Trends in Analytical Chemistry, Vol. 27, No. 11, 2008
c) it enables additional tools for identification pur-
shows how these 628 compounds give a MW range of
poses, either via isotope-abundance analysis (IAA)
303.8–304.4 amu, with a distribution (full-width at half
or via accurate-mass-related elucidation of the
maximum, FWHM) of 0.166 amu. This does not include
empirical formula.
the numerous possibilities for fragment ions in this m/z
region that can further complicate the situation, espe-
techniques in LC-MS often yield the (pseudo)-molecular
cially considering the typically low residue levels for the
ion (e.g., [M+H]+), formation of adduct ions with sodium
analyte among complex matrix components at high
or ammonium ions also occurs frequently. In electron
ionization (EI), the molecular ion is practically absent in
If the target analyte contains many elements with a
about 30% of the mass spectra and such a high
large mass defect, such as Cl and Br, then HRMS truly
probability of its absence introduces doubt that the
excels in the suppression of matrix interference. Whereas
highest mass ion present in a mass spectrum is indeed
hydrogen adds only +0.008 amu mass defect per H
the molecular ion. In fact, the highest mass spectral peak
atom, each Cl atom has mass defect of
could be a high mass fragment ion or emerge from an
Br has an even greater mass defect of
impurity. A unique way to enhance molecular ions in EI
Other common elements in organic molecules have the
is by using supersonic molecular beam (SMB)-MS, which
has been reviewed recently With cold EI (EI of
vibrationally cold molecules in SMB-MS) the molecular
. The relatively large mass defects for Cl and Br cause
ion is enhanced and is practically always observed
a significant shift from the center of the mass distribution
while cluster chemical ionization can be used to
for a given MW (see which improves selectivity of
ascertain further the validity of the suspected mass-
analysis for HRMS. Other MS approaches also give en-
spectral peak as the molecular ion
hanced detection of halogenated compounds, such asnegative-ion chemical ionization or even EI in full-scan
3.2. High-resolution MS and accurate mass
mode with isotope-abundance analysis. However, the
High-resolution MS (HRMS) is becoming more popular
superior capabilities of HRMS are fully demonstrated in
in laboratories, particularly in the form of time-of-flight
the analysis of dioxins, PCBs, and similarly multi-halo-
(TOF) MS, while it can also be found in magnetic sector,
genated analytes
Fourier transform (FT) MS, Orbitrap, and even quadru-
Otherwise, most organic chemicals (analytes and
pole MS with software-calibration enhancements. For
matrix components) do not contain Cl or Br, and are
this discussion, we must distinguish between accurate
located near the center of the histogram, as shown in
mass and high resolution. For example, TOF-MS can
for diazinon. In this example with resolution of
have mass accuracy of 2 ppm while its full-width half-
10,000, about 72 other compounds (11.5% of those
maximum resolving power is of the order of only
with MW = 304 amu) in the NISTÕ02 library would
10,000. Thus, in comparison with FTMS or Orbitrap,
have the same measured mass, so a nine-fold improve-
which has similar mass accuracy but a higher resolvingpower of 100,000 TOF-MS has less power to re-duce matrix interferences than its capability to generatepossible elemental formulas.
Accurate-mass TOF is more powerful in LC-API-MS
(because the pseudo-molecular ion is nearly alwayspresent in the spectrum) than in GC-EI-MS because, asstated in Section 3.1, about 30% of the chemicals do notexhibit the molecular ion Also, common EI-MSlibrary-search programs require unit-mass resolution towork properly (although that could be simulated withappropriate software).
Usually, the various vendors who sell high-resolution
TOF instruments describe the power of high resolution inqualitative terms, which makes it hard to compare it toother options (e.g., MS2), so we wish to provide a more
Figure 2. Histogram showing the exact-mass distribution of the
quantitative picture of the merit of HRMS. For this
628 compounds with nominal molecular weight of 304 amu in
purpose, we used the NISTÕ02 library, making the
the NISTÕ02 mass-spectral library. The calculated average mass is
assumption that the library accurately represented the
304.123 amu with a standard deviation of 0.083 amu. The exact
mass of diazinon is marked, and the depicted region demonstrates
MW = 304 amu, the library contained 628 compounds,
how 72 other compounds in the library cannot be resolved fromdiazinon with MS resolving power of 10,000.
including our example-target analyte, diazinon.
Trends in Analytical Chemistry, Vol. 27, No. 11, 2008
ment in selectivity over unit mass-resolution instru-
that the evaluated mass spectral peak is indeed the
ments is realized, which is good but not sufficient for full
molecular ion.
elimination of matrix interference.
However, the ability to determine accurate mass
3.4. Comprehensive two-dimensional GC·GC-MS
enables the analyst to have a table of possible elemental
Comprehensive GC·GC-MS serves as another method for
formulas listed in declining order of matching the
improved sample identification through improved GC
experimental mass. Such a measurement provides a
separation and hence reduced matrix interference. Typ-
strong, independent (orthogonal) tool for analyte iden-
ically GC·GC has a theoretical gain in separation power
tification, which can serve to confirm other methods of
by about 20 through having a second GC time window
identification. To a certain extent, accurate mass can
of 4 s with an average peak width of 0.2 s How-
also be obtained with ‘‘unit'' resolution quadrupole MS,
ever, in reality, the GC·GC separation power is signifi-
and commercial software algorithms can accurately
cantly lower because a proper GC·GC analysis requires
locate the center of the mass-spectral peak. If a typical
that each first-dimension GC peak width will have to
sample mass (e.g., diazinon at m/z 304) is measured with
accommodate three to four GC·GC cycles. As a result,
5-ppm accuracy, this translates into 0.0017-amu
the GC·GC analysis time is generally longer, due to the
accuracy, which reduces the number of possible com-
need to generate broader GC peaks, so the GC·GC sep-
pounds by an impressive factor of 100. However, the
aration power should be compared with a one-dimen-
level of confidence in sample identification by HRMS
sional GC separation with a longer column. Also, the
cannot be greater than the level of confidence that the
second-dimension separation is often not fully orthogo-
evaluated mass-spectral peak is indeed the molecular
nal to that of the first dimension (see Section 2.1).
ion, and this should not be underestimated. If the ana-
Despite the above criticism, GC·GC-MS is a powerful
lyzed ion is actually a fragment ion, unknown to the
analytical tool that can particularly excel in the sepa-
analyst, then the identification will be wrong.
ration of polar samples from non-polar matrix interfer-ences, and, as a result, improve library identification.
3.3. Isotope-abundance analysis
Furthermore, the added separation power of GC·GC-MS
The relative abundances of the various isotopomers
may allow high-quality identifications using two ions (in
(molecular ions with different isotopes) can provide
reconstructed SIM) instead of following the traditional
accurate elemental formulas. Traditionally, the IAA ap-
guideline of three ions. This would provide an even lower
proach has served to help elucidate molecular weights
LOI since the third ion is typically the one with the least
and chemical structures of synthesized organic com-
abundance and greatest chance for matrix interferences.
pounds Recently, a unique IAA method and soft-
ware were developed , among other things, to link
column also increases S/N, which has the effect of fur-
with the NIST MS library and automatically support or
ther lowering LOI compared with one-dimensional GC
reject the proposed library identification. In case of a
rejection, the IAA software independently provides a listof elemental formulas with declining order of matching
the experimental data, similar to accurate-mass mea-
MS2 is a powerful technique that uniquely combines
surements, but IAA does not require costly accurate-
improved sensitivity and selectivity; however, like SIM, it
mass MS instrumentation.
comes with the price of being a target-based method,
Due to the low intensity of the isotopomer ions, key for
which misses any compound that is not in its target list
use of the IAA approach is very low noise with few
(so there is an inherent chance for many false negatives).
chemical and background interferences.
While there are many examples in the literature
There is also demand for absence of protonation (due
showing how MS2 excels in reducing matrix interference
to chemical ionization or self-chemical ionization).
and typically lowers the LOD, we are not aware of any
SMB-MS and SMB-MS2 can provide high sensitivity
publication that has explored how and when MS2 fails
with low background noise while their collision-free fly-
and the factor by which it improves selectivity. In EI-
through EI process excludes undesirable molecular ion
MS2, the inherent dissociation energies for any given
protonation which allows the use of IAA even at
molecule lead to the same fragments as it generates in EI
mass spectra but with different relative intensities. Thus,
Although the IAA software was developed and tested
the MS2 product ions are not a matter of random dis-
mostly with GC-SMB-MS, IAA is also being used effec-
tribution of all masses from m/z 1 to the precursor ion
tively to complement and to supplement accurate-mass
minus 1. If we analyze the NIST library of EI mass
data with high-resolution LC-MS. Again, as with accu-
spectra, we conclude that, for a parent ion with m/z 250,
rate mass, the level of confidence in any sample identi-
only about eight masses are preferred in the mass range
fication, including combined IAA and NIST library
m/z 50–249. For example, hydrogen loss is usually un-
search, cannot be greater than the level of confidence
likely and the next option is a loss of CH3, which is a
Trends in Analytical Chemistry, Vol. 27, No. 11, 2008
difference of 15 amu apart with a void mass space of 14
With respect to choice of ions and their relative
amu. In LC-API-MS, the use of (pseudo)-molecular ions
selectivity, as shows in the case of the NISTÕ98
as MS2 precursor ions further restricts the number of
database, ions of higher m/z have less chance of potential
available possibilities for product ions.
interferences than those with lower m/z. In a real
Moreover, a small percentage of organic chemicals are
application involving the GC-MS analysis of pesticides in
not suitable for MS2 analysis, particularly when EI is
very complicated spice extracts, the degree of matrix
used, due to ion instabilities (or excessive stability to
interference was shown to reduce exponentially by a
allow further fragmentation with typical instruments),
factor of 20-fold per each 100-m/z increase This
lack of enough product ions, or low formation of high-
study also demonstrated how the presence of the
mass ions in MS to allow MS2 (e.g., in terms of identifi-
molecular ion in the spectrum greatly enhanced the
cation power, MS2 lacks the ability to exclude 150,000
ability to isolate the analyte peak in a complicated
other compounds in full-scan spectral libraries). How-
matrix; however, this pattern is a general trend and any
ever, in terms of targeted analysis, MS2 has a second
particular situation depends on the combination of
dimension of selectivity in its collision-induced dissocia-
analyte, concentration, matrix and method.
tion voltage settings, which is part of the optimization
Some chemists have set policies for MS-identification
decisions to account for the general trend shown inFor example, some criteria had dictated that ionswith m/z <91 or many ions from a chlorine or bromine
4. Limitations of current guidelines and rules
cluster should not be used for identification purposes. The intention of such policies is to reduce the
Milman has given an excellent overview of different MS
chances of false positives by analysts with poor judg-
identification criteria used by different organizations
ment, but their arbitrariness can eliminate valuable
and we do not repeat it here, but the information is
information and preclude the use of new technologies
helpful in order to improve understanding of the dis-
that could result in improving identifications. For in-
cussion in this section.
stance, chlorine-ion or bromine-ion clusters indicateboth the presence and the number of those atoms in the
4.1. Targeted monitoring in MS and choice of ions
molecule, and that eliminates a great number of other
There are some major difficulties in implementing typical
possible chemicals in the identification.
MS-identification guidelines and rules, especially with
Another common predisposition is to avoid use of ions
GC-quadrupole MS because its use in the analysis of
with <10% relative-ion abundances . A very impor-
chemical residues tends to require SIM mode to achieve
tant aspect to consider when choosing quantitation and
the necessarily low LOQ. It is quite common for analytes
qualifier ions in MS is to maximize S/N and thereby
to yield only one or two ions of adequate intensity, even
minimize LOD or LOQ and LOI, but this does not mean
in EI, so it is impossible to identify many analytes at
that ions with relative abundance <10% will not yield
reasonably low concentrations using three or four ions.
higher S/N than ions with higher intensity. As shown in
Even when an analyte has P 3 intense ions, some of
an example for permethrin the presence of the
the ions chosen for analysis over the range of targeted
molecular ion at the proper tR with adequate S/N, even if
compounds frequently have an interference from thesample extracts at the tR of interest, so different ionsmust be chosen, depending on the analyte-matrix pair,
to increase selectivity and minimize LOI. The choice ofSIM ions can be difficult, especially in multi-residueanalysis of complex matrices with scores of analytes in
Relative Abundance >5%
the method. A trade-off must be made in the number of
y = 48608e-0.0111x
analytes that can be included in the method vs time and
R2 = 0.9866
the number of qualifier ions, all depending on back-
ground matrix interferences.
Furthermore, in targeted approaches (e.g., SIM and
Relative Abundance = 100%
MS2), ‘‘false negatives'' are guaranteed to occur for non-
y = 2013.8e-0.0107x
R2 = 0.9057
targeted compounds if they occur in the sample. Simi-larly, chemicals of interest or importance that are not
recovered or detected by any particular method could be
considered false negatives if they occur in the sample.
This is why scope of analysis (or analytical range) is
Figure 3. Number of spectra in the NISTÕ98 mass-spectral library
often the most important feature in a method, particu-
plotted vs m/z of the base peak (relative abundance = 100%) andm/z of peaks with relative abundance >5%.
larly in regulatory screening applications.
Trends in Analytical Chemistry, Vol. 27, No. 11, 2008
it was only 1% relative-ion abundance, is a powerful aid
ion-ratio restrictions, and higher mass resolution),
in the identification of analytes. IAA also works on the
which tends to increase selectivity in general, but not
premise that small isotopomer ions may be used to assess
necessarily in particular. An example is how MSn of the
molecular formulas of the chemical to help make iden-
same product ion generated from different compounds
tifications. Such powerful approaches should not be
accumulates more information, but has no value in
distinguishing between the different compounds.
demonstrates the effect of choosing different MS
demonstrates this situation in the case of MS2 of ethion
ions in GC-MS (SIM) for the same application in a dif-
and terbufos using the same fragment, m/z 231, which is
ferent matrix is demonstrated in the analysis of pirimi-
the base peak in their full-scan mass spectra. Another
phos-methyl in carrot and orange extracts vs added
example is that higher mass resolution to obtain exact-
concentrations. Note how the relative intensities and
mass measurements cannot distinguish between isomers
variabilities of the ion ratios ‘‘change'' for the same
with the same molecular formula. Some molecules (e.g.,
analyte in different matrices, due to chemical noise
polychlorinated biphenyl or dioxin congeners) produce
(matrix interferences). For carrots, m/z 125 should be
almost identical mass spectra, and a high-resolution
avoided, but, for oranges, the base peak with m/z of 290
chromatographic separation must be relied upon to help
has a co-eluting interference. This is not unusual, and
make the identification In that case, highly selec-
the analyst has to be aware of the different situations
tive sample preparation followed by high-resolution GC
and take appropriate precautions. However, the wide
(or GC·GC) with selective detection of chlorinated com-
diversity and large number of food, environmental,
pounds would probably be more selective than a rapid
forensic, clinical, and other possible matrices makes it
GC-MS method using high-resolution TOF. In essence,
impractical to choose different qualifier ions depending
the central failing of prescribed mass-spectral identifica-
on the interferences in each matrix . Ultimately, the
tion rules is that they do not apply in all situations .
analyst must take good care and use sound judgment,
The lack of a unified theory for the concept of selectivity
especially when using SIM to make identifications.
in analytical chemistry is an underlying cause of theproblem . For example, it is impossible to provide the
degree of selectivity in a method to such an extent that it
The key to identification of chemicals is not necessarily
demonstrates true specificity (the result can only originate
to characterize the analyte to the greatest extent possi-
from the analyte and no other factor). In science, a
ble, but to exclude the possibility that any other molecule
hypothesis, such as ‘‘this chemical is in the sample,''
possesses the same measured trait(s). Current guidelines
cannot be proved with 100% confidence, and experi-
and rules narrow the focus (e.g., more ions, MS2, tighter
mental evidence can only lend further support to or dis-
Concentration in Carrot (ng/g)
vs. m/z 75
%Rel. Abund.
Concentration in Orange (ng/g)
Figure 4. Ratios of qualifier ions for pirimiphos-methyl vs the base
peak of m/z 290 in GC-MS (SIM) analysis of carrot and orange ex-tracts (n = 4 at each concentration). In carrot, m/z 125 had a
Figure 5. MS2 spectra of terbufos and ethion on an ion-trap instru-
chemical interferent, and in orange, an interferent occurred at
ment using EI. Precursor ion = m/z 231, excitation storage le-
vel = m/z 102, excitation amplitude = 62 V, non-resonant mode.
Trends in Analytical Chemistry, Vol. 27, No. 11, 2008
prove the hypothesis, not prove it. Unlike quantitative
incorrect. The chemist claimed to have identified oxamyl
analysis, qualitative analysis does not provide the actual
in the sample by a GC method employing splitless injec-
degree of confidence that a result is accurately known
tion, but oxamyl is thermally labile and cannot be ana-
because chemical selectivity does not follow a normal
lyzed directly at the GC conditions used. The method had
distribution pattern. We have already mentioned that
actually detected oxamyl oxime, which was produced by
spurious errors probably predominate and that mistakes
thermal decomposition of oxamyl at the hot GC injector.
also do not follow a normal distribution. In trace-residue
Typically, LC methods are used for the analysis of
analyses in complex matrices, perhaps qualitative an-
carbamate pesticides. Alternatively, as shown in ,
swers with 95% confidence can be determined through
they can be analyzed directly by GC-SMB-MS ,
a practical validation process , but statements of
which can provide EI spectra at
>99.9% or >99.99% confidence become problematic
prominent molecular ion, particularly for thermally-
without a solid theoretical foundation for measuring
labile chemicals (e.g., oxamyl), and that is a major
selectivity and/or the analysis of many samples in care-
advantage of SMB over traditional GC-MS.
fully controlled and perfectly conducted experiments.
It is hard to fault the MS expert in misidentifying
oxamyl in the conventional GC-MS analysis. Theoxamyl-oxime metabolite appears in the NIST MS
5. Real-world pitfalls in qualitative analysis
library, but it is listed as ethanimidothioic acid,2-(dimethylamino)-N-hydroxy-2-oxo, methyl ester. The
Everybody makes mistakes, and every analyst can give
lower traces in show the mass spectra of oxamyl
examples of circumstances that led to a mistake. Learning
and its oxime in the NISTÕ02 library (upper trace is the
from the mistakes of others is better than learning from
mass spectrum of cold oxamyl obtained using GC-SMB-
our own mistakes. In this section, we present some
MS ). Given the choice between the exotic sounding
examples of real situations that have led the analyst either
name given above and oxamyl, which was the injected
to make a misidentification or to take extra precautions in
compound after all, the chemist figured that oxamyl was
qualitative MS analyses. We provide these examples of
the correct molecule and unwittingly entered the spec-
real-world pitfalls to underscore the points made in this
trum of oxamyl oxime as oxamyl into the MS data-pro-
article, and perhaps help the reader recognize when a set
cessing software.
of circumstances leading to a possible identification orconfirmation is not as clear as it may seem.
5.1. Chemical degradation (e.g., GC-MS of carbamatepesticides)In a presentation at a pesticide residue workshop, an MSexpert showed how a targeted pesticide was detected in adifficult pepper sample using GC-MS in full-scan data-acquisition mode with mass-spectral deconvolution andcontemporaneous mass-spectral library matching. Thecalibration curve was linear and no matrix interferenceswere observed. The deconvoluted mass spectrum (whichwas well isolated from the complex background) wasessentially identical with the reference spectrum for thepesticide in his mass-spectral library of targeted analytes,which was generated by contemporaneous injection ofcertified pure reference standards using the same methodon the same instrument. The spectrum obtained with EIpossessed five intense ions of >15% relative abundancethat gave a very similar pattern to the spectrum in theNIST mass-spectral library for the pesticide. The tR wasexactly the same, as was the peak shape, with respect tothe reference standard for the analyte. The combinationof all of these factors met the GC-MS confirmation cri-teria established for chemical residue analysis by several
Figure 6. Comparison (top to bottom) of cold EI mass spectrum of
organizations so, for all intents and purposes,
oxamyl using GC-SMB-MS and the NIST library mass spectra
the pesticide was ‘‘identified''.
of oxamyl and oxamyl oxime. Conventional GC-MS using EI does
Despite this, knowledgeable and experienced pesticide
not yield the m/z 219 molecular ion, and hot splitless injectioncauses the conversion of oxamyl to its oxime in the inlet.
chemists in the audience knew that the finding was
Trends in Analytical Chemistry, Vol. 27, No. 11, 2008
This type of mistake is not uncommon; the compound
tissues, two analytes with the same nominal molecular
injected is not necessarily the compound detected or
weight [i.e. albendazole sulfone (MW = 297.328) and
assigned in reference libraries (analysts have also
hydroxy mebendazole (MW = 297.313)] co-eluted using
uncovered misnamed spectra in commercial mass-spec-
a C-18 column (15 cm · 3 mm i.d. with 5-lm particles)
tral libraries), especially if the molecular ion is absent in
with a typical water/MeOH/MeCN buffered reversed-
the EI mass spectrum of that compound. There can be
phase gradient . shows their structures and
other sources of errors (e.g., mislabeling the reference
fragmentation pattern in electrospray positive MS2.
standard, assigning the wrong chromatographic peak as
Without knowledge of this situation, the analyst would
the analyte in a mixture, or not performing background
not be able to distinguish between the two chemicals in
real samples using typical MS2 instruments with unit-
In another example, the MS expert had also included
mass resolution. The situation is further complicated
methomyl in the list of pesticide analytes in the GC-MS
given that hydroxy mebendazole needs a much higher
method. Like oxamyl, methomyl is thermally labile and
regulatory LOD than albendazole sulfone in the EU (there
readily converts to methomyl oxime in certain solutions
is no LOD for mebendazole and its important metabolites,
(e.g., methanol) and/or during traditional GC analysis at
whereas the maximum residue limits (MRLs) for alben-
both the injector and column. Again, the analyst ob-
dazole and its important metabolites are 100 ng/g in
tained a consistent peak with strong ions of m/z 105, 88,
milk and 1 lg/g in bovine liver). In this case, a high-
and 58, which looked much like the NIST library spec-
mass-resolution instrument could help resolve these
trum for methomyl. Unlike oxamyl and its oxime,
compounds, but improving their chromatographic sep-
methomyl oxime is not included in the library and the
aration with a narrower column, smaller particles, and/
US tolerance definition for methomyl does not include its
or more selective stationary phase would also provide a
oxime To further complicate matters, the maxi-
good solution.
mum regulatory limit for a separate pesticide, thiodicarb,includes methomyl in its tolerance definition. If the
5.3. Identifying unknowns (e.g., isofenphos-methyl)
pepper sample had contained methomyl rather than
One of the most difficult jobs for an analyst is to find a
oxamyl and/or oxamyl oxime, then it would have ap-
completely unknown chemical in a sample. Unscrupu-
peared to the analyst that the presence of methomyl was
lous athletes, racing-animal trainers, and food producers
identified in the sample with little doubt. In fact, the
use illegal drugs or pesticides in an attempt to gain a
chemist would have unknowingly identified the presence
competitive or monetary advantage, and they try to use
of methomyl oxime in the sample, which has no regu-
chemicals that cannot or will not be detected. For
latory bearing in the U.S.A., and which could have
example, farmers in Spain apparently obtained an
arisen from unquantifiable concentrations of methomyl,
unregistered pesticide, isofenphos-methyl, from a man-
methomyl oxime, and/or thiodicarb in the original
ufacturer in China and applied it to a crop of peppers
Since there was no registration for the pesticide,
In these cases, the presence of the molecular ion in the
the analytical laboratories had no knowledge of it and
mass spectra of the parent molecules would have en-
there was no reference standard for it. It is common
sured that the oxime metabolites were not confused with
practice in some chemical-residue laboratories to use
the larger pesticide analytes. This example illustrates
GC-MS (SIM) and LC-MS2, which can detect only tar-
both importance of the presence of the molecular ion and
geted analytes. These methods are guaranteed to miss
also the selection of an appropriate analytical method-
any chemical of possible interest that is not targeted in
ology that eliminates undesirable analyte conversion. In
most cases, despite identification guidelines that typicallyrequire more ions for structure elucidation, it is actuallythe presence of distinct ions and other critical informa-
tion (e.g., analyst knowledge and sample history) that
tends to yield better results. Moreover, the use of a more
appropriate type of approach, such as LC-MS2, would
have provided the evidence needed to find the errors.
These examples also demonstrate the importance of
using orthogonally selective (independent) approaches
for the purposes of confirmation.
5.2. Similar analytes (e.g., LC-MS2 of anthelmintic
residues)In another example involving LC-MS2 method develop-
Figure 7. Structures and ESI(+)-MS2 fragmentation patterns ofalbendazole sulfone and hydroxyl mebendazole.
ment for the residue analysis of veterinary drugs in cattle
Trends in Analytical Chemistry, Vol. 27, No. 11, 2008
residues were initially found in food extracts as a curious
There are certain circumstances in which no blanks
extraneous peak in GC using nitrogen-phosphorus
can be found because all samples tested contain the
detection. Full-scan GC-MS was used to make the iden-
suspected chemical. A few examples of this are ultra-
tification and to obtain the multi-analysis confirmation
trace findings of certain ubiquitous persistent organic
after a reference standard of the illegally applied chem-
pollutants in environmental samples or semicarbazide in
ical had been synthesized.
shrimp using very sensitive MS instruments. In these
Some analysts choose to use GC-MS in full-scan mode to
cases, the ultralow background levels are averaged, and
provide the chance to find unknowns (and increase the
a positive finding of interest only occurs when the cal-
potential number of analytes targeted). One advantage for
culated concentration exceeds an ‘‘action level'' or the
those laboratories faced with the isofenphos-methyl
situation was that they re-examined the total-ion chro-
In other cases (e.g., the occurrence of acrylamide in
matograms from earlier analyses to identify the previously
certain types of processed food), the parameter of interest
non-targeted analyte. In the case of LC, full-scan MS with
cannot be isolated in a blank without also generating the
API is not so useful because typically only the pseudo-
chemical that needs to be identified. In the case of
molecular ion appears, and that is insufficient for making
acrylamide, no blanks of crackers, potato chips, French
identifications even in high-resolution applications. The
fries, and similar cooked products containing asparagine
ability to assess the MW or the molecular formula of a
and reducing sugars could be prepared because
chemical is helpful, but additional information is needed
acrylamide is formed during the process used in making
to make an identification in MS. Ferrer and Thurman
the food products, so, when acrylamide was found in
showed that it was possible to obtain in-source fragmen-
these food products using LC-MS2, there was still a
tation in LC-TOF-MS to analyze 101 pesticides while
chance that there had been a false identification (or
maintaining acceptable sensitivity.
quantitative overestimation) because chemical interfer-
In full-scan EI-MS, there are typically so many matrix
ents could not be excluded using a true matrix blank.
peaks (with many overlaps) in chromatograms that it is
Acrylamide is a small molecule (71 amu), and an
much too time consuming to evaluate the spectra for
interferent in potato chips (presumably the amino-acid
each peak . Deconvolution software exists to help
valine that generates the m/z 72 immonium ion) was
, but most chemicals in the chromatograms (e.g.,
observed during sample-preparation experiments .
the examples of isofenphos-methyl and methomyl oxime)
In this situation, confirmation by an alternate method,
do not appear in even the most extensive mass-spectral
namely GC-MS2, was done to verify that the acrylamide
libraries. Furthermore, spectral matching with a con-
had been found by both distinct methods at similar
temporaneously analyzed reference standard is com-
concentrations in the samples. But, even then, the
monly required to make an identification, and simple
analyst had to demonstrate that analytical artefacts had
matching with a library spectrum is not sufficient for a
not occurred, because acrylamide can be formed from its
variety of reasons (e.g., mistakes in the library, differ-
starting materials in the extract during the hot injection
ences between instruments and conditions, or incorrect
process in the GC inlet (or during any sample-
tR). This topic has been debated in the literature
preparation steps using enough heat). Incredibly, it is
and experts in the American Society for Mass Spec-
possible that the analyst can follow all standard practices
trometry have devised guidelines that emphasize the
for two independent MS identifications for orthogonally
need for contemporaneously analyzed reference stan-
selective confirmation of acrylamide, and still be wrong if
an interferent is misidentified in LC-MS2 and acrylamideis formed during sample preparation and/or during
5.4. Situations without a true blank (e.g., acrylamide in
injection (split/splitless) in GC-MS2. In these instances,
there is no substitute for an informed, careful chemist
To ensure the quality of analysis, one of the most
using well-validated methods.
important factors, which is commonly absent from typ-ical identification criteria, is that the analyte signal does
5.5. Contaminated reference standards (e.g., erucamide
not occur in the analysis of a blank matrix. A basic
control experiment in scientific investigations is to isolate
One of the most insidious problems in developing tar-
the parameter of interest (in this case, ‘‘that the signal
geted MS methods (e.g., SIM and MS2) is when a refer-
occurs because the analyte is present in the sample, and
ence standard is mislabeled, has degraded, or contains a
not an artefact''). Analysis of reagent and matrix blanks
contaminant. Just because a bottle has a chemical listed
is standard practice in QC procedures, which are
with a given purity does not mean it is always correct,
designed to exclude certain factors (e.g., instrument-
but this type of false information can be difficult to un-
memory effects, carry over, laboratory or reagent
contamination, or misidentification due to matrix
In an example, reference standards for 3-acetyl-
deoxynivalenol (DON) and 15-acetyl-DON were con-
Trends in Analytical Chemistry, Vol. 27, No. 11, 2008
taminated with erucamide. LC-MS2 conditions are typi-
rinsed the apparatus very well – but not well enough, it
cally optimized using infusion, and the largest peak is
turned out. The pesticides were confirmed to be present
assumed to be the standard, since it typically has nearly
in the extracts by independent GC-MS analyses, but the
100% purity. The erucamide [M+H]+ was m/z 338, with
extracts were contaminated during the filtration steps, so
its C-13 isotope at m/z 339, and, unfortunately, the
the source of the pesticides in the samples had not been
[M+H]+ for 3- and 15-acetyl-DON was also m/z 339.
With infusion, one could not therefore optimize the
In a blind comparison study of samples shared be-
system for either of those compounds, since erucamide at
tween two laboratories to compare the different methods
m/z 339 created analyte-optimization problems, but an
used, a student was asked to review the findings from the
uninformed analyst would not recognize the issue.
GC-MS (full scan) because the chemist did not trust the
In this example, the analyst was aware of the issue
software to do the job automatically. Indeed, the
and chromatographically resolved erucamide from the
instrument missed a large peak for a pesticide (phosmet,
target mycotoxins in LC-MS2 rather than using infusion.
which gives only two strong ions), but the student was
Column fractionation of the impurity and analysis by
day-dreaming during the tedious review process and also
UV/VIS revealed that it gave no UV/VIS signal, yet
missed the obvious peak. The chemist was embarrassed
fractionation of the standard showed a UV/VIS spec-
when the other laboratory identified the pesticide resi-
trum. Analysis of the contaminated standard by LC-UV/
due, and the chemistÕs laboratory had not. The studentÕs
VIS-MS2 also indicated this. So, it was possible in this
excuse was: ‘‘I was part of the method, and if there was a
example to have an impurity in the standard that could
false negative, then it is still the methodÕs fault.''
be detected by LC-MS2. However, other detectors couldassist the analyst in demonstrating the impurity chro-
5.7. Analyte derivatization (e.g., nitrofurans and
matographically had no impact on the identification and
quantitation of the target compound, but would have
The use of the antibacterial agents nitrofurans is banned
resulted in seriously incorrect optimization for the target
in many countries, due to their mutagenic and carcin-
compounds. Basically, information from an independent
ogenic effects. In the analysis of nitrofurans in animal
method (e.g., UV/VIS in this example) can be used in
tissues, the common analytical method calls for over-
addition to MS techniques to aid in identifications as well
night acid hydrolysis and derivatization of the sample
as making confirmations.
followed by LC-MS2 analysis . Nitrofurans arerapidly metabolized to smaller molecules, [e.g., furazoli-
5.6. Spurious errors (e.g., misteaks)
done to 3-amino-2-oxazolidinone (AOZ) or nitrofurazone
We mentioned that spurious forms of errors are probably
to semicarbazide (SEM)] that bind to proteins in the tis-
the most common reasons for misidentifications in
sue. For nitrofurazone converting to SEM, shows
working laboratories. There are numerous examples of
how there are different possibilities that can lead to the
human mistakes that we have found, but the more
same detected derivatized analyte(s) in the method.
troubling factor relates to the mistakes we have not
Another possibility is that the chemicals can already be
found. In the following paragraphs, we give just a few
present in the sample, but originate from a source other
examples that illuminate potential laboratory errors.
than nitrofurans.
An analyst in a laboratory had been preparing a
When the new nitrofuran method was first imple-
concentrated stock solution of many pesticides using
mented, 30% of positive nitrofuran findings by an EU
disposable pipette tips. Even though the used tips were
veterinary-drug reference laboratory in The Netherlands
placed in a box clearly labeled as waste, they were
were for SEM, mainly in prawns and shrimp, at typical
accidentally confused as being new. A different analyst
concentrations of 1 ng/g The other analyte
then used these tips to transfer sample extracts for rou-
commonly detected was AOZ, which like SEM (MW = 75
tine monitoring to autosampler vials for analysis. Each
amu), is a rather small molecule (MW = 102 amu)
extract in the sequence was found to contain a different
comprising common elements C, H, N, and O. Many
pesticide at a rather high level. All identification criteria
enforcement actions have been taken on nitrofuran
were met in each case, and the matrix and reagent
findings in shrimp imported to the EU, which has led to
blanks were clean, but the analyst questioned the re-
large economic losses by exporters and producers .
sults, and, fortunately, the cause of the problem was
In the meantime, SEM was shown to occur in packaged
foods, most probably arising from plastic sealing rings
In another example, high concentrations of endosul-
and carrageenan . However, we should note that
fans were found in river water . Unknown to the
initially no actions were taken until field investigations
analysts, the same pesticide had been applied in an
found containers of nitrofurans in the possession of
agricultural field trial, and run-off water samples were
shrimp producers.
prepared using the same filter apparatus in the labora-
This leads to the question of whether the EU require-
tory. The other group of analysts thought they had
ments, which pertain to the number of IPs achieved in MS
Trends in Analytical Chemistry, Vol. 27, No. 11, 2008
analysis, are valid in this sort of situation. The purpose of
found in the pet food, did not match the pathologic
analysis is to determine if nitrofurans were used, and
causes of the pet deaths. It is possible that folic acid was
inherent aspects of the analytical method do not eliminate
confused with aminopterin, as they have similar struc-
other factors that could lead to the same result. The acid
tures, as shown in .
hydrolysis and derivatization steps affect the degree of
Later, a combination of high concentrations of mela-
selectivity in the overall method, and, although the IP
mine and cyanuric acid were found to be the cause of
system very probably leads to proper MS identification of
kidney failure of the pets that had died from eating the
the derivatized analytes, those are not the compounds of
contaminated pet food.
regulatory interest. The IP rules were developed with the
This example illustrates the need for confirmation by a
selectivity of different MS techniques in mind, and they do
second laboratory in such a high-profile case, but, even
not take into account the lesser degree of selectivity of
within the same laboratory, use of a second method
chemical reactions during sample preparation. An ana-
would have made it less likely that aminopterin would
lyst can therefore achieve >10 IPs of the end products of
have been confirmed, and the press report would not
the method using many sophisticated MS analyses, but
have been issued.
provide no additional information to determine if nitro-furans were used or not. To address this concern, at leastin the case of chickens, investigations have been con-
6. Assuring data quality
ducted to monitor for the parent drug directly in eyeballs,where they may accumulate .
The point of these true stories and real-world discussion
When SEM was found to be occurring naturally and as
is to emphasize that seemingly straightforward analyses
a contaminant in the samples, this created a conundrum
and the analytical decisions that come from them can
for the regulators because their inflexible IP rules (EU/
still be wrong, despite strict analytical requirements and
657/2002) required that regulatory laboratories take
enforcement action when four identification points were
Oxamyl and methomyl are just two among scores of
met, even at extremely low concentrations of <0.1 ng/g.
pesticides that have ‘‘complicating details'' in their anal-
This led to revision of the EU legislation to devise MRPLs,
ysis and regulatory control, and any method that ana-
which essentially set a limit of 1 ng/g for nitrofuran-
lyzes something other than the true compound of interest,
marker metabolites .
as in the nitrofuran example, should face increased
In a similar situation involved in the regulatory
scrutiny and lead to caution in interpretation of results. In
analysis to determine if dithiocarbamate fungicides were
todayÕs environment, greater pressure is placed on the
used illegally in fruit and vegetable production, the most
laboratories to follow ISO 17025 practices and obtain
common approach has been to analyze CS2 liberation
accreditation for their analytical methods, but accredita-
from the samples when treated with tin chloride
tion alone cannot guarantee data quality. In nearly all the
The method detects and quantifies CS2, but there are
examples given, a checklist, such as the one discussed in
known sources of matrix components leading to CS2
Section 6.2, would have helped the analysts avoid mis-
other than just dithiocarbamates, particularly in bras-
takes and make more accurate identifications.
sica-vegetable species, and that makes regulatory actions
The reliance on recipe-type instructions reduces the
questionable. To address this problem, chemists have
role of the thinking analyst and displaces the scientific
been investigating methods to analyze the dithiocarba-
burden of proof (‘‘it wasnÕt me, it was the method'').
mates in the food directly .
That is not to say that analytical procedures should notbe followed precisely or that an organized system and
5.8. Multi-laboratory confirmation (e.g., aminopterin
proper documentation should not be in place, but, in any
or folic acid, or melamine in pet food?)
laboratory, the most critical factor that leads to high-
In this last example, sometimes the stakes are so high
quality analytical data comes from skilled, informed,
that many laboratories are needed to confirm an ana-
responsible analysts, who have sufficient resources and
lytical result. An example of this was the large interna-
time to do their job.
tional investigation to find the cause of a rash of pet
Independent of laboratory accreditation, a valuable
deaths in 2007, which had been associated with pet food
tool for improving the performance of analysts comes
Before obtaining confirmation from other labora-
from the feedback provided in PT-sample programs, as
tories, one organization announced that aminopterin
documented in many examples shows an
had been identified in the pet food and they believed it to
example in the case of European Proficiency Testing
be the cause of the deaths. Although a second laboratory
(EUPT) among approximately 130 accredited laborato-
did confirm the presence of aminopterin at the non-toxic
ries for analysis of pesticide residues in fruits and vege-
level of <4 ng/g, other laboratories could not detect that
tables. The quantitative results improved by about 50%
chemical in the pet food. The known toxicological
from PT 4 to PT 8 in terms of inter-laboratory repro-
symptoms of aminopterin, and the low concentrations
ducibility, and the qualitative performance also im-
Trends in Analytical Chemistry, Vol. 27, No. 11, 2008
Figure 8. Analytical approach for the analysis of nitrofurazone in animal-tissue samples and possible pathways for false-positive results for thecompound detected, NP SEM.
m/z 441 → 294, 175, 120
labs reported 11
Number of Labs 10
Number of Pesticides
m/z 442 → 295, 176, 120
71% of labs
reported 1
reported 1 -16 pesti
-16 pesti
Figure 9. Molecular structures of aminopterin and folic acid.
proved in terms of fewer false negatives, as shown in
for Tests 6 and 8. In many cases, the laboratories
Number of Labs
had to expand the scope of analysis to reduce the
number of ‘‘false negatives,'' and, in other cases, thefeedback obtained by participating in the programs
helped analysts better identify the analytes in blind
Number of Pesticides
Figure 10. Improvement in qualitative results from Tests 6 to 8
Also, the number of false positives reported during
involving 128–130 laboratories participating in the European Profi-
that EUPT program has also reduced over time. In the
ciency Testing Program for pesticide residue analysis of fruits and
first five EUPT sample sets (1–5), 20 false positives
vegetables. In Test 6, there were 13 pesticides present in the
were reported among the laboratories in each test round,
sample, and, in Test 8, there were 16 pesticides present in the sam-
but, in the next five sample sets (6–10), this number
decreased to 610 false positives per round among the130 laboratories. This decrease is attributed to the
6.1. Qualitative method validation
wider implementation of MS systems and improved skills
Due to the lack of a solid theoretical basis for qualitative
and experience among analysts in using the instruments
identifications, qualitative methods must typically be
and the techniques, especially LC-MS2.
assessed empirically, just as quantitative methods are
Trends in Analytical Chemistry, Vol. 27, No. 11, 2008
validated. The real chances of false positive and false
selectivity of the chromatographic separation;
negative results in identifications, which constitute the
instrument performance and proper tuning;
most important qualitative performance factors, need to
be determined with respect to analyte, concentration,
samples and that no analyte carryover or
matrix, and method parameters. Proposals have been
laboratory contamination has occurred;
made to do this during method development
prior knowledge about the history of the sample
and inter-laboratory trials have been performed to
in terms of likelihood that the analyte would be
measure the confirmatory ability of MS techniques in
present, such as its potential use;
blind analyses of real-world samples .
whether the MS-fragmentation pattern can be ex-
The empirical approach is daunting if the measure-
ment is to be made to a high level of confidence (e.g.,
the detected chemical in the extract makes sense
>99%), but a reasonable degree of validation can be
(e.g., analyte-stability issues, concentration, clean
performed to eliminate the use of poor methods (e.g.,
up done, and derivatizations);
95% level of confidence) .
consistency of the analytical result with previous
A multivariate statistical model has been demon-
analyses of the sample(s);
strated to lower the rates of both false positives and false
prior experience of the analyst; and,
negatives in GC-MS identification vs arbitrarily
elimination of other possible compounds that
chosen criteria , but it may not be practical to
could lead to the same result.
analyze enough samples in a blind fashion to determine
One must also recognize that more than a single
the optimized criteria.
compound is often present at any given time in a chro-
Essentially, this practical difficulty leads back to the
matogram, which can generate mixed spectra that can
simpler concept of confirming analytes in the sample by
confuse the identification. For this reason, proper back-
repeating the analysis of a duplicate sample using an
ground subtraction, which may entail mass-spectral
orthogonally selective, validated method (perhaps even
deconvolution of co-eluted peaks, is critical to aid and to
in a second laboratory if the importance of the analysis
improve the process.
warrants that).
6.3. Determination of ion ratios
6.2. Factors in making identifications
A common flaw in typical MS-identification guidelines is
Section 5 presented several examples of how current
the way used to set permissible variability in the relative
approaches to identification can lead the analyst astray,
ion-abundance ratios. The IP rules dictate that, for GC-
not necessarily due to the identification criteria per se,
MS in EI mode, relative ion abundances must be within
but because not all the factors were considered in the
±10% (relative, not absolute value) of the ion intensity of
overall method. In making qualitative decisions, the
the reference spectrum for ions >50% relative abun-
analyst (and/or software) should take into consideration
dance, ±15% for ions <20–50%, ±20% for ions >10–
a number of factors, many of which, but not all, appear
20%, and ±50% for ions 610% In chemical ioni-
in current guidelines . These factors can be included
zation, API and other techniques, a wider acceptable
in developing checklists for analysts to compile the
degree of variability is permitted. Different confirmation
information needed systematically to satisfy fit-for-pur-
criteria set different standards for acceptability, as shown
pose identification criteria:
in a previous report but none are based on specific
information about elemental composition gained
empirical measurements. In reality, some molecules,
from other detectors, isotopic patterns (e.g., Cl
techniques and instruments yield variability in the mass
and Br), or the nitrogen rule (for molecules com-
spectra greater or less than others.
prising C, H, N, O, P, S, Si, and halogens, an odd-numbered ion indicates that an odd number of Nmolecules occurs in the ion, and an even-num-
7. Proposed alternatives
bered ion means that no or an even number ofN molecules appears in the structure);
7.1. Determining variability of ion ratios
correct isotopomer patterns for the assigned
During the validation of a quantitative method using MS
molecular formula;
detection, many analyses are done to evaluate recover-
the choice of qualitative ions and the presence of
ies, repeatability, reproducibility, linearity of calibration
the molecular ion in the spectrum, even if it is
and other factors. The same data can be used to deter-
mine the variability of the ion ratios from day-to-day
actual trends and variability of the chromato-
with respect to concentration in the sample matrix(es).
graphic tR and appearance of peak shape of the
Indeed, some chemists have chosen to measure these
analyte (e.g., peak shifts, tailing factors, and iso-
factors to help set the identification criteria for their
Trends in Analytical Chemistry, Vol. 27, No. 11, 2008
The error bars in indicate the standard deviation
means to ensure that the criteria are reasonable and
(SD) of the measurements of pirimiphos-methyl with
realistic. The multivariate statistical approach should
n = 4 at each concentration (taken from calibration
work even better if enough samples can be evaluated to
standards in matrix analyzed on four different se-
fit into the model
quences). In carrot, the average %relative ion ratios vs
The assessment of ion ratios is just one facet of the
m/z 290 for concentrations P 38 ng/g were 99 ± 3 for
overall process, and the acceptability criteria should not
m/z 276, 86 ± 2 for m/z 125, and 68 ± 2 for m/z 305;
be set so strictly that they act to increase unnecessarily
whereas, in orange, the same approach on the same
the LOI without actually providing a greater degree of
instrument a few months later led to values of 110 ± 7,
selectivity in the analysis. At this time, rather than
108 ± 6, and 73 ± 4 for the same ions vs m/z 290,
defining criteria that are assumed to decrease the rate of
respectively. The change in mass spectra over time,
false positives, such as defining an arbitrary points
including the base peak in this case, due to different
system and ion-abundance windows, the actual mini-
tuning parameters or other factors, is not an uncommon
mization of false positives is better addressed through
occurrence and should be taken into account in the
checklists, many analyses of positive samples and blanks,
confirmatory procedures. Thus, reference spectra should
analyses of blind PT samples, better analyst training, and
be updated frequently, ideally after each tuning.
knowledgeable judgment, taking into account all avail-
Assuming normal (Gaussian) distributions of ion
able information.
intensities, the setting of variability criteria for relativeion intensity dictates the rate of false negatives that will
7.2. Matching factors
occur. For example, at 10 ng/g in carrot, the limiting
In library searches, ion ratios are calculated relative to
third ion for pirimiphos-methyl (m/z 305, the molecular
the base peak, so more emphasis is placed on the base
ion) averaged 72 ± 8% relative abundance vs the m/z
peak than other ions in calculating the spectral fit. The
290 base peak. This indicates that, if the ion-ratio
switch in base peak for compounds, such as pirimiphos-
window is set at ±10% (relative value, thus an accept-
methyl in , can lead to lower match factors, despite
able range of 65–79% relative abundance) as the IP
the analyteÕs presence. Endosulfan is an example where
system requires, then the rate of false negatives at 10 ng/
its spectrum contains dozens of ions, and its pattern is
g in carrot would be 33% according to Gaussian dis-
thereby easily recognizable by the analyst (provided it
tribution theory. A window of ±2 SD (which should be
does not co-elute with chlordane), but the variability of
determined from P 16 measurements) can be used in-
the relative ion intensities is high, resulting in low
stead, corresponding to 5% false negatives. Thus, a
matching factors. Identification criteria should not re-
relative ion-abundance range of 56–88% for m/z 305
quire high spectral match factors (e.g., >90%) that do
would yield an LOI of 10 ng/g for pirimiphos-methyl in
not correspond to reality. Just as the ion ratios can be
carrot by the GC-EI-MS (SIM) method used. Further-
empirically measured during method validation, so can
more, an ion-ratio window of ±3 SD relates to a mere
match factors vs reference spectra to determine the more
0.3% chance of false negatives, which would lead to a
realistic setting for the required match factor at the LOI
relative ion-abundance range of 48–96% in the pirimi-
needed. As discussed previously, contemporaneous
phos-methyl example given.
comparisons with the standard spectra are critical, ide-
In terms of potential false positives, comparison of
ally from both solvent and matrix-matched standards at
these ion ranges in NISTÕ98 with the more narrow
a concentration similar to that of the analyte in the
window defined by the IP system made no difference in
being able to isolate the spectrum of pirimiphos-methyl
In practice, matching thresholds for targeted pesticides
from all others in the library using the three most in-
are typically set very low (e.g., 40%) in order to minimize
tense ions of higher mass.
false negatives, so then the analyst can use judgment to
We suggest use of an empirical approach such as this:
assess whether the preliminary software finding is
first, to determine the variability of the ion ratios
correct or not. Ideally, software would be trusted to
in matrix extracts at the desired concentration; and,
identify any (or all) compound(s) in the chromatogram
then an ion-ratio window can be defined to set the iden-
automatically, but this is seldom the case.
tification criteria for ion ratios at that concentration.
Another important, useful tool is the identification-
Instrument software already commonly allows the
probability factor provided by the NIST library and the
user to set qualifier - ion ratios or spectral matching
listed order of hits. It is better to have a worse fit but
factors in the method, and the software could be further
better hit in the sense of having the suspected analyte
devised to measure variability of mass spectra. By using
come first in the identification-probability list and with a
real data rather than arbitrary criteria, the rate of false
high level of confidence than necessarily with a high
negatives should be reduced. Afterwards, a general
matching factor. A known example to illustrate this
assessment of false positives can be made using blanks,
concept is the mass spectra of simple normal alkanes
blind analyses, mass-spectral library searches, and other
that provide very high matching factors to all, but they
Trends in Analytical Chemistry, Vol. 27, No. 11, 2008
are usually useless for identification because, without a
system suitability and/or QC tests be made before and
molecular ion, they all show the same mass spectra. The
during analytical sequences to ensure that instrument-
use of identification-probability factors (e.g., provided in
performance criteria are being met (e.g., proper calibra-
NIST mass spectral matching software) is therefore pre-
tion, tR, peak shapes, and MS tuning factors).
ferred over simple direct-matching factors, since theidentification probability also takes into account the
7.4. Assessment of false positives and false negatives
degree to which other candidates in the library, which
An important possible measurement of method perfor-
may also have high matching factors, can be excluded as
mance is to determine the percentage of sample analyses
candidates in the identification. The ability to exclude
that give erroneous results using given identification
other library candidates through MS library search
criteria. One simple approach is to analyze in a blind
represents an important advantage of full-scan GC-MS.
fashion a number of samples fortified (or not) withanalyte(s) of interest in matrices of interest. If only final
7.3. Assessment of chromatographic factors
extracts of blanks need to be fortified, then time-con-
Another important factor in MS detection in combina-
suming sample-preparation steps can be minimized. MS
tion with chromatography is the selectivity of the sepa-
specialists often believe such an approach would be a
ration. Direct infusion of the sample into an MS
waste of time because they are confident in their
instrument may be reasonable for screening methods,
instruments and techniques, but when blind analyses
but, for complex samples, an analytical separation
have been conducted in practice (e.g., PT samples), the
should be coupled to the MS detector for increased
number of erroneous results reported can be surprising
selectivity. Just as the MS-ion ratios can be assessed
This approach for perhaps 20 blind samples
empirically, the assignment of the acceptable tR window
would certainly be useful to indicate a problem with the
and peak shape can also be validated empirically during
qualitative aspects of a method (or to verify that there is
the quantitative method-validation process. Measure-
not a serious problem).
ment of tR is a central to chromatography and it is col-
One of the authors of this report challenged application
lected in each injection sequence with respect to system-
chemists from different MS manufacturers to provide re-
suitability tests and calibration standards, and that
sults using their instruments on a series of fruit and
provides a realistic range of values that can be measured.
vegetable extracts that have been fortified (or not) at
Moreover, the different ions used in an MS identification
relevant concentrations with 16 pesticides. Among five
should co-elute with each other (within the scan speed
chemists who accepted the challenge, none has been able
and dwell times of the detection process).
to perform the task well at the first attempt and all re-
provides the average, SD and %relative SD
quired feedback to help them learn from their mistakes.
(%RSD) in the tR of 16 pesticides from 24 injections of
Independent of instrumentation, qualitative analysis
calibration standards in carrot extracts over the course of
cannot be done properly without experience with the
four sequences without performing instrument mainte-
analytes and the matrices. The blind analysis of samples
nance. One possibility is for the tR window in identifica-
in a realistic situation is an excellent way to train ana-
tion criteria to be established based on this kind of
lysts, and it is recommended that this practice be
empirical measurement. For example, using a ±3 SD
window for dimethoate would require that its peak must
chemical-residue analysis. Perhaps in future, an auto-
have tR of 11.841–12.045 min to be considered for
mated system of analyte identification using a neural-
identification purposes. Again, contemporaneous assess-
network approach, in which the software is trained,
ment of tR makes good analytical and practical sense.
much as an analyst is trained by experience, can be
Peak shape and isomeric patterns should also be taken
developed, but, until that time, analyst training is an
into account. For example, the four isomers of cyper-
essential part of qualitative analysis.
methrin that occur in the pesticide formulation present aspecific pattern, which, when observed in the correct tR
7.5. Spectra with too few ions
window, essentially guarantees that cypermethrin has
One of the main difficulties in GC-MS identification by
been detected independent of MS information. Peak
traditional criteria (either EI or CI in full scan or SIM)
shapes, tR values, and asymmetry factors provide further
relates to how many pesticides have mass spectra with
information to help make an identification. The presence
only one intense ion. Notable examples include per-
of a tailing peak at the tR for a compound that gives
methrin (m/z 183), fenthion (278), methoxychlor (227),
sharp peaks (or vice versa) indicates a possible false po-
phosmet (160), terbufos (231), pendimethalin (252),
sitive. Again, the use of contemporaneous injections of
piperonyl butoxide (176), disulfoton (88) and pyri-
standards or standard addition to the injected sample (in
methanil (198). Many other pesticides [e.g., methida-
matrix and at similar concentration level) would ac-
thion (m/z 145 and m/z 85), give only two intense ions.
count for such chromatographic changes. In fact, it is
To obtain three or more ions that meet typical MS-
common for standard operating procedures to require
identification criteria leads to LOIs that are >100 ng/g,
Trends in Analytical Chemistry, Vol. 27, No. 11, 2008
2) enforcement applications to support legal actions.
Table 2. Reproducibility of tR in the GC-MS analysis of 16 pesti-
Different degrees of confidence in identification and
cides in calibration standards prepared from blank carrot extracts(n = 24)
confirmation results are required for these two purposes.
The current guidelines do not differentiate very well
between these different needs, so they do not optimally
suit either purpose of analysis.
The current criteria are not based on objective or
empirical forms of measurement, but are based on
arbitrary criteria using subjective assessments about the
degree of selectivity provided by different MS techniques.
The assessments are generally correct, but there are
many exceptions, depending on the analyte-matrix pair,
the concentration, the MS ions detected, the analytical
technique, and the importance of the results.
The current rates of false negatives are thought to be
too high in both types of application (enforcement and
non-enforcement) because the identification criteria are
too stringent. Furthermore, spurious errors are not typ-
ically addressed to reduce false positives, which are bestaddressed through confirmatory analysis.
especially in complicated matrices, which is too high for
As the word implies, ‘‘confirmation'' requires the
risk assessment and other common regulatory purposes.
results from at least two analyses (which should ideally be
However, if the one or two strongest ions are used, then
orthogonally selective and at least one of them should
the LOI can be reduced greatly.
involve MS detection after an analytical separation),
Independent of the number of ions, we propose that
which, in enforcement applications, must agree with each
enforcement actions continue to entail a second injection
other in terms of analyte identity and concentration.
of the original, re-extracted sample using a different
For non-enforcement purposes, a single analysis using
analytical approach (e.g., LC vs GC), column phase (with
MS detection should be satisfactory, or two analyses
orthogonal selectivity), detector (e.g., element-selective
using non-MS methods, provided that there has been an
detector vs MS), and/or MS technique (e.g., CI vs EI, or
empirical demonstration that acceptably low rates of false
MS2 vs MS) plus inclusion of reagent and matrix blanks
positives and false negatives occur for the analytes at an
. Due to the inability of non-MS methods to provide
adequately low concentration in the targeted matrices.
structural information, at least one of the analytical
An MS qualitative screening approach may be the best
methods should use MS for detection, if the instrumen-
option, followed by a different method to make the con-
tation is available to the laboratory.
firmation and to improve the quantitative determination.
For analyses that do not involve high stakes, such as
We summarize the arguments and the points made in
non-violative findings for risk-assessment purposes, a
this article in the following proposed sequence for
single injection using MS detection should be demon-
quantitative and qualitative method validation.
strated as fit-for-purpose. The method-validation process
Define the need for the analysis (scope of analytes and
would provide a realistic estimate of the LOI, and blind
matrices, concentration ranges, acceptable degree of
analyses would have to be done to demonstrate that no
accuracy, and tolerable rate of false negatives);
false positives occurred in P 10 analyses of different
Define the primary analytical method(s) to meet
sample types using blanks and samples fortified at the
the needs of the analysis most effectively and
LOI level. Again, non-MS techniques do not provide
efficiently (e.g., GC-MS, LC-MS2, GC with selective
structural information, so a second analysis should be
detection, LC-fluorescence, and others);
done using MS or an alternate approach if a non-MS
detector is used in the first analysis.
method(s) for the targeted analyte(s) and ma-trix(es) (or representative analytes and matrices)to determine quantitative characteristics of the
8. Summary of proposed confirmation system for
method(s) according to accepted practices;
chemical residues
At the same time as quantitative method validation,empirically determine MS criteria using samples of
In the chemical-residue analysis of foods and environ-
known concentrations on different days in fortified
mental samples, the main purposes of analysis relate to:
matrices to achieve the desired identification limit
1) surveillance monitoring and data collection for
(with set criteria to minimize false negatives);
non-enforcement reasons; and,
Trends in Analytical Chemistry, Vol. 27, No. 11, 2008
Measure rates of false positives while also training
[9] A. Gentili, D. Perret, S. Marchese, Trends Anal. Chem. 24 (2005)
analysts through analysis of blind fortification sam-
[10] L. Rivier, Anal. Chim. Acta 492 (2003) 69.
ples ( P 10–20 blanks from different sources and a
[11] B. Maralikova, W. Weinmann, J. Chromatogr. B 811 (2004) 21.
similar number of fortified samples at a variety of
[12] B.L. Milman, Trends Anal. Chem. 24 (2005) 493.
concentrations); [Note: this can be done over time
[13] G. Stoev, A. Michailova, J. Chromatogr. A 1031 (2004) 11.
during routine analyses, and such tests should be
[14] E. Soboleva, K. Ahad, A
´ . Ambrus, Analyst 129 (2004)
part of the QC requirements];
[15] G. Stoev, A. Mihailova, J. Chromatogr. A 1047 (2004) 263.
Devise a second approach to be used for indepen-
[16] General Inspectorate for Health Protection, Analytical Methods for
dent confirmation of suspected violative samples
Pesticide Residues in Foodstuffs, 6th Edition, Ministry of Public
from the initial analysis, and conduct blind analy-
Health Welfare and Sport, Amsterdam, The Netherlands, 1996.
ses of extracts to estimate rates of false positives
[17] M. Mezcua, C. Ferrer, J.F. Garcı´a-Reyes, M.J. Marti´nez-Bueno,
and false negatives vs concentration, as in Step (5).
M. Sigrist, A.R. Ferna´ndez-Alba, Food Chem. 112 (2009) 221.
[18] S. Dagan, J. Chromatogr. A 868 (2000) 229.
The proposed validation sequence covers only general
[19] A.B. Fialkov, A. Gordin, A. Amirav, J. Chromatogr. A 991 (2003)
factors, but, in future, we hope that more specific,
practical procedures can be described, just as in the
[20] J.A. Sphon, J. Assoc. Off. Anal. Chem. 61 (1978) 1247.
case of quantitative method validation for chemical
[21] W.C. Brumley, J.A. Sphon, Biomed. Mass Spectrom. 8 (1981) 390.
residue analysis.
[22] R. Baldwin, R.A. Bethem, R.K. Boyd, W.L. Budde, