL
             O
1.4


1.2


1.0


0.8


0.6


0.4


0.2
                      0.1      1       10     100    1000

                            Concentration  (ng/mL)
Figure 12.  Dose response for Aroclor 1262.
                                    24

-------
EXTRACTION SOLVENT EFFECT ON ASSAY DOSE RESPONSE

       Immunoassay is generally conducted in aqueous  solutions which have been buffered.
Typically, the immunochemical systems tolerate a high percentage of a number of organic solvents
such as dimethyl sulfoxide (DMSO), acetonitrile, 2-propanol, methanol and ethanol.  Other solvents,
such as hexane, may be tolerated by immunoassays, however in significantly reduced volume fractions.
Methanol was chosen as the solvent of choice, based in large part on previous extraction studies
employing  a range of solvents (Spittler, 1986).

       Aroclor 1248 standards, prepared in methanol, were run using the standard assay procedure,
varying the concentration of methanol composition in the assay solution from 5, 10, 15 and 20
percent.  Dose response curves fit by the 4 parameter logistic fit are illustrated in Figures 13-16.
Based  on the comparability of curves for the series of 5  to 15 percent methanol, 15 percent methanol
was chosen for routine use in the assay, as this will result in the lowest detection limit when sample
extracts are added to assay buffer.  The dose response for 20%  methanol was very erratic, as seen in
Figure 16.

CROSS-REACTIVITY STUDIES

       Many compounds possibly present in environmental samples may potentially bind, or cross-
react, with  the anti-PCB antibodies due to some structural similarity with PCBs. The degree of
binding will be reduced from that of PCB binding, but nevertheless any cross-reactivity will give rise
to a measured level of PCBs higher than that actually  present.  Of particular interest are 1,2,4-
trichlorobenzene arid 2,5-dichlorophenol. The trichlorobenzene is analogous to the distal phenyl ring
of the  immunization hapten, while  the dichlorophenol  is  analogous to the proximal dichlorophenoxy
moiety of the immunization  hapten. It is also possible that because the hapten is linked to the long
aliphatic  spacer arm via an phenyl-ether linkage, that chlorophenol ethers might be potentially more
cross-reactive than hydroxyl analogues.

       To  address this issue, serially diluted standard solutions of a number of potentially cross-
reacting compounds, based on structural similarity,  were prepared in methanol.  These solutions were
carried through the  standard assay procedure, enabling calculation of I50 values. These data are
summarized in Table 2. In the case when no cross-reactivity was observed at very high
concentrations, the I50 is reported as greater than the highest concentration of the solution assayed.

       The data show that none of the potentially cross  reacting compounds exhibits  greater than 3%
cross-reactivity relative to Aroclor  1248.  A 2.7% cross-reaction was observed for 2,4,5-trichloro
phenol which is not unexpected considering that the 2,4,5-trichloro  substitution pattern was present on
the immunizing hapten. This clearly demonstrates the role of hydrogen bonding in this particular
cross-reaction, because the corresponding chlorine substituted benzene (1,2,4-trichloro) exhibited
insignificant cross reactivity (<0.1% relative  to Aroclor 1248).

EXTRACTION EFFICIENCY STUDIES

       Extraction of PCBs from soil samples played a significant role in the overall performance  of
the present ELISA based analytical procedure. In the  current study, it was desired that laborious,
overnight extraction procedures be  avoided.  In order to  take maximum advantage of the high
throughput potential of the quantitative PCB  plate ELISA, it was determined that a relatively high

                                               25

-------
throughput extraction procedure should be employed, and thus the extraction procedure was derived
from known procedures used for field applications (U.S. Environmental Protection Agency 1992a,
1992b).

       Generally, these extraction methods entail placing the soil sample, drying agent and extraction
solvent in a suitable vessel, followed by hand shaking for several minutes.  The method derived from
these procedures and used in the current study is more vigorous, in that it employed a mechanical
shaker of much greater intensity than obtained by handshaking, and a longer time period.

       The question of determining extraction efficiency of a method as applied to real-world samples
is never directly answerable unless one has an infinite amount of time and resources to enable
preparation of spiked samples under the same real-world conditions as those  which produced  the
samples in the first place.  As a substitute for this impractical situation, a three pronged approach was
taken in the current study.

       The first part of the approach involved spiking of commercially obtained PCB standard
reference soils (SRMs) and real-world samples with known quantities of radiolabeled analyte, followed
by sample aging and then  subsequent extraction and counting of the extracted radioactivity.  Carbon-
14 ring labeled 2,2',5,5'-tetrachloro biphenyl was chosen as the spiking material.  In the case of PCBs,
this does not strictly emulate extraction of real world Aroclor contaminated soils, due to the fact that
                                              26

-------
             (D
             D
             O

             -4—'
             Q_

             O
    0.7


    0.6


    0.5


    0.4


    0.3


    0.2
                 0.1  L
                      1           10         100        1000

                            Concentration  (ng/mL)


Figure 13.  Dose response, 5% methanol.
    1.0 -



•~S  0.8 -



    0.6
             c
             (D
             O
             O

             'a.  0.4
             O


                 0.2
          1
                                 10        100        1000

                            Concentration  (ng/mL)


Figure 14.  Dose response, 10% methanol.
                                    27

-------
    1.6

    1.4

S   1.2

    1.0
               c
               N
              |  0.7


              I  0-6

              "5  0.5
               y

              "o.  0.4
              O

                  0.3


                  0.2
                                   A -  0.985

                                   B -  0.882

                                   C -  35

                                   D -  0.25
                             10          100        1000

                             Concentration  (ng/mL)
Figure 16. Dose response, 20% methanol.
                                   28

-------
TABLE 2. CROSS-REACTIVITY OF POSSIBLE CO-CONTAMINANTS
COMPOUND
Anisole, 2,3,4-trichloro
Anisole, 2,4,6-trichloro
Anisole, 2,6-dichloro
Anisole, 2-chloro
Anisole, 3,5-dichloro
Anisole, 4-chloro
Benzene, 1,2,3,4-tetrachloro
Benzene, 1,2,3-trichloro
Benzene, 1,2,4,5-tetrachloro
Benzene, 1,2,4-trichloro
Benzene, 1,2-dichloro
Benzene, 1,3,5-trichloro
Benzene, 1,3-dichloro
Benzene, 1,4-dichloro
Benzene, chloro
Biphenyl
Butyric acid, 4-(2,4,5-trichlorophenoxy)
DDE
DDT
Phenol, 2,3,4-trichloro
Phenol, 2,3,5,6-tetrachloro
Phenol, 2,3,5-trichloro
Phenol, 2,3,6-trichloro
Phenol, 2,3-dichloro
Phenol, 2,4,5-trichloro
Phenol, 2,4,6-trichloro
Phenol, 2,4-dichloro
Iso, ng/mL
>20000
>20000
>20000
>20000
>20000
>20000
>20000
>20000
8920
>20000
>20000
>20000
>20000
>20000
>20000
>20000
5300
>1400000
>205500
3940000
1280
>20000
>20000
29300
914
>20000
7990
                         29

-------
COMPOUND
Phenol, 2,5-dichloro
Phenol, 2,6-dichloro
Phenol, 2,6-dichloro
Phenol, 2-chloro
Phenol, 3,4-dichloro
Phenol, 3,5-dichloro
Phenol, 4-chloro
Phenol, pentachloro
I50, ng/mL
8740
81900
>20000
>20000
98000
1290
>20000
>20000
       Aroclors are of course mixtures of many congeners. Unfortunately, radiolabeled Aroclors are
not commercially available.  In order to use radiolabeled Aroclors, it would be necessary to chlorinate
radiolabeled biphenyl under conditions which essentially emulated the industrial preparation of the
Aroclors.  This was clearly undesirable in terms of the additional research effort needed. The
tetrachloro congener named above was chosen as the closest approximation from the limited list of
commercially available radiolabeled congeners because it has been determined that this congener
typically comprises up to 8.4% in Aroclor  1242 and 1248 (Albro et al., 1981).

       The  second approach by  which method extraction efficiency was  ascertained was by extraction
of commercially available PCBs  in soil standard reference materials (SRMs), followed by analysis of
the extracts by the quantitative PCB plate ELISA.  This approach may potentially be limited by  the
fact that the  data is confounded by ELISA  performance and by the fact that the soil matrix may  not
necessarily emulate the actual real-world samples.  In the current study, two Aroclor 1248 SRMs
representing 2 PCB levels were extracted.  In addition, five Aroclor 1242 SRMs representing 5 PCB
levels were extracted.

       The  third approach to determining extraction efficiency involved spiking of approximately 10
percent of the real world samples with known quantities  of the relevant Aroclor, followed by
extraction and subsequent analysis by the ELISA.

EXTRACTION OF C-14 RADIOLABELED TETRACHLORO BIPHENYL SPIKED SOILS

Samples

       Four separate sets of extraction experiments were conducted aimed at examining the effect of
PCB level on the extraction efficiency, the effect of changing the spike level,  and the effect of
different soil types on the extraction efficiency.  All spikings  were carried out employing spiking
Method A.
                                              30

-------
Extraction of USATHEMA Soil

       The first set of extractions utilized clean standard soil obtained from the United States Army
Toxic and Hazardous Environmental Material Agency (USATHEMA). A total of thirty 5-gram
samples were spiked; 10 of these samples were spiked with 58 pL of C-14 spiking solution to give a
total PCB concentration of about 0.1 mg/Kg.  Twenty 5-gram samples were spiked with 125  pL of C-
14 spiking solution and half were fortified with 32 pL of 1.1 X 10"5 g/mL of Aroclor 1248 standard
solution, to give a total PCB level of 1 mg/Kg. The other half were fortified with 454 pL of the 1248
standard solution to give a total PCB level of 10 mg/Kg. In addition, three unspiked blanks were
carried through the procedure.

       The samples were extracted with 5 mL methanol according to the described procedure.  The
spiking vials were also extracted with 5 mL methanol to determine if any losses of activity occurred
during the spiking procedure.  The extract was added to liquid scintillation cocktail (400 uL extract/10
mL cocktail), and the samples were counted.  In addition, calibration standards were prepared by
adding 125 pL of the C-14 spiking solution to 5, 7, 10, and 15 mL of methanol. The standards  were
added to scintillation cocktail (same  as above, in triplicate); these standards correspond to 100, 71, 50
and 33 percent recovery for the 1 and 10 mg/Kg samples and 216,  153, 108 and 72 percent recovery
for the 0.1 mg/Kg spikes.  The standards were counted along with the spiked samples, and a linear
least-squares fit  was used to calculate percent recovery.  The fit is shown in Figure 17. The  results are
presented in Table 3.
 o
 o
 CD
cr

 c
 CD
 
-------
          TABLE 3.  RECOVERY OF RADIOLABELED TETRACHLORO BIPHENYL
                                FROM USATHEMA SOIL
Total
PCB Level, mg/Kg
0.1
1.0
10
Mean Percent
Recovery of C-14
84.7
89.1
89.9
RSD,
%
0.2
1.5
0.6
       No significant activity was counted in the spike vials, thus the calculated recoveries represent
actual recoveries of material from the soil.  This data demonstrates that the C-14 labeled tetrachloro
biphenyl can be recovered with high efficiency in the presence of up to  10 mg/Kg Aroclors.

Extraction of Commercial PCB Reference Soils

       The second set of C-14 extractions was carried out on seven commercially obtained PCB in
soil SRMs (Environmental Research Associates, Arvada, CO). Five levels of Aroclor 1242 were
represented, along with two levels of Aroclor 1248. The samples were weighed out in duplicate, and
were prepared, extracted and analyzed using the described procedures. The calibration curve for this
extraction is shown in Figure 18.  The results are presented in Table 4.
              CD
              >
              O
              (J
              CD
              CD
              U
              L_
              CD
              D_
                    00
80
60
                    40  L
         y = 2.94 + 0.012x

         r2 = 0.995
                          4000    5000    6000    7000

                                   Decay Events/Minute

Figure 18. Reference Soil C-14 extraction calibration curve.
                                              8000
                                           32

-------
          TABLE 4.  RECOVERY OF RADIOLABELED TETRACHLORO BIPHENYL
                                     FROM SOIL SRMS
Aroclor, SRM Concentration (mg/Kg)
1248, 33.9
1248, 33.9
1248, 282
1248, 282
1242, 0.5
1242, 0.5
1242, 1.5
1242, 1.5
1242, 8.0
1242, 8.0
1242, 25.0
1242, 25.0
1242, 45.0
1242, 45.0
Percent C-14 Recovery
87.6
94.5
87.0
96.0
84.0
90.8
88.5
86.0
94.0
89.0
91.0
90.0
88.0
88.0
       The mean extraction efficiency of the C-14 radiolabeled PCB in the presence of widely
varying PCB levels and two types of Aroclors is 89.4%, with a RSD of 3.7%.  Again, no significant
activity remained on the spiking vessel walls, thus the calculated recoveries represent about an 11 %
residual non-recovery in the soil samples.

Extraction of C-14 Spiked Real-world Samples

       The third set of extractions was carried out on approximately 10%  of one of the sets of real-
world samples included in the present report.  These samples were collected at  Kansas City, MO, and
have known analytical values for Aroclor 1248 contamination levels, as to  be discussed later.  Spiking,
extraction, and analysis were carried out in duplicate, identical to procedures described above, with the
exception that the samples were dried and powdered prior to spiking due to the difficulties encountered
in spiking firm clay samples.  The results for the extractions are given in Table 5.
                                             33

-------
               TABLE 5. RECOVERY OF RADIOLABELED TETRACHLORO
                 BIPHENYL FROM KANSAS CITY SOIL/CLAY SAMPLES
Sample
ID
KC043D
KC043D
KC046
KC046
KC049
KC049
KC061
KC061
KC068
KC068
KC081
KC81B
KC083D
KC083D
KC085
KC085
KC088
KC088
Percent
C-14 Recovery
88.7
94.0
87.9
88.4
87.4
87.4
90.8
97.5
90.7
96.9
89.8
92.7
89.2
94.6
98.9
98.1
89.0
93.7
Aroclor 1248
level, mg/Kg
1.74
1.74
<0.033
<0.033
<0.033
<0.033
580
580
0.504
0.504
0.687
0.687
0.413
0.413
428
428
2.70
2.70
       The mean recovery was found to be 92.0%, with a RSD of 4.1%. The results indicate that the
extraction procedure can remove with a greater than 90% efficiency a major congener of the Aroclors
of interest in the current study, and further, can do so regardless of level of Aroclor 1248 present.

ELISA PERFORMANCE CHARACTERISTICS

       The quantitative PCB plate ELISA performance was characterized in terms of detection limit,
assay accuracy, assay precision and the practical  quantitation range.
                                           34

-------
Assay Detection Limit

       The assay detection limit is defined as being three standard deviations above the zero standard.
Optical density data for twelve separate determinations spanning an approximately 5.5 month period
were used in the calculation.  The standard deviation for the zero standard (triplicate wells) was
multiplied by three and subtracted from the optical density of the zero standard (recall the inverse
relation between concentration and optical density). The resulting optical density value was substituted
into the respective four-parameter fit for each dose response curve.

       The resultant concentration values of Aroclor 1248  corresponding to the assay detection limit
are summarized in Table  6:
                  TABLE 6.  ELISA DETECTION LIMIT FOR AROCLOR 1248
                              Calculated Detection Limit, ng/mL
                                             0.742
                                             0.336
                                              1.56
                                             0.608
                                             2.80
                                             0.767
                                             2.34
                                              1.55
                                              1.40
                                             2.90
                                             0.646
                                             0.527
                             Mean Detection Limit ± 1 Standard
                             Deviation, ng/mL =  1.34 ± 0.87
       Similar data were acquired for Aroclor 1242 using identical procedures.  These data are
summarized in Table 7.
                                              35

-------
                  TABLE 7.  ELISA DETECTION LIMIT FOR AROCLOR 1242
                                        Calculated Assay
                                     Detection Limit, ng/mL
                                              1.62
                                              1.56
                                             0.608
                                              1.83
                                             0.776
                                              2.33
                                              2.32
                                  Mean Detection Limit ±  1
                                  Standard Deviation =
                                  1.57 ng/mL ± 0.62 ng/mL
       It is important to note that these are not detection limits for soil samples, but rather for
concentrations in the assay solution.  When converted to corresponding soil levels, these values are
corrected for the dilution factor of 6.67 which occurs when adding 15% soil extract to 85% buffer.
Hence the limit of detection  for 1248, in soil, is 8.95 ng/g ± 5.8 ng/g.  The calculated detection limit
for Aroclor 1242 in soil is 10.5 ng/g ± 4.1 ng/mL.  The calculated standard deviations are fairly large
fractions of the calculated detection limits, which in large part is due to the fact that calculations of
concentrations within in the  region near the zero standard are inherently of low mathematical accuracy
when the slope of the four parameter fit approaches zero.

Standard Curve Characteristics

       The practical quantitation range for the ELISA corresponds to the linear region of the
sigmoidal curve where the optical density begins to fall off rapidly as PCB concentration increases.
This region can be conveniently fit mathematically  by a linear equation of the form:
                                OD = A + B*log(Concentration)
Where:
      OD  = measured optical density
        A  = intercept value
        B  = slope of the line
        C  = Concentration of PCBs in ng/mL

     A typical calibration curve is shown in Figure 19.  Data points are the mean of triplicate
measurements with error bars of + 2 standard deviations.
                                              36

-------
                      0.9

                      0.8
                      0.6

                 "5    0.5
                 o
                 "a.   0.4
                 O
                      0.3
y  =  1.015  -  0.337x

 2  =  0.993
                      0.2  L-
                             0.5          1.0         1.5         2.0

                                        log(concentration)

Figure 19.  Optical density versus log Aroclor 1248 concentration.

     As can be seen, the correlation coefficient, r2, is greater than 0.99; the upper and lower limits for
quantitation can be empirically reached by including standards which typically fall in the high and low
regions of the four parameter curve where the response begins to roll off. Calibration curves with r2
values less than 0.99 were never used for calculation of Aroclor concentration in samples.  Rather, the
points at the end of the curve were discarded, resulting in appropriate inclusion of the linear region.

     The current assay has been employed in the range of about 3 ng/mL to 200 ng/mL for Aroclors
1248 and 1242 in assay solution, corresponding to 20 to  1333 ng/g in soil. These concentrations
represented the maximum assay working range, as slight roll-off was occasionally observed for these
boundary concentrations as a result of inter-assay variation. Typically,  inclusion of these points (if
rolling off) would lower the  correlation coefficient into the 0.97 region, in which case these boundary
points were discarded.

Long Term Assay Precision

     To address the question of assay stability and reproducibility, standard reference materials (SRMs)
were assayed during each analytical run.  For Aroclor 1248, two commercially obtained soil SRMs
(Environmental Resource Associates, Arvada, CO) were used, with certified Aroclor 1248 levels of
33.9 and 282 mg/Kg of soil.  For Aroclor 1242, five commercially available  soil SRMs
(Environmental Resource Associates, Arvada, CO) were used, with certified Aroclor 1242 levels of
0.5,  1.5, 8, 25 and  100 mg/Kg of soil.

     In all cases, the soil SRMs were extracted employing the standard  extraction procedure as
described, and the extracts stored in borosilicate glass vials sealed with  teflon lined caps.  The extracts
were diluted with methanol to bring the PCB levels into the assay range.
                                              37

-------
Long Term Assay Precision for Aroclor 1248--
     The Aroclor 1248 SRMs were extracted in duplicate and are designated as "A" and "B."  These
extracts were assayed over a 6 month period, and thus, the resulting data represent inter-assay
variability  for a six month period.  No apparent trends, either upward or downward, were noted as a
result of storage artifacts. The post-extraction dilution for the Aroclor 1248 SRMs corresponded to a
dilution factor of 334 (before being added to assay) for the low level SRM and a dilution factor of
2263 (before addition to assay) for the high level 1248 SRM. Results for the Aroclor 1248 SRMs are
summarized in Table 8.

     The optical densities for the diluted Aroclor  1248 SRMs corresponded to optical densities in the
range corresponding to the region between the 38 and 50 ng/mL calibration standards, thus these
sample dilutions fell into a mathematically optimal region, i.e., the central part of the curve.  The
reported error does not include an  estimate of extraction or dilution error.

Long Term Assay Precision for Aroclor 1242-
     The five Aroclor 1242 SRMs were extracted on two separate occasions and analyzed  by ELISA.
It was observed that the results for the first set of extracts did not differ from those of the  second set
of extract.  Due to the large number of wells needed to assay two sets of five standards (three
wells/sample x five x two sets = 30), it was decided that one set/analysis was the practical limit which
still  enabled analysis of a reasonable quantity of samples on a plate.  The stored extracts were
analyzed over a 3 month period, to give Aroclor 1242 SRM data representing 3 month variation data,
as summarized in Table 9.

     The measured  levels of Aroclor 1242 for all  soil  SRMs  are on the low side, corresponding to
about 53-91% extraction efficiency, assuming  100% accuracy for the ELISA.  According to SW-846
Method 8080, a measured recovery of  39-150% for solution phase Aroclor 1242 measurements meets
quality control  criteria for acceptable results. The quoted Method 8080 recovery criteria does not take
into  account the additional error incurred as a  result of soil extraction, thus, it can be concluded that
the ELISA results represent acceptable performance, at least based on comparability to gas
chromatographic reference data.

     Based on data which will be discussed in the sections addressing spike recovery and matrix
effects, it can be concluded that the ELISA itself is accurate, and thus it can be stated that the low
results are  most likely due to low extraction efficiency for the Aroclor 1242 SRMs.

EFFECT OF SAMPLE MATRIX ON ASSAY PERFORMANCE

     It is imperative to determine what effect,  if any, non-analyte related sample characteristics had on
assay performance.   In the case of immunoassay,  such factors as ionic strength, metal and  other
inorganic content, humic acid content, etc., may result in an alteration of the assay performance.  The
effect of matrix upon assay performance, when observed, is of a non-specific nature.  By this, it is
meant that  the matrix effect is not  due  to binding  events with the antibody itself, as is the case with
analyte or a cross-reacting compound, but rather with  such ill-characterized effects such as alteration of
antibody conformation.  This latter event could alter the binding constant for the antibody-analyte
interaction.

     If any such effects  were to occur,  there would be a loss of correspondence between the dose
response for a set of "clean" standard solutions, and a set of "matrix effect containing" samples. As  a

                                               38

-------
result, the accuracy and possibly precision of the analytical results could differ greatly from the
characterized accuracy and precision of the assay system.

     An effective method by which to detect such matrix effects is to carry out a series  of dilutions on
the samples, assay these sample diluates, and then determine mathematically whether the results are
related.  Such a scheme is termed a determination of parallelism.  If the  analytical results for the
diluates fall within the working range  of the PCB assay, two conclusions can be drawn  about the
measured PCB values.

     First, the quantity of PCB measured for all the diluates from  a given sample should be identical
within the bounds of method accuracy after mathematical correction for dilution. Second, it is clear
that with respect to PCB  concentration, the serial dilutions for the samples are essentially equivalent to
the serial dilutions used to prepare the calibration standards, and as such, the sample diluates ought to
produce a dose response or calibration curve mathematically identical to that of the standard curve
obtained during the same analytical run.  If this is not the case, it  becomes apparent that other matrix
related effects are skewing the assay response.

     Importantly, it should be noted that such a scheme cannot determine whether cross-reacting
compounds are contributing to the measured apparent PCB concentration,  because the cross-
reactivity is an equilibrium binding event just as is the PCB-antibody binding event.  Essentially,
the two events differ only in the magnitude of their respective binding constants.

Parallelism Determination Results

     Parallelism studies were carried out on each of the three sets of real world  samples.  Serial
dilutions were performed by adding 1 mL of the methanolic sample extract to 1 mL of  pesticide grade
methanol.  Following mixing, 1 mL of this diluate was added to another mL of pesticide grade
methanol and then mixed. This procedure was repeated until dilutions which gave results within the
working range of the  assay were obtained. The resulting diluates  had dilution factors of 2n, where n is
any integer between 1 and up to 8. The results are summarized in two ways. The results  for all
diluates of a given sample falling within the assay working range  were averaged, and a RSD
calculated. In addition, the data are plotted superimposed onto the assay dose response curves  for the
standard solutions.
                                               39

-------
TABLE 8. ELISA RESULTS FOR AROCLOR 1248 SOIL SRM EXTRACTS
High A (mg/Kg)
258
275
296
284
261
299
291
242
219
313
295
316
263
239
302
245
264
247
273

Expected: 282
Mean = 273
Percent Error = -3.2
SD = 26.6
RSD = 9.8%
High B (mg/Kg)
238
247
288
273
244
287
287
248
218
291
286
285
245
231
284
237
256
236
274

Expected: 282
Mean = 261
Percent Error = -7.4
SD - 23.4
RSD = 8.8%
Low A (mg/Kg)
37.8
39.6
33.5
34.6
40.6
37.9
39.2
33.0
26.6
39.3
36.7
44.4
39.8
33.7
36.2
34.6
38.4
33.2
37.2

Expected: 33.9
Mean = 36.6
Percent Error = 7.9
SD = 3.76
RSD = 10.3%
Low B (mg/Kg)
35.0
35.3
	
31.8
33.5
33.0
33.6
29.2
26.9
39.6
37.4
39.9
33.4
31.2
32.3
32.0
37.2
30.4
38.6

Expected: 33.9
Mean = 33.9
Percent Error = 0
SD = 3.49
RSD = 10.2%
                          40

-------
TABLE 9. ELISA RESULTS FOR AROCLOR 1242 SOIL SRMS
ERA 0.5 mg/Kg
0.258
0.283
0.179
0.205
0.230
0.309

0.361

0.357

0.301

0.179

0.283
0.226
0.237
0.329
0.226

0.254
0.266

Expected:
0.500
Mean = 0.264

Percent Error =
-47
SD = 0.054
RSD = 20.0%
ERA 1.5 mg/Kg
0.940
0.879
0.718
0.801
0.820
0.904

0.929

0.980

0.779

0.756

0.879
1.261
0.988
0.911
0.751

0.946
0.916

Expected:
1.50
Mean = 0.892

Percent Error =
-40.5
SD = 0.122
RSD = 13.7%
ERA 8 mg/Kg
5.93
5.44
4.64
5.07
4.96
4.53

670

7.67

3.67

497

5.44
7.00
5.96
6.67
441

6.23
5.72

Expected:
8.00
Mean = 5.22

Percent Error =
-35
SD= 1.65
RSD = 31.0%
ERA 25 mg/Kg
17.7
16.7
14.8
16.8
15.9
16 2

16 6

19 1

17 7

16 9

16.7
18.8
18.3
18.7


17.5
17.1

Expected:
25.0
Mean - 17 221

Percent Error =
-31
SD= 1.11
RSD = 6.5%
ERA 100 mg/Kg
99.6
91.3
94.3
91.4
87.3










91.3
88.9
94.1
83.3


96.7
87.6

Expected:
100
Mean — Q1 444

Percent Error=
-8.6
SD = 4.41
RSD = 4.8%
                     41

-------
Parallelism Results for Aroclor 1248 From the Kansas City Samples-
     Approximately ten percent of the Kansas City SITE samples were non-randomly selected based
on having concentration ranges between 818 and 2498 ug/Kg, as previously determined by ELISA.
These samples were selected because the concentrations were such that several of the serial dilutions
starting directly from the undiluted sample extracts would have concentrations falling within the assay
working range. If higher concentration samples had been chosen, the number of serial dilutions
needed to bring the concentrations into the assay working range would be so high such that the matrix
effect, if any,  would have already been diluted out.  Initial concentration below 800 ug/Kg were not
suitable because, once diluted several times, the concentrations would quickly fall off the lower end of
the assay working range.

     The results for the  calculated concentrations are summarized in Table 10.

                   TABLE 10.  ELISA RESULTS FOR SERIALLY DILUTED
                               KANSAS CITY SOIL EXTRACTS
1
Sample
KC17
KC22D
KC39
KC42
KC53
KC56
KC60D
KC68
KC83
Mean PCB
Concentration (ug/Kg)
2798
406
968
1514
1707
924
1590
1252
777
RSD (%)
4.4B
13.8C
14.0C
0.9B
11. 7C
16A
11.9"
7.1C
8.8C
Dilution Range
Measured
22..24
2\.24
2'..24
22..24
2'..24
2\22
21..23
2>..24
21..24_
A: n = 2, B: n = 3, C: n = 4

    The average RSD for all samples (excluding KC56, for which only two points fell within the
assay working range) is 9.0%. This is within the precision performance of the assay for repeated
samples at a fixed  concentration.  It can be concluded that, based on the above results, no significant
matrix effect was observed in the Kansas City samples.  The data for the Kansas City diluates are
plotted in Figures 20-22.  Each plot represents data collected during one analytical run. The data for
samples is plotted along with the calibration  standards used in the corresponding run.

    The assay response for serially diluted sample extracts is essentially identical to assay response
for calibration standards, thus it can be concluded that no matrix effects were present.
                                              42

-------
                                          O  Standards
                                          v  KC17
                                          n  KC22D
                                          A  KC39
                                          O  KC42
                           10     20     30     40     50

                            Concentration (ng/mL)

Figure 20. Dose response for serially diluted Kansas City samples, group 1.
             >N
             -4—'
             cn
             c
             CD
             O

             O
             0

             Q_
             O
O  Standards
v  KC53
D  KC56
A  KC60D
                          20     40     60     80    100

                            Concentration (ng/mL)

Figure 21. Dose response for serially diluted Kansas City samples, group 2.
                                  43

-------
                                                  O  Standards
                                                  v  KC68
                                                  n  KC83
                               20      40     60      80     100

                                Concentration  (ng/mL)

Figure 22.  Dose response for serially diluted Kansas City samples, group 3.

Parallelism Results for Aroclor 1242 From the Allied Paper/Portage Creek/Kalamazoo River Samples-
    Eleven samples from the Allied Paper/Portage Creek/Kalamazoo River Superfund site were
serially diluted and subsequently the diluates were analyzed by ELISA as described.  The results are
summarized in Table 11.

     TABLE 11.  ELISA RESULTS FOR SERIALLY DILUTED ALLIED PAPER SAMPLES
Sample
19A
24B
20B
27B
28C
30C
37C
31C
34C
25C
38C
Mean Aroclor 1242
Concentration,
(mg/Kg)
98.3
130
74.4
112
157
143
186
113
172
132
135
RSD (%)
9.4B
8.4C
5.9s
0.4B
3.4B
3.7B
8.4C
7.9C
6.2C
6.8A
4.9C
Dilution Range
22-24
22-25
23-25
23-25
22-25
22-25
22-25
22-25
22-25
23-24
22-25
A: n = 2, B: n = 3, C: n = 4
                                         44

-------
    The data are plotted along with the corresponding calibration standard data in Figures 23-26.  The
dose response for the sample diluates is essentially identical to the dose response for calibration
standards.

    In addition, it can be seen that the RSDs for each of the parallel analyses are not distinguishable
from analytical  variation.  Hence it can be concluded that no significant matrix effects enter into the
response of the  PCB ELISA, for the samples obtained at the Allied Paper/Portage Creek/Kalamazoo
River Superfund site.
           en
          o
1.4


1.2


1.0
           Q_
          0   0.6
               0.4
                   0
                                                 O   Standards
                                                 v   28C
                                                 n   30C
                                                 A   37C
                                                 O   31C
                             50       100      150      200

                             Concentration  (ng/mL)

Figure 23. Dose response for serially diluted Allied Paper/Portage Creek samples, group 1.
                                          45

-------

>>
en
c
(U
Q

D
U
CL
0


1
1

0


0

0
0


0

8


6

4
2
0
\
b n
\A A
\-\?Cx
C^:^
"^^^^
^^^^
-
i i

0 50 100 1
Standards
19A
25C
38C



'Sv-,,
-^S^^,

i

50






^Zz-~~^ s — v
^e-v
i

200
                              Concentration  (ng/mL)

Figure 24. Dose response for serially diluted Allied Paper/Portage Creek samples, group 2.
              en
              CD
                  2.5
                  2.0
              D   -i
              o   ' -
              Q_
              O
                  1.0
O  Standards
v  23A
n  20B
A  27B
                      0        50       100      150     200

                              Concentration  (ng/mL)

Figure 25. Dose response for serially diluted Allied Paper/Portage Creek samples, group 3.
                                      46

-------
                    1.6
              :£   1.4
               (n
               c
               (U
              Q   1.2
               a
              .y   1.0
               CL
              O
                   0.8

                   0.6
                       0
                     O  Standards
                     V  24B
                     D  34C
 50        100       150       200

Concentration  (ng/nnL)
Figure 26.  Dose response for serially diluted Allied Paper/Portage Creek samples, group 4.

Parallelism Results for NEIC Samples-
    The NEIC samples were extracted and analyzed as described. During this analysis, only two data
points fell on the calibration curve for each sample, and thus, the variation between duplicates is
expressed as the relative percent difference (RPD). The data are summarized in Table 12.  With the
exception of sample 03-008-88-02-S, it can be seen that error for the diluates is essentially within the
bounds of assay error, hence it can be concluded that no matrix effects were observed for the NEIC
samples.

Spike Recovery for Real-World Samples

    Approximately ten percent of the real-world samples were spiked, extracted and analyzed with the
quantitative PCB plate ELISA, along with the corresponding non-spiked samples.  An equivalent
amount  of the spiking solution itself was pipetted into volumes of methanol equal to the extraction
volume.  These latter solutions correspond to 100% extraction recovery, and were analyzed directly in
the same ELISA as the  sample spikes. In addition, the "100% recovery" solutions were taken through
the entire extraction procedure to determine potential loss during the extraction manipulations.

    Nine samples taken from the Kansas City SITE (to be discussed later) study were spiked with
Aroclor 1248 using spiking procedure B. Five samples from the Allied Paper/Kalamazoo River/
Portage  Creek Superfund site study were spiked with Aroclor 1242 using spiking procedure A.
                                             47

-------
          TABLE 12. ELISA RESULTS FOR SERIALLY DILUTED NEIC EXTRACTS
Sample
03-002-88-02-S
03-003-88-02-S
03-004-88-02-S
03-005-88-02-S
03-006-88-02-S
03-007-88-02-S
03-008-88-02-S
03-009-88-02-S
ELISA Result
(mg/Kg)
62.6
107
81
95.4
110
160
77.3
55.9
RPD
1.8
5.6
12.9
9.4
1.8
15.6
28.4
0.8
Results for Kansas City Soil Spikes--
    The Kansas City samples were spiked with 250 uL of 1.1 x 10"4 g/mL standard solution of
Aroclor 1248, which corresponds to a level of 5.5 mg/Kg in the soil.  Three of the nine samples were
aged approximately 1.5 months prior to extraction, the remainder overnight. The resulting extracts
were serially diluted in methanol to bring them into the assay range. The measured concentrations for
all serial dilutions within the assay range (typically  1:2 to 1:32) were averaged to arrive at a final
concentration value.  The expected value was calculated as the sum of expected spike value plus the
measured value of the un-spiked sample.  The results are summarized in Tables  13-15. Each Table
represents  a separate analytical standardization.  The samples identified with asterisks were aged
overnight.

                 TABLE 13.  SPIKE RECOVERY OF AROCLOR 1248 FROM
                          KANSAS CITY SOIL SAMPLES, SET ONE
Sample
Spike Solution,
Extracted
Spike Solution
KC13
KC22
KC13 Spike
KC22 Spike
Mean Aroclor Result
(ug/Kg), RSD
5722, 9.5%
6167, 10.8%
937
365
6860, 8.4%
5598, 18.0%
Expected Value
(pg/Kg)
5500
5500
NA
NA
6437
5864
Percent Spike
Recovery
104.0
112.0
NA
NA
106.5
95.5
                                            48

-------
TABLE 14. SPIKE RECOVERY OF AROCLOR 1248 FROM
      KANSAS CITY SOIL SAMPLES, SET TWO
Sample
Spike Solution
KC31
KC43D
KC81
KC31 Spike
KC43D Spike*
KC81 Spike*
Mean Aroclor Result
(ug/Kg), RSD
5894, 9.7%
324
7587
931
6931, 5.5%
18704, 11.3%
5788, 10.8%
Expected Value
(ug/Kg)
5500
NA
NA
NA
5824
13087
6431
Percent Spike
Recovery
107.2
NA
NA
NA
119.0
143.0
90.0
TABLE 15. SPIKE RECOVERY OF AROCLOR 1248 FROM
     KANSAS CITY SOIL SAMPLES, SET THREE
Sample
Spike Solution
KC83D
KC86D
KC90D
KC83D Spike*
KC86D Spike
KC90D Spike
Mean Aroclor Result
(ug/Kg), RSD
4803, 12.6%
612
2053
1957
6199, 5.1%
7380, 8.5%
6059, 11.8%
Expected Value
(pg/Kg)
5500
NA
NA
NA
6112
7553
7457
Percent Spike
Recovery
87.3
NA
NA
NA
101.4
97.7
81.3
                    49

-------
Results for Allied Paper/Portage Creek/Kalamazoo River Soil Spikes-
    The Kalamazoo samples were spiked with 500 pL of a 5 x 10"5 g/mL standard solution of
Aroclor 1242, to give a PCB level of 5000 pg/Kg (for clean soils). The soils were
aged two days wet and 4 days dry.  The results after analysis by ELISA are summarized in Table 16.

                   TABLE 16. AROCLOR 1242 SPIKE RECOVERY  FROM
                        ALLIED PAPER/PORTAGE CREEK SAMPLES
Sample
Spike Solution
5A
12A
4B
13B
38C
5A Spike
12A Spike
4B Spike
13B Spike
38C Spike
Mean Aroclor Result
(pg/Kg),
4732, 2.7%
464
7659
960
9480
3182
6685
13166
7309
14997
6630
Expected Value
(ug/Kg)
5000
NA
NA
NA
NA
NA
5464
12659
5960
14480
8182
Percent Spike
Recovery
94.6
NA
NA
NA
NA
NA
122.3
104.0
122.6
103.6
81.0
Summary of Spike Results for Real-World Samples-
    For Aroclor 1248, spiked into real world clay samples at the 5.5 mg/Kg level, and subsequently
analyzed by ELISA, it was found that the mean recovery was 104.3%, with a RSD of 17%.  For
Aroclor 1242, spiked into real-world soil, waste paper, and sediment samples at the 5 mg/Kg level,
and subsequently analyzed by ELISA, it was found that the mean recovery was 106.7%, with a RSD
of 14%.

    Of course, recovery for extracts as determined by ELISA is confounded by the precision and
accuracy of the ELISA itself.  Based on recoveries determined for all spike solution analyses (n =  5),
it was found that the mean recovery or accuracy of the ELISA at this Aroclor concentration was
101.0%, with a RSD of 8.9%. The accuracy levels attained in both Aroclor 1248 and Aroclor 1242
extractions  are nearly identical to the accuracy of the spike solution recovery, thus it can be inferred
that the RSD for the extract analyses is additive.  The RSD for the extraction itself, then, is roughly <
8%, a result which is within a factor of two of that obtained for extraction of the C-14 spikes.

    The results obtained during the spike recovery studies demonstrate that the assay  produced
accurate determinations of spiked soil extracts. Further, the results demonstrate good extraction
performance, at least for fortified real-world samples. Finally, accurate spike recovery is  an additional
demonstration of freedom from matrix effects for the real-world samples.
                                             50

-------
RESULTS FOR DUPLICATE ANALYSES
    A statistically large subset of sample extracts was subjected to independent re-analysis in order to
obtain a measure of reproducibility of the quantitative PCB plate ELISA .  The population of data
actually consists of three sub-populations:  1) Kansas City SITE samples re-assayed after greater than
3 months between analysis 2) Kansas City SITE samples re-assayed after several days and 3) Allied
Paper/Portage Creek/Kalamazoo River samples re-assayed about 2 months later.  The data are
summarized in Table 17.

                 TABLE 17.  ELISA RESULTS FOR DUPLICATE ANALYSES
Sample
KC71*
KC31
KC96*
KC57*
KC97*
KC87*
KC87D*
5A
KC81
KC83
4B
KC56
KC83D
KC39
KC22D
KC68
KC60D
KC53
KC13
KC22
KC42
KC90D
KC86D
Result 1 (mg/Kg)
0.057
0.148
0.155
0.171
0.173
0.199
0.235
0.601
0.707
0.818
0.965
0.989
1.02
1.11
1.15
1.20
1.37
1.63
1.64
1.90
2.14
2.23
2.36
Result 2 (mg/Kg)
0.044
0.324
0.055
0.112
0.141
0.115
0.136
0.464
0.930
0.777
0.960
0.924
0.612
0.968
0.406
1.25
1.59
1.71
0.936
0.364
1.51
1.96
2.053
                                           51

-------
Sample
KC17
KC98*
KC106*
KC55*
12A
KC43D
KC112*
KC103*
Result 1 (mg/Kg)
2.5
2.89
3.56
4.77
6.82
7.18
112
120
Result 2 (mg/Kg)
2.80
2.644
4.97
5.34
7.66
7.59
88.0
94.0
*Time between re-run < 4 days.

    The duplicates are plotted against each other in Figure 27, excluding the two highest level
samples, which statistically skew the regression line.  The slope approaches unity, with a correlation
coefficient of 0.95, indicating a high degree of reproducibility between duplicate analyses, most of
which were separated in time by several months.

ELISA ANALYSIS OF REAL-WORLD SAMPLES

    The final stage of the current project  entailed the analysis of three sets of samples obtained from
EPA SITE demonstrations, EMSL-LV Technical Support Center demonstrations and regulatory
activities. During the design stage of the project, it was decided that three  sets of samples of widely
varying matrix composition should be employed in order to demonstrate  wide applicability of the
quantitative PCB plate ELISA based analytical procedure.  The  samples were collected as splits and
extracted and analyzed according to the procedures outlined in the present report. The analytical
results obtained by ELISA were compared to analytical data collected for the same samples using
standard reference methods, as outlined below.
                               y =  -0.25  +  1.117x
                                                  468

                                Results  For Analysis  1  (ng/g)

Figure 27.  Duplicate ELISA analyses of Kansas City samples.
                                              52

-------
Analysis of Kansas City Site Samples

     The first set of samples to be considered were collected in conjunction with an EPA Super-fund
Innovative Technology Evaluation (SITE) of field methods for PCBs. The location chosen was a U.S.
Department of Energy (DOE) facility located about 20 miles south of downtown Kansas City,
Missouri. The facility has been used since 1949 for the manufacture of various non-nuclear
components destined for use in nuclear weaponry systems. The study site is located in a former
channel of Indian creek, and contains a former storm water outfall that discharged from the DOE
facility into the creek.  During the early 1970's, Indian Creek was rerouted as part of a flood control
project, and the former channel was covered with about 10 feet of fill.

     In early 1988 sampling was conducted and PCBs were detected in the soil around the site.  The
contamination is believed to have occurred in the 1960's and  1970's as the result of occasional spills
of a Therminol heat-transfer unit that subsequently drained into the outfall. In 1989, the DOE
conducted a Resource Conservation and Recovery Act (RCRA) facility investigation (RFI) and
corrective measures  study (CMS),  (U.S. Department of Energy, 1989), and as a result, the PCB
concentrations are well mapped  spatially at the site.  The soil is contaminated with a wide range of
PCB concentrations, from non-detect below 0.16 ug/Kg up to 9,680 mg/Kg.

     Sediments overlying the bedrock consist of soft, dark brown to  gray, homogeneous, medium to
high plasticity, moist silty clay with traces of fine sand.  This material varies in depth from 7 to 15
feet, and appears to have low  permeability.

Description of Site Design Factors Relevant to the Plate ELISA--
     The PCB SITE demonstration was designed as a comparative evaluation of four field screening
technologies, all referenced to a common, accepted analytical method for PCB determination.  Samples
were collected at a site where PCB contamination, predominantly by Aroclors  1248 and 1242, had
been well characterized in terms of concentration, allowing for collection of samples most suitable for
challenging the efficacy of the technologies.  Samples were collected and split five ways, one sample
for each of the field technologies and one for the confirmatory method.

Sampling-
     The basis for the experimental design was the control of with-in sample homogeneity.  Samples
on the order of 2 to  3 Kg were collected, physically homogenized and split among the four field
screening technologies and the confirmatory method.

     A sampling plan was designed such that coverage of a full range of concentrations was obtained,
with an emphasis on the lower concentrations, near 10 mg/Kg. The plan called for the collection of
the following samples: 20 samples from areas containing more than  1000 mg/Kg PCBs, 20 samples
containing between 100 to 1000 mg/Kg, and 60 samples from areas containing less  than 100 mg/Kg.

     Prior to the start of the SITE demonstration, samples were collected from selected locations and
sent to all involved technology developers as well as to EPA, EMSL-Las Vegas. Duplicate analyses
were carried out using the field screening methods utilizing samples  taken from the same sample
container. Many of these duplicate analyses gave widely varying results, which, after checking for and
finding no analytical error, led to the suspicion that homogeneity could be a potentially serious
problem, particularly due to the fact that many of the samples were the consistency  of modeling clay.
Drying of the sample followed by thorough mixing of the resultant powder as typically carried out in
classical laboratory methods was deemed to be unsuitable, as the method would no longer emulate
                                              53

-------
procedures carried out in field screening settings.  Due to the fact that inter-method comparison was
pivotal to the SITE study design, further measures were needed.

    Due to the concern with the effects of possible sample inhomogeneity, it was decided that field
duplicates would be randomly generated at a rate of about 20 to 25%.  These field duplicates were
taken as splits of the initially collected sample, and are therefore essentially (or should be) equivalent
to the splits taken for the CLP analysis and ELISA analysis.  The field duplicates were subjected to
analysis by the quantitative PCB plate ELISA as well as by the  CLP method.  Analysis of such a
relatively large number of duplicates by the reference method was specified in order to allow a
statistically meaningful measure of the success in sample homogenization.

Reference Method--
    Trie method used for obtaining reference data was the EPA Contract Laboratory Program (CLP)
method outlined in the CLP Statement of Work (SOW) for Organics Analysis (U.S. Environmental
Protection Agency, 1990a). This method is a gas chromatography method employing an electron
capture detector. PCB extraction is carried out via Soxhlet extraction using a hexane/acetone mixture.
The current study is intended to be a  comparative study between the plate ELISA results and those
obtained  by the accepted standard method.  As such, the CLP method results are taken to be the "true"
values.

Results for Kansas City Site Samples-
    Of the 142 samples collected during the SITE demonstration, 110 passed QA/QC requirements
during CLP analysis.  Samples from the set of 110 samples were selected at random  and analyzed over
a period of six weeks using the  ELISA method as outlined previously. The method was calibrated
with Aroclor  1248 and at least two soil SRM extracts were analyzed concurrently (typically, four soil
extracts were run concurrently).  The  results for the ELISA analysis are tabulated in Table 18, along
with the corresponding results by the  CLP lab.  CLP method detection limit was reported as 0.033
mg/Kg. Error for the ELISA  represents ± 2 standard deviations, as derived from the standard
deviation for  1248 SRM analysis.

    The error for the CLP method (which is a derivative of SW-846 Method 8080/81) is harder to
estimate.  According to SW-846 Method 8080, the standard deviation,  s, for determination of Aroclor
1248 in solution by a single analyst is -0.17 X, where X is the  average analytical result in ug/L (U.S.
EPA, Office of Solid Waste and Emergency Response,  1986).  The analytical error (± 2 standard
deviations) for the CLP method results will thus be  closely approximated as ± 34% of the reported
value.  It is important to note that these measures of variation are for the measurement stage of the
method, not the entire method inclusive of the extractions.

          TABLE 18.  ELISA AND CLP RESULTS FOR KANSAS CITY SOIL SAMPLES
Sample
ID
KC46D
KC46
KC49
KC48
KC45
ELISA RESULT
(mg/Kg)
0.224
0.098
0.165
0.108
0.031
ELISA ERROR
(mg/Kg)
0.067
0.043
0.022
0.018
0.002
CLP RESULT
(mg/Kg)
<0.033
<0.033
<0.033
<0.033
<0.033
CLP ERROR
(mg/Kg)
0.011
0.011
0.011
0.011
0.011
                                              54

-------
Sample
ID
KC35
KC99
KC41
KC35D
KC92D
KC87D
KC71D
KC82
KC77
KC70
KC57
KC51
KC69D
KC69
KC109D
KC109
KC12
KC110
KC111
KC24
KC27
KC21
KC87
KC67
KC03
KC14
KC105
KC28
KC82D
ELISA RESULT
(mg/Kg)
0.020
0.048
0.022
0.026
0.284
0.235
0.076
0.029
0.080
0.055
0.171
0.253
0.134
0.159
0.054
0.263
0.093
0.025
0.042
0.071
0.052
0.156
0.199
0.120
0.288
0.234
0.034
0.113
0.063
ELISA ERROR
(mg/Kg)
0.005
0.008
0.009
0.002
0.010
0.036
0.012
0.024
0.012
0.014
0.008
0.062
0.084
0.014
0.012
0.070
0.025
0.054
0.034
0.006
0.028
0.022
0.044
0.028
0.042
0.018
0.006
0.037
0.030
CLP RESULT
(mg/Kg)
<0.033
<0.033
<0.033
<0.033
<0.033
<0.033
<0.033
<0.033
<0.033
<0.033
<0.033
<0.033
<0.033
<0.033
<0.033
<0.033
<0.033
<0.033
<0.033
0.055
0.057
0.063
0.076
0.081
0.114
0.180
0.210
0.216
0.244
CLP ERROR
(mg/Kg)
0.011
0.011
0.011
0.011
0.011
0.011
0.011
0.011
0.011
0.011
0.011
0.011
0.011
0.011
0.011
0.011
0.011
0.011
0.011
0.019
0.019
0.021
0.026
0.028
0.039
0.061
0.071
0.073
0.083
55

-------
Sample
ID
KC31
KC93
KC83D
KC81D
KC83
KC56
KC42
KC22
KC07
KC60D
KC01
KC39
KC06
KC58
KC81
KC22D
KC98D
KC53
KC90
KC84D
KC13
KC30
KC84
KC98
KC92
KC86D
KC05
KC90D
KC86
ELISA RESULT
(mg/Kg)
0.148
0.124
1.020
0.669
0.818
0.989
2.14
1.90
3.31
1.37
4.41
1.11
2.89
2.50
0.707
1.15
2.09
1.63
2.28
5.89
1.64
0.479
5.90
2.89
1.69
2.36
10.1
2.23
2.68
ELISA ERROR
(mg/Kg)
0.024
0.004
0.104
0.044
0.048
0.112
0.416
0.122
0.540
0.366
0.160
0.128
0.266
0.224
0.014
0.042
0.104
0.112
0.162
0.294
0.136
0.059
0.119
0.342
0.290
0.184
0.502
0.098
0.142
CLP RESULT
(mg/Kg)
0.263
0.295
0.413
0.450
0.484
0.505
0.517
0.535
0.552
0.577
0.593
0.676
0.679
0.681
0.687
0.718
0.825
0.958
1.01
1.08
1.13
1.15
1.16
1.17
1.21
1.25
1.37
1.40
1.42
CLP ERROR
(mg/Kg)
0.089
0.100
0.140
0.153
0.165
0.172
0.176
0.183
0.189
0.196
0.202
0.230
0.231
0.232
0.234
0.244
0.281
0.326
0.343
0.367
0.384
0.391
0.394
0.398
0.411
0.425
0.467
0.476
0.483
56

-------
Sample
ID
KC02
KC43D
KC88D
KC102D
KC26
KC66
KC08
KC78
KC62
KC55
KC106
KC17
KC88
KC65
KC50
KC80
KC52
KC40
KC50D
KC19
KC104
KC59
KC15
KC25
KC74
KC73
KC95
KC23
KC75
ELISA RESULT
(mg/Kg)
7.90
7.18
4.26
6.82
1.68
4.45
24.8
5.90
2.54
4.77
3.56
2.50
3.02
33.7
13.6
7.85
2.72
13.9
12.6
8.49
28.1
11.2
6.23
54.8
14.5
33.8
47.2
18.7
14.6
ELISA ERROR
(mg/Kg)
0.130
0.198
0.310
0.334
0.342
0.466
1.578
0.078
0.674
0.152
1.77
0.213
0.170
15.2
2.37
0.148
1.06
4.84
1.19
0.577
5.26
1.95
0.836
5.66
2.18
11.3
11.2
3.24
1.60
CLP RESULT
(mg/Kg)
1.50
1.74
1.77
1.77
1.96
1.98
2.00
2.27
2.35
2.40
2.50
2.55
2.70
3.08
3.60
3.77
4.21
4.25
4.41
6.70
7.66
7.86
9.13
11.7
13.3
15.8
17.5
20.8
23.0
CLP ERROR
(mg/Kg)
0.510
0.592
0.602
0.602
0.666
0.673
0.680
0.772
0.799
0.816
0.850
0.867
0.918
1.05
1.22
1.28
1.43
1.44
1.50
2.28
2.60
2.67
3.10
3.98
4.52
5.37
5.95
7.07
7.82
57

-------
Sample
ID
KC34
KC79
KC89
KC18
KC76
KC32
KC100D
KC100
KC102
KC112
KC103
KC85
KC85D
KC61
KC36
KC91
KC91D
KC16
ELISA RESULT
(mg/Kg)
20.0
68.9
25.7
127
84.8
15.7
638
616
6.93
119
120
39.0
415
570
2310
935
1160
607
ELISA ERROR
(mg/Kg)
5.70
5.60
1.45
30.2
3.75
3.50
127
51.7
0.984
67.6
9.32
3.12
131
12.3
109
152
264
74.4
CLP RESULT
(mg/Kg)
34.0
42.8
45.0
45.2
46.7
47.6
167
177
293
315
403
428
465
580
816
1630
1704
2110
CLP ERROR
(mg/Kg)
11.6
14.6
15.3
15.4
15.9
16.2
56.8
60.2
99.6
107.1
137
146
158
197
277
554
579
717
     Another, possibly more informative, representation of the results is given in Figures 28-31. Each
grouping of two bars represents the ELISA result and the CLP result for a given sample.  The data are
grouped by concentration range to allow full axis expansion.  The non-detects (< 0.033 mg/Kg) are
depicted as 0.033 mg/Kg for convenience only.

     The most striking trend in the results is that, generally speaking, the PCB levels, as measured by
quantitative  PCB plate ELISA, are higher than the corresponding value obtained by CLP analysis.
Based on earlier discussions, it can be concluded that the ELISA itself is accurate.  For Aroclor spiked
directly into extraction solvent, it was seen that determinative accuracy was very high.  The spike
recovery values also show good accuracy and precision. Further, performance of the method as
applied to the two  Aroclor 1248  soil SRMs shows that accuracy of the total method is high.
                                               58

-------
              en
              en
                  0.3
              c  0.2
              o
              D
              CD
              O
              c
              o
              o
                  0.1
                  0.0
                              ELISA

                              CLP
                      0      5      10     15    20    25


                                        Sample


Figure 28. ELISA and CLP results for Kansas City samples, group 1.
              ^  7
               en

              \  6

              ^  5

               c  4
               o  ^
              CD
              o
              o
                  0
                               ELISA

                               CLP
                      . , , ^ j 1 1
                    0
                           5      10     15     20     25     30

                                       Sample


Figure 29. ELISA and CLP results  for Kansas City samples, group 2.
                                       59

-------
                  70
              en
             *   60


              ^  50


              c   40
              o

             "o   ^


             "c   20
              CD
              U
              c   10 h
              o

             0    o
                               ELISA

                               CLP
                                             JUL
                     0      5       10     15     20


                                        Sample


Figure 30. ELISA and CLP results for Kansas City samples, group 3.
                                                           25
en
N- 2500
~\
en
E 2000
o 1500
0
-b 1000
c
CD
c 500
o
(J
n
| | ELISA
i^H CLP
^^^^
-
-




_
_ _^^.

j
I









TT
Nil









^











































                       0        5        10       15


                                         Sample


Figure 31. ELISA and CLP results for Kansas City samples, group 4.
                                                           20
                                       60

-------
    The data sets can be compared using a paired t-test.  Careful consideration must be given to the
proper application of this statistic as applied to a data set such as the results for the Kansas City
samples, where the numerical values of the data points extend over almost 5 orders of magnitude.  It is
necessary to use a modified form of the t-test:
Where

          d   =   is the mean of relative sample differences = £ ((ResultELISA-ResultCLp)/ResultCLp)
          5   =   is the population variance mean difference (0 if the data sets are equivalent)
          s2   =   is the estimated variance of the differences
                  ResultELISA = ELISA result for sample i)
                  ResultcLp   = CLP result for sample i)
          n   =   number of samples

     Assuming the CLP result to be the "true" value, division of the difference between the sample
pairs by the CLP result converts the difference into a relative difference.  If the two data sets are
equivalent, the relative mean difference will be zero, analogous to the case where the mean difference
should be zero as well.

     If the mean differences are not converted to relative differences, the magnitude of the mean
differences will be larger than the magnitude of most of the results  themselves, due to the sample
distribution.  The differences for the lower concentration samples will become statistically insignificant
relative to the differences for the high concentration samples, and hence, the t statistic will be skewed
by the few high level samples. Carrying out the modified t-test calculation, we get:

                                 d  =  1.5469
                                 s2=   2.2502
                                 n  =  110
                                 t  =  1.08

for 109 degrees of freedom, p « 0.005, thus, statistically, the two data sets are not equivalent. The
mean relative difference  shows that the ELISA results are on the average greater than the CLP results
by a factor of about 2.5.

     In addition to the ELISA analysis of methanol shake extracts described above, a subset consisting
of 23 samples was randomly selected, extracted using supercritical fluid extraction, and then analyzed
by ELISA.  An intriguing picture begins to emerge when one considers the ELISA results for
methanol shake extracts  as compared to both the ELISA results for supercritical fluid extracts and CLP
results. The data are summarized in Table  19, and the  SFE-ELISA data are plotted against the
corresponding CLP data in Figure 32.
                                                61

-------
TABLE 19. ELISA, SFE-ELISA AND CLP RESULTS FOR
         KANSAS CITY SOIL SAMPLES
Sample
KC12
KC99
KC41
KC27
KC03
KC14
KC31
KC83
KC56
KC39
KC13
KC30
KC05
KC90D
KC86
KC26
KC17
KC88
KC40
KC33
KC19
KC25
KC23
MeOH Shake
ELISA
(ing/Kg)
0.093
0.048
0.022
0.052
0.276
0.234
0.148
0.818
0.989
1.11
1.14
0.479
16.9
2.23
2.68
1.68
2.50
3.02
13.9
11.2
8.49
59.9
18.7
SFE-ELISA
(mg/Kg)
0.039
0.013
0.035
0.028
0.536
0.166
0.387
0.363
0.304
0.436
1.25
0.105
12.0
1.52
1.61
0.638
1.21
1.58
8.40
5.40
4.49
11.6
9.45
SFE-ELISA
ERROR
(mg/Kg)
0.002
0.001
0.003
0.002
0.015
0.011
0.024
0.009
0.013
0.014
0.029
0.006
0.675
0.274
0.198
0.074
0.072
0.102
0.277
0.149
0.254
1.36
0.794
CLP
(mg/Kg)
<0.033
<0.033
<0.033
0.057
0.114
0.180
0.263
0.484
0.505
0.676
1.13
1.15
1.37
1.40
1.42
1.96
2.55
2.70
4.25
6.00
6.70
11.7
20.8
                     62

-------

en
"^
CH

C
O
D
C
CD
0
C
O
o

/ w
60


50
40

30
20


10

0

| | ELISA
F=^ SFE-ELISA
1 '
- |H CLP
...

--
...


—

«_ *- fc, n, -T

0 5 10













:int lh nllrflff












1
L ll M
(II





















15 20 2
                                               Sample

Figure 32.  SFE-ELISA and CLP results for re-extracted Kansas City Samples.

     In general, the SFE-ELISA gave results which are lower than the results for ELISA analysis of
samples extracted with methanol.  The SFE-ELISA results, in general, are convergent with the CLP
results.

     A paired t-test was carried out on the SFE-ELISA data and the CLP data with the null hypothesis
that the data sets are equivalent.  Carrying out the calculation with this hypothesis gives:

                       t = 0.8729
                       p = 0.39

     The two data sets are thus equivalent by the t-test. In contrast, if the methanol shake
extraction/ELISA data and the CLP data for the same samples are compared using the paired t-test, a
t-value of 2.118 is  obtained, with a corresponding probability of only 0.046.  The methanol
shake/ELISA results and the CLP results are not equivalent to the 0.05 level. The SFE-ELISA data
results are equivalent to the CLP results, with a mean RPD between SFE-ELISA results and CLP
results of -14%.  The mean RPD between methanol shake/ELISA and CLP results, in comparison, is
35%.

     Given the fact that the ELISA performance itself is well characterized statistically (RSD < 10%
for replicate inter-assay analyses, based on Aroclor 1248 SRM data), it is clear that the difference
between the two ELISA results is due to the extraction procedure.  Of course, sample homogeneity
will play a  role, but it could be expected to be random with regard to PCB level.  Clearly, this is not
the case.

     As discussed earlier, extraction efficiency using methanol is expected to be greater than that for
hexane, or hexane/acetone mixtures (Spittler, 1986).  The CLP method specifies acetone/hexane, a
solvent system which apparently  was optimized for the suite  of analytes covered by the method (non-
volatile organochlorines and pesticides).
                                              63

-------
     Based on this knowledge, it can be hypothesized that the results for the ELISA analyses and the
CLP analyses differ, in large part, due to the different extraction procedures employed.  For example,
the lower extraction efficiency of the hexane/acetone mixture could come into play for that subset of
the Kansas City samples which are more "difficult" to extract than the others.  The ideal case for
comparison of ELISA data to GC data would entail co-analysis of extract splits generated  by whatever
particular extraction method was employed. Unfortunately, this was beyond the scope of the current
study.

Analysis of Allied Paper/Portage Creek/Kalamazoo River Superfund Site Samples

Background--
     The samples analyzed with the quantitative PCB plate ELISA in the current study were collected
as part of an EPA Technical Support demonstration project referred to as the "Field Screening
Technologies Analysis of Polychlorinated Biphenyls At The Allied Paper/Portage Creek/Kalamazoo
River Superfund Site,  Kalamazoo, Michigan."  The quantitative PCB plate ELISA was not an integral
component of the demonstration project, rather, the participants in the project generously agreed to
collect sample splits which were intended for use by investigators working in the Immunochemistry
Program at EMSL-LV.

     The Allied Paper/Portage Creek/Kalamazoo River Superfund Site is located in the southeast
corner of Michigan's lower peninsula.  The site covers 35 miles of the Kalamazoo river, from the
confluence with Portage Creek, and approximately 3 miles of Portage creek. PCBs are the main
contaminant of concern at this site, and according to the Michigan Department of Natural  Resources
(MDNR), are due mainly to carbonless copy paper recycling during the 1950's through the 1970's.

     This site has been extensively characterized over the past several decades, and the soil types have
been characterized as ranging from poorly to well drained soils that have loamy to sandy and loamy
subsoils formed in glacial outwash (Blasland and Bouck, 1991).  During pre-demonstration activities, it
was determined that Aroclor 1242 was the main Aroclor of interest in the context of the field
demonstration.  Three types of sample matrices were collected: soil, sediment and paper waste.  For
further detail, refer to  the Demonstration and Quality Assurance Project Plan for  the  "Field Screening
Technologies Analysis of Polychlorinated Biphenyls At The Allied Paper/Portage Creek/Kalamazoo
River Superfund Site,  Kalamazoo, Michigan,"  available through the Technology Support Center,
Environmental Monitoring Systems Laboratory, Las Vegas, NV (U.S. EPA, 1992a).

Analysis of Samples by Quantitative PCB Plate ELISA--
     Thirty-eight samples were extracted and analyzed by quantitative PCB plate ELISA according to
procedures described.  Each analytical run included analysis of the five commercially obtained soil
SRMs as described in Section 4.  Separate splits of all the samples were analyzed by  the MDNR using
SW-846 Method 8081 according to quality control/quality assurance measures specified by that
method (U.S. EPA, Office of Solid Waste and Emergency Response, 1992c).  The samples are
designated with a letter prefix A, B, or C and  a number.  The A corresponds to river  sediment,  B to
soil,  and C to paper waste and the numbers, to a specific sampling location.  These locations are
arbitrary in terms of the present study.

     The results are summarized in Table 20:
                                              64

-------
    TABLE 20. ELISA AND SW-846 METHOD 8081 RESULTS FOR
ALLIED PAPER/PORTAGE CREEK/KALAMAZOO RIVER SOIL SAMPLES
Sample ID
IB
2B
3B
4B
5A
6B
7B
8A
9A
10B
11B
12A
13B
14A
15A
16A
17A
18B
19A
20B
21B
22C
23A
24B
25C
26C
27B
ELISA Result
(mg/Kg)
1.00
0.204
1.58
1.81
0.552
2.73
2.41
2.90
7.93
3.23
11.0
6.82
9.48
8.85
12.5
17.4
10.8
67.2
26.0
48.0
37.8
28.5
23.3
146
47.7
91.4
56.3
ELISA Error
(mg/Kg)
0.201
0.041
0.316
0.362
0.110
0.546
0.482
0.580
1.59
0.645
2.20
1.36
1.90
1.77
2.51
3.49
2.16
13.4
5.20
9.61
7.55
5.69
4.65
29.2
9.53
18.3
11.3
Method 8081
Result (mg/Kg)
<0.5
<0.5
0.540
0.600
0.630
1.02
2.60
2.75
3.60
6.00
8.90
11.0
11.6
12.0
17.9
21
34.0
74.0
81.0
88.0
95.0
106
107
136
139
140
160
Method 8081
Error (mg/Kg)
0.170
0.170
0.184
0.204
0.214
0.347
0.884
0.935
1.22
2.04
3.03
3.74
3.94
4.08
6.09
7.14
11.6
25.2
27.5
29.9
32.3
36.0
36.4
46.2
47.3
47.6
54.4
                          65

-------
Sample ID
28C
29C
30C
31C
32C
33C
34C
35C
36C
37C
38C
ELISA Result
(mg/Kg)
44.6
240
39.9
41.2
120
46.0
122
37.1
60.6
100
78.7
ELISA Error
(mg/Kg)
8.91
48.0
7.98
8.23
24.0
9.19
24.4
7.42
12.1
20.0
15.7
Method 8081
Result (mg/Kg)
170
180
180
186
190
204
240
242
260
267
268
Method 8081
Error (mg/Kg)
57.8
61.2
61.2
63.2
64.6
69.4
81.6
82.3
88.4
90.8
91.1
     The MDNR laboratory detection limit was reported as 0.5 mg/Kg. Again, method error for
ELISA is taken as 2 standard deviations derived from soil SRM analysis.  Method 8081 error is 34%,
as discussed earlier.  The data are summarized graphically in Figures 33 and 34.  The non-detects (<
0.5 mg/Kg) are depicted as 0.5 mg/Kg for convenience only.

     Figure 33 shows that for the "lower level" samples (< 30 mg/Kg), there is close agreement
between the PCB plate ELISA results and the Method 8081 results;  most  of the ELISA and Method
8081 results overlap within the error limits of the respective methods. Using the modified paired t-test
with the null hypothesis that the data sets are equivalent gives:

     t= 1.53
     p =  0.14

The two  sets of "low level"  results are thus equivalent by the t-test at the  0.14 level.

     In contrast, the  results for the "higher level" samples (>30 mg/Kg) show a different trend.
Generally, the PCB ELISA results are significantly lower than the Method 8081 results, by up to a
factor of 6.5, as can  be seen in Figure 34.  Performing the paired t-test on these data gives:

     t = -8.46
     p « 0.001

     It was reasoned that since the problem lies mainly with the high level samples,  there may be a
non-equilibrium partitioning with the shake extraction procedure.  Contrasted with Soxhlet extraction,
which essentially provides multi-stage extraction, the shake procedure provides only  a one-stage
equilibrium partitioning between the solid sample and solution.  In the case of the high level samples,
the quantity of non-extracted PCBs heldup on the solid sample matrix could be significant.
                                               66

-------
en
*: 30
^
c 20
o
o
-t— '
c 10
CD
O
C
0
0
n

I | ELISA
HI Method 8081
_

T

T
...luit.itilli


I


I



I

'i




-r
[i











                           1   2  3   4  5  6  7  8  9  10  1 1 12 13 14 15 16


                                            Sample



Figure 33. ELISA and SW-846 Method 8081 results for Allied Paper/Portage Creek/Kalamazoo River

samples, group 1.
               en
               en

               £
               c
               o
               c
               0)
               
-------
     In an effort to verify this hypothesis, the high level samples were re-extracted using the Soxhlet
 procedure specified in SW-846 Method 3540A (U.S. EPA, 1990b) with the exception that methanol,
 rather than hexane/acetone, was used.  After overnight extraction, the methanolic extracts were serially
 diluted  and analyzed by the ELISA. The data are summarized in Table 21.

         TABLE 21.  ELISA RESULTS FOR SOXHLET EXTRACTS AND METHOD 8081
               RESULTS FOR ALLIED PAPER/PORTAGE CREEK/KALAMAZOO
                                    RIVER SOIL SAMPLES
Sample
19A
20B
22C
23A
25C
27B
28C
30C
31C
33C
36C
38C
37C
ELISA Result
(mg/Kg)
94.5
66.5
104
66.5
132
112
157
143
113
131
196
173
186
ELISA Error
(mg/Kg)
18.9
13.3
20.9
13.3
26.4
22.5
31.5
28.6
22.6
26.3
39.2
34.6
37.2
Method 8081
Result (mg/Kg)
81.0
88.0
106
107
139
160
170
180
186
204
260
268
267
Method 8081
Error (mg/Kg)
27.5
29.9
36.0
36.4
47.3
54.4
57.8
61.2
63.2
69.4
88.4
91.1
90.8
     The data are represented graphically in Figure 35.

     Each group of three bars represent ELISA results for the methanol shake extraction, ELISA
results for the methanolic Soxhlet extracts and the Method 8081 results.  Two points are evident.
First, all ELISA results for the Soxhlet extracts fall within 2 standard deviations of error for the
Method 8081 results.  Analytically, the Soxhlet/ELISA and the Method 8081 results appear to be
equivalent. Second, the differences in ELISA results (shake versus Soxhlet) are a consequence of
extraction procedure only, as the two sets of results overlap within the error limits of the respective
methods.  The only difference in the extraction was vigor, rather than solvent or sample clean-up etc.,
and thus the difference in ELISA results is due almost solely to extraction method.  These experiments
demonstrate that, coupled to an appropriate sample preparation procedure, the ELISA is capable of
providing highly accurate analytical data.
                                             68

-------
                 cr>
                o
                c
                o
                o
350
300

250
200
150
100

50
n
[ | Shake Extraction/ELISA
Y//A Soxhlet Extraction/ELISA
•• Method 8081
-
.- 1
-\\]
', i
- ' 7- '
i (• i;

I

[:
i

7
f





1


7
1 '
J


I;
\
y

I

'
f:




7
(




1







1



y



]

1


-r








                              1   2   3   4   5   6   7   8   9   10111213
                                                Sample

Figure 35.  ELISA and SW-846 Method 8081 results for Soxhlet extracts of high level samples.

     This conclusion that the Soxhlet/ELISA and Method 8081 data sets are equivalent was tested by
carrying out the modified paired t-test on the data.  Carrying out the calculation gives:

     d  = -0.21
     s2 = 0.029
     t = -4.59
     p < 0.005

     The hypothesis that the data sets are statistically equivalent must be rejected on the basis of the t-
test. The failure of the test in this case is due in large part to the small variation seen in the relative
difference values. The ELISA results are biased low, on the average, by a factor of about 0.83.

     From a practical standpoint, a 17% bias is of little importance; inter-lab bias of such a magnitude
is commonly observed even when using the same method. In the present case of comparing
quantitative PCB plate ELISA data to Method 8081 data, not only is an inter-method comparison
being made, but a confounding inter-lab comparison is being made as well.  It can thus be concluded
that for real-world use, the quantitative PCB plate ELISA provided accurate results when coupled with
a rigorous extraction procedure.

     The results for the Allied Paper/Portage Creek/Kalamazoo River Superfund site samples
demonstrate the potential dangers of abbreviated extraction procedures.  The 20 minute mechanical
rocker shaking employed in the current work is more vigorous than many of the rapid hand shake
extraction procedures common to the commercial ELISA based screening methods, yet the 20 minute
mechanical shake extraction procedure was shown to have an  extraction efficiency far lower than an
overnight Soxhlet extraction using the same solvent. Perhaps  many of the "false negative" results
observed in field screening applications are due to poor extraction performance, rather than poor
ELISA performance.  It appears that the tough question of extraction validation may need to be more
adequately addressed.
                                              69

-------
Analysis of EPA NEIC Samples

Background-
     Samples employed during this phase of the current study were obtained from the EPA National
Enforcement and Investigation  Center (NEIC) in Denver, Colorado. The sample splits were taken
from archival samples which had already been subjected to analysis by NEIC chemists using the
method described in "PCB Analytical Program Standard Operating Procedure" (Hill et al., 1989).  This
GC/ECD method is a modification of a previously published method (Bellar et al., 1982). Based on
reference data provided by NEIC, single analyst standard deviation was calculated to be 18.5%, thus
method error was taken to be ± 37%.  Nine samples characterized as being  contaminated with a
mixture of Aroclors 1242/1254/1260 were obtained.  The sample matrices ranged from soil to oily
sludge.

Analytical Procedure--
     The samples were extracted  according to the described procedures.  Extracts were serially diluted
in 1 mL of methanol, typically up to 1:256 dilution.  Aroclor 1242 was chosen as the calibrator.
Samples were analyzed in triplicate wells according to the described procedures.

Results-
     The results for ELISA and NEIC GC analysis are summarized in Table 22.

             TABLE 22. ELISA AND NEIC RESULTS FOR NEIC SOIL SAMPLES
Sample
01-012-92-03-S
03-002-88-02- S
03-003-88-02- S
03-003-93-02- S
03-004-88-02- S
03-005-88-02- S
03-006-88-02-S
03-007-88-02- S
03-008-88-02-S
03-009-88-02-S
ELISA
Result
(mg/Kg)
172
62.6
107
770
81
95.4
110
160
77.3
55.9
ELISA
Error
(mg/Kg)
7.83
3.63
3.52
17.3
5.10
19.1
3.45
9.00
9.10
3.61
NEIC
Results
(mg/Kg)
46.0
37.0
61.0
307-841
46.0
62.0
63.0
235
101
43.0
NEIC
Error
(mg/Kg)
17.0
13.7
22.6
-
17.0
22.9
23.3
87.0
37.4
15.9
    The data are represented graphically in Figure 36.  The data point for sample 03-003-93-02 has
been removed because of the large uncertainty of the NEIC result.  On average, it can be seen that the
ELISA values tend to run higher than the corresponding NEIC values. This is most likely not a
matrix effect, as variation for sample diluates is low, and furthermore, the samples are not necessarily
spatially related, and indeed, obvious diffences were observed between samples.  Several samples were
oily clays, whereas several of the other samples were moist sediments or soils.
                                             70

-------
                 en
                 o
                 L_
                -4-J
                 c
                 0)
                 o
                 c
                 o
                o
250

200

150

100

 50  |-

   0
ELISA
NEIC
                               I
         4567
                                               Sample
                                                                   8
Figure 36.  ELISA and NEIC results for NEIC samples.

     A number of factors could come into play, the most important probably being the fact that, as
reported by NEIC, a portion of the Aroclor mixture is Aroclor 1254 and 1260.  As reported in Table
1, the ELISA is approximately 2 times as sensitive to Aroclor 1254 and about 1.6 times as sensitive to
Aroclor 1260 relative to Aroclor 1242, thus the ELISA response for the NEIC samples is  somewhere
between a factor of 1 to 2, dependant upon the fractions of 1254 and  1260 present.

     A paired t-test with the null hypothesis again being that two data sets are statistically equivalent
gives:

     d =0.6640
     o2  = 0.6948
     t =2.3894
     p <0.05

     The null hypothesis must thus be rejected; the ELISA results are, on the  average, biased high by
a factor of about 1.7, which is consistent with, but does not confirm the statements made in the
proceeding  paragraph.  Calibration of the quantitative PCB plate ELISA with  Aroclor 1254 would shift
the curve such that measured Aroclor levels would  be one half that reported using Aroclor 1242
calibration.  The ELISA results would thus be biased only slightly low from the NEIC GC results.

     These  results are illustrative of a common problem in PCB calibration, whether for GC or ELISA
analysis, namely, that of determining the most suitable Aroclor calibrant.  The Aroclor chosen for
calibration  must emulate the actual composition of the samples. This is true  whether the
determinative method is by ELISA or GC.
                                             71

-------
Summary of Results for Real-World Samples

    Two large sets of unrelated environmental samples obtained from Superfund sites were analyzed
using a simple, shake  extraction with methanol, followed by determination of PCBs with the
quantitative PCB plate ELISA.  In general, it was seen that the quantitative PCB plate ELISA results
were statistically biased  away from the corresponding results obtained using standard  GC methods.  In
the case of one set of samples, the clays from Kansas City, the results were biased high, on the
average, by a factor of approximately 2.5.  For the second set of samples, the soil, sediment and paper
waste samples from the  Allied Paper/Portage Creek/Kalamazoo River site, the ELISA results were
biased low, on the average, by a factor of 0.83.

    Following re-extraction of a randomly selected sub-set of the Kansas City samples with
supercritical fluid extraction, followed by measurement of PCBs with the quantitative PCB plate
ELISA, it was found that the ELISA results for this sub-set of samples were essentially equivalent to
the GC results. This result indicates that the quantitative PCB plate ELISA effectively measured PCB
levels in  the Kansas City clay samples when applied to appropriately prepared sample extracts.

    The quantitative PCB plate ELISA performed well for shake extracts of Allied Paper/Portage
Creek/Kalamazoo River samples with PCB levels below 30 mg/Kg; the ELISA results were not
statistically different from the GC results for this sub-set of the samples.  In contrast,  the PCB plate
ELISA measured significantly lower PCB levels than expected, based on the GC results, in methanol
shake extracts of samples with PCB levels  greater than 30 mg/Kg.

    This latter sub-set of samples were re-extracted using a more rigorous, overnight extraction
procedure, and re-analyzed  using the quantitative PCB plate ELISA.  The ELISA results overlap the
GC results, within the error limits of the respective  methods. The ELISA results were still biased low,
but on the average by only  about 20%, which is of little practical concern considering that this
comparison of results  is  actually an inter-lab, inter-method comparison.  The ELISA results  for the
Allied Paper/Portage Creek/Kalamazoo River samples demonstrated that the quantitative PCB plate
ELISA can provide accurate determination of PCBs in soil, sediment and paper waste provided it is
used in conjunction with adequate sample preparation procedures.

    Assorted environmental samples obtained form NEIC were analyzed using  the quantitative PCB
plate ELISA calibrated with Aroclor 1242 standards.  The results  were biased high, on the average, by
a factor of 1.7 relative to the GC results.  It was later reported by NEIC that the samples contained
mixtures  of Aroclors 1242/1254/1260. The calibration standards were thus somewhat inappropriate for
these  particular samples.

    Based on performance data for the quantitative PCB ELISA, it can be stated that simply by
switching to Aroclor 1254 calibration standards, the quantitative PCB plate ELISA would have
provided results which, on the  average, would be biased low by 30%, which is acceptable performance
for inter-method comparisons based on inter-lab data. This is particularity true for standard methods
intended  for analysis of PCBs.
                                              72

-------
                                   REFERENCES
1.   Albro, P. W., and C. E. Parker., Comparison of the Compositions of Aroclor 1242 and
    Aroclor 1016.   J. Chromatogr. 169:163, 1979.

2.   Albro, P. W., J. T. Corbett, and J. L. Schoeder.  Quantitative Characterization of
    Polychlorinated Biphenyl Mixtures (Aroclors 1248, 1254 and 1260) by Gas
    Chromatography Using Capillary Columns.  J. Chromatogr. 205: 108, 1981.

3.   Bellar, T.A, and JJ. Lichtenberg.  The Determination of Polychlorinated Biphenyls in
    Transformer Fluid and Waste Oils.  EPA-600/4-81-045. U.S. Environmental Protection
    Agency, Cincinnati, Ohio. 1982.

4.   Blasland & Bouck Engineers.  1991. "Allied Paper, Inc./Portage Creek/Kalamazoo River
    Superfund Site: Description of the Current Situation."  Blasland & Bouck Engineers,
    P.C.  Syracuse, NY.

5.   Gee, S., Hammock, B. D., and Van Emon, J. M.  A User's Guide To Environmental
    Immunochemical Analysis. U.S. Environmental Protection Agency, Environmental
    monitoring Systems Laboratory.   Las Vegas, Nevada, 1994.

6.   Hill, D. F., and A. Palomares. PCB Analytical Program Standard Operating Procedures.
    EPA-330/9-89-004.  U.S. Environmental Protection Agency, Office of Enforcement and
    Compliance Monitoring, National Enforcement Investigation Center, Denver, Colorado.
    1989.

7.   Hutzinger, O., Safe, S., and Zitko, V., The Chemistry of PCB'S, CRC Press, Cleveland,
    Ohio, 1974.  pps. 9, 23, 33.

8.   Jensen, S. New Sci., 32:612,  1966.

9.   Schmidt, D.  J., Clarkson, C. E., Swanson, T. A., Egger, M. L.,Carlson, R. E., Van
    Emon, J. M., Karu, A. E., Monoclonal Antibodies for Immunoassay of Avermectins, J.
    Agric. Food Chem., 38: 1763-1770, 1990.

10. Spittler, T. M.  Field Measurement of Polychlorinated Biphenyls in Soil and Sediment
    Using a Portable Gas Chromatograph. In: Environmental Sampling for Hazardous
    Wastes. Schweitzer, G. E., Santolucito, J. A., Eds.  American  Chemical Society:
    Washington, D.C.  1984; vol.  267, pp 37-42.

11. U. S. Department of Energy.   1989.  RCRA Facility Investigation and Corrective
    Measures Study for the Abandoned Indian Creek Outfall.  U.S. Department of Energy.
    Albuquerque Operations Office, Environment and Health Division, Environmental
    Programs Branch, Kansas City, Missouri.
                                         73

-------
12. U. S. Environmental Protection Agency.  1986. Method 8080 - Organochlorine
    Pesticides and PCBs. Manual SW-846,  U.S. Environmental Protection Agency, Office
    of Solid Waste and Emergency Response, Washington, D. C. 8081-1 to 8081-27.

13. U.S. Environmental Protection Agency. 1989.  Development of Immunochemical
    Techniques for Residue Analysis of Environmental Contaminants.  EPA Contract 68-03-
    3511, Work Assignment 1-14.  U. S. Environmental protection Agency,  Environmental
    Monitoring Systems Laboratory-Las Vegas, Las Vegas, Nevada.

14. U. S. Environmental Protection Agency.  1990a. Method 3540A - Soxhlet Extraction.
    Manual SW-846,  U.S. Environmental Protection Agency, Office of Solid Waste and
    Emergency Response, Washington, D. C.

15. U. S. Environmental Protection Agency.  1990b. Statement of Work for Organics
    Analysis, OLM01.0. U.S. Environmental Protection Agency, Office of Solid Waste and
    Emergency Response, Analytical Operations Branch, Environmental Protection Agency
    Contract Laboratory Program, Washington, D. C.

16. U. S. Environmental Protection Agency.  1992a. Demonstration & Quality Assurance
    Project Plan: Field Screening Technologies Analysis of Polychlorinated Biphenyls  At The
    Allied Paper/Portage Creek/Kalamazoo River Superfund Site, Kalamazoo, Michigan.
    U.S. Environmental Protection Agency, Technology Support Center, Las Vegas, Nevada.

17  U. S. Environmental Protection Agency.  1992b. Demonstration and Quality Assurance
    Plan For PCB Immunoassay And Field Screening Technologies. U.S. Environmental
    Protection Agency, Office of Research and Development., Cincinatti, Ohio.

18. U. S. Environmental Protection Agency. 1992c.  Method 8081- Organochlorine
    Pesticides, Halowaxes and PCBs as Aroclors by Gas Chromatography: Capillary Column
    Technique. Manual SW-846.  U. S. Environmental protection Agency, Office of Solid
    Waste and Emergency Response, Washington, D. C.

19. U.S. Environmental Protection Agency. 1993.  Method 3541- Automated  Soxhlet
    Extraction.  Manual SW-846, Proposed Update II.  U. S. Environmental protection
    Agency, Office of Solid Waste and mergency response, Washington, D. C.
                                         74

                                                 •&V.S. GOVERNMENT PRINTING OFFICE: 1994 • 550-001/00204

-------