Evaluation of Analytical Methods for EDCs and PPCPs Via ...

Evaluation of Analytical Methods for EDCs and PPCPs Via
Interlaboratory Comparison [Project #4167]
ORDER NUMBER: 4167
DATE AVAILABLE: March 2012
PRINCIPAL INVESTIGATORS:
Brett J. Vanderford, Jörg E. Drewes, Christiane Hoppe-Jones, Andrew Eaton, Ali
Haghani, Yingbo Guo, Shane A. Snyder, Thomas Ternes, Michael Schluesener, and
Curtis J. Wood
OBJECTIVES:
The objective of this project was to evaluate existing methods for the analysis of
contaminants of emerging concern (CECs) at low ng/L detection levels in water and
provide analytical guidelines for future work.
BACKGROUND:
CECs, such as endocrine disrupting compounds (EDCs) and pharmaceuticals and
personal care products (PPCPs), are organic contaminants that have been detected in
wastewater (WW), surface water (SW), and drinking water (DW) throughout the world
(Ternes et al. 1999, Kolpin et al. 2002, Boyd et al. 2003, Benotti et al. 2009). Because
CECs represent a broad spectrum of compounds, the development of techniques for their
analysis is quite challenging. In addition, CECs vary widely in their physico-chemical
properties and their concentrations in the environment can be quite low (typically sub-
g/L). These and other issues related to CEC analysis (blank contamination, matrix
effects, no standardized/fully validated methods, etc.) have led to the question of whether
results generated by a given method accurately depict the true concentration of each
contaminant in the water.
APPROACH:
The first task was a literature review that would be used to develop a target analyte list. In
addition, a questionnaire was sent to all participating laboratories that requested the
parameters used during all stages of analysis. In the second task, sample collection and
preservation studies were performed to assess preservation strategies and residual oxidant
quenching agents, sample storage times, and bottle types commonly described in the
literature.
The third task was comprised of a series of single-blind interlaboratory comparisons
using unspiked and spiked whole volume aqueous samples prepared by a third party. The
comparisons were divided into three events, each performed in a different aqueous matrix
(deionized water [DI], DW, and SW). The matrices varied in quality and composition to
sufficiently challenge the methods with the goal of identifying which factors in the
analytical process were most important in achieving reliable, accurate results.
©2012 Water Research Foundation. ALL RIGHTS RESERVED.

The fourth task involved the selection and validation of the most promising sampling
parameters and analytical methodologies from the sample collection/preservation studies
and the interlaboratory comparisons. The validation was accomplished with initial
demonstrations of capability (IDCs), showing accuracy and precision in DI and DW, and
a lowest concentration minimum reporting level (LCMRL) study for each analyte to
determine RLs across multiple laboratories. The results of the study were then
summarized to provide guidance for utilities and analytical laboratories on the analysis of
CECs in water.
RESULTS/CONCLUSIONS:
Twenty two target compounds were selected based on the consideration of a number of
factors (physicochemical properties, degree of use/sales, published occurrence and fate
both in the environment and during DW treatment, etc.). The target compounds included
acetaminophen, bisphenol A (BPA), caffeine, carbamazepine, ciprofloxacin, diclofenac,
erythromycin, 17β-estradiol (E2), estrone (E1), 17α-ethynylestradiol (EE2), fluoxetine,
gemfibrozil, ibuprofen, naproxen, 4-nonylphenol (NP), 4-tert-octylphenol (OP),
primidone, progesterone, sulfamethoxazole, testosterone, triclosan, and trimethoprim.
Sample collection/preservation studies
Of the bottle types tested, unsilanized amber glass bottles had the least impact on the
target analytes. As many of the target compounds showed poor recoveries in high density
polyethylene (HDPE) bottles, caution is advised when using these bottles to sample
CECs. Three residual oxidant quenching agents were evaluated: ascorbic acid, sodium
thiosulfate, and sodium sulfite. The majority of the target compounds were compatible
with all of the quenching agents. Since ascorbic acid showed the least interaction, its use
is supported for the target analytes.
Various combinations of temperature, pH, and chemical compounds were investigated as
preservation agents. The optimal preservation conditions were found to be compound
dependent; however, a preservation scheme of sodium azide and storage at 4°C was
found to have the least impact on the target compounds. In addition, sodium omadine was
shown to be a suitable alternative to sodium azide for many of the target analytes. A
holding time of 28 days for SW and DW was demonstrated for sodium azide with storage
at 4°C for the majority of the target compounds. A holding time of 72 hours with storage
at 4°C and no chemical preservatives was also investigated and showed promising results
for many of the compounds.
Interlaboratory comparisons
Twenty-five laboratories participated in the three interlaboratory comparisons, using a
total of 51 methods. The comparisons included representatives of the following types of
samples for every analyte: laboratory reagent blanks (LRBs), low level laboratory
fortified blanks (LFBs), high level LFBs, and/or low and high level laboratory fortified
sample matrices (LFSMs). The target compounds were randomly placed into three
separate groups and spiked, meaning that there were no samples that were exclusively
©2012 Water Research Foundation. ALL RIGHTS RESERVED.

one type, but every analyte had a representative of each type of sample in the
comparison.
For the low spikes, concentrations were set at environmentally relevant levels, while still
maintaining a sufficient number of laboratories that could analyze each compound. As
there were no widely accepted levels of the compounds that were considered harmful to
human health and no regulatory basis for the RLs of the compounds, it was decided to
choose spike amounts that would allow 75% of the laboratories to analyze the low level
spike. The high spikes were set at approximately 5–25 times the concentration of the low
spikes.
Interlaboratory consensus
The percent relative standard deviation (%RSD) was calculated for each target compound
from the results of the interlaboratory comparisons and, overall, %RSDs were compound
dependent with the majority of compounds having %RSDs in the 20–30% range after
outliers were removed. Even with outliers removed, six compounds had %RSDs > 30%,
three of which (ciprofloxacin, NP, and OP) were ≥ 50%. With outliers included, the
%RSDs of all target compounds were > 25% and half were > 40%, showing the presence
of extreme outliers and highlighting the need for the standardization of methods for all
target compounds.
A notable trend in the data was the higher %RSDs exhibited by nine compounds
(ciprofloxacin, erythromycin, E1, fluoxetine, gemfibrozil, ibuprofen, naproxen,
primidone, and trimethoprim) in the first comparison relative to the second and third.
This is most likely due to the use of standardized materials after the first comparison. The
large decrease in %RSD for some compounds suggests that they may degrade easily or
the methods of standard preparation may vary from laboratory to laboratory.
False positives and false negatives
Most compounds showed false positive rates of < 10% and did not demonstrate the need
for RL adjustments based on blank contamination or misidentification of background
interferences. However, six compounds did have false positive rates ≥ 10% (BPA,
caffeine, ciprofloxacin, NP, OP, and triclosan). Most of the compounds had false
negative rates of < 10% with only three compounds having false negative rates ≥ 10%
(ciprofloxacin, EE2, and NP). In most cases, false negatives were the result of only a few
methods reporting RL in the low spike samples, the high
spike samples, or both.
Individual compound analysis
The results of all three comparisons were collectively assessed on a compound by
compound basis and the techniques of choice for the detection and quantification of the
target compounds were overwhelmingly identified as LC-MS/MS and isotope dilution
(ID), respectively. A combination of the two was recommended for 18 of the 22
©2012 Water Research Foundation. ALL RIGHTS RESERVED.

compounds. No conclusions could be made for three compounds (ciprofloxacin, NP, OP).
Based on the analytical difficulty demonstrated here, it is recommended that these
compounds be analyzed by experienced laboratory personnel.
Method analysis
In addition to compound analysis, individual method performance was studied. The target
compounds were divided into three groups (PPCPs, steroids, and NP/OP) and each
laboratory’s methods were assigned to one of those groups. For each method in each
group, a number of factors were considered to identify the best overall method.
PPCPs. Four methods for PPCPs had overall median average absolute biases (MAABs) ≤
10% and 13 of 23 methods had overall MAABs < 20%. Four methods had overall
MAABs > 30%, including two methods with overall MAABs of 64% and 73%. The high
MAABs for some methods indicate that there may be methods currently in use that
cannot accurately quantify many of the target compounds. This also suggests that
comparisons of data generated at different laboratories should be done with caution.
Method 18CD showed low rates of false positives/negatives, analyzed most of the target
compounds, and had low levels of AAB throughout the interlaboratory comparisons. In
addition, this method had an overall MAAB < 10%. Therefore, method 18CD was
selected for validation. This method used LC-ESI-MS/MS for analysis and ID for
quantification.
Steroids. Several of the methods were able to accurately quantify all five steroids, as
three methods had overall MAABs < 10% and seven were < 20%. All of the methods
used LC-MS/MS for detection, with one exception, suggesting that this technique is
appropriate for steroid analysis. Six of the seven methods with the least average MAABs
used ID, suggesting this procedure may be the most accurate means of calibration.
As method 14B had a low overall MAAB, all AABs ≤ 15%, RLs ≤ 1.0 ng/L, and no false
positives/negatives, it was selected for validation. This method uses LC-ESI-MS/MS and
ID for analysis and quantification, respectively. However, as the USEPA had already
published EPA Method 539, method 14B was not carried through the validation process.
NP/OP. Only eight methods analyzed NP/OP in two of the three comparisons and all had
an overall MAAB > 40%. Much of the inconsistency was the result of blank
contamination that affected NP/OP in the second and third comparisons, making accurate
determinations of NP/OP very difficult. As even the top three methods did not provide an
analysis technique that was able to accurately quantify NP/OP in all three matrices at all
spike levels, a method for them was not carried through the validation process.
Method validation
Method 18CD was implemented in a total of six laboratories. The method validation
followed protocols that the USEPA used for the validation of USEPA Method 539, which
©2012 Water Research Foundation. ALL RIGHTS RESERVED.

equired (1) an initial demonstration of capability for low system background and
acceptable accuracy and precision, and (2) the determination of LCMRLs.
Initial demonstration of capability (IDC)
All laboratories were able to demonstrate low system background and, with only two
exceptions, %RSDs calculated from LFBs and LFSMs for every compound were ≤ 10%
in DI and DW, indicating a high level of precision in these matrices. The same set of
replicate data was used to calculate average percent recoveries of the compounds to
evaluate the accuracy of the method. With the exception of one laboratory for one
compound, all mean recoveries were between 90–120%. Therefore, it is expected that
this method will provide similar LFSM recoveries for DI and most DW matrices in the
mid-level calibration range.
Determination of LCMRLs
The LCMRL method was chosen for this study because of its use of both precision and
accuracy and because it has been recently adopted by the USEPA. Five laboratories
participated in the LCMRL determination. Overall, the LCMRLs from the laboratories
were similar and were determined to be at low ng/L levels. The maximum LCMRL for
most compounds was < 10 ng/L, indicating that this method would be expected to yield <
20 ng/L RLs for all target compounds in most laboratories.
APPLICATIONS/RECOMMENDATIONS
Quality assurance/quality control (QA/QC) recommendations
Accuracy
It is recommended that frequent LFSMs be performed to determine the analytical
accuracy in each matrix. In addition, control limits should be determined for each
compound individually due to the compound dependent variations seen in the
interlaboratory comparisons. LFSMs should be performed at both low and high levels
relative to the RLs of the method and the expected concentrations in the samples. A
higher degree of bias was observed at low spike levels, so testing at those levels is
strongly encouraged.
As shown in the final method validation for the PPCPs studied in this project, average
LFSM recoveries in both matrices for these compounds were between 90–120%. This
indicates that LFSM recoveries can be obtained for many of the CECs in drinking water
at ng/L levels. As such, it is recommended that methods for these compounds be able to
demonstrate that they can achieve these recoveries before any samples are collected.
As discussed above, the standardization of solutions had a significant impact on the
outcome of the interlaboratory comparisons. Ciprofloxacin, erythromycin, E1, fluoxetine,
gemfibrozil, ibuprofen, naproxen, primidone, and trimethoprim displayed > 25%
©2012 Water Research Foundation. ALL RIGHTS RESERVED.

eductions in %RSD from the first to the second comparison after standardized solutions
were used. The large decrease in %RSD for some compounds suggests that they may
degrade easily and be more unstable than the others. Therefore, it is recommended that
standards of these compounds are frequently checked for accuracy and freshly prepared
on a regular basis.
Precision
It is recommended that frequent replicates be performed to determine the analytical
precision in each matrix. Most of the target compounds showed average %RSDs > 20%
overall; however, during the method validation, it was shown for PPCPs that %RSDs <
10% were attainable using a single method in multiple laboratories. Accordingly, it is
realistic to expect methods for these compounds to produce %RSDs < 10% in drinking
water. On the other hand, methods for NP/OP were found to be highly variable, in part
due to frequent blank contamination. Methods for these compounds should be thoroughly
assessed in terms of precision to ensure that reproducible results can be obtained.
Blanks
As false positives were recorded for all but one of the target compounds, it is highly
recommended that frequent laboratory, travel, and field blanks be incorporated into any
sampling campaign for these compounds to monitor contamination. This is especially
true for BPA, caffeine, ciprofloxacin, NP, OP, and triclosan, which showed relatively
high false positive rates during the interlaboratory comparisons. Reporting limits should
be assessed in terms of observed contamination and false positive rates and adjusted
accordingly.
Reporting limits
False negatives were frequently observed for many of the target compounds, especially
the steroids. It is therefore recommended that formal RL determinations using established
protocols be conducted to help ensure the reliability of the RLs used for any project
involving these compounds. In addition, LFBs and LFSMs near the RLs should be
analyzed when a method for these compounds is established and at regular intervals
thereafter to verify their validity in the matrices being tested.
MULTIMEDIA:
This report includes a CD-ROM that contains the following material:
A summary of the methods found in the literature review
Summaries of questionnaire responses for each laboratory/method
Detailed analysis of each compound for each interlaboratory comparison
Master data spreadsheets containing raw data from all three interlaboratory
comparisons
SOP of method 18CD that was validated during the project
This material is also available on the 4167 project page under Web tools.
©2012 Water Research Foundation. ALL RIGHTS RESERVED.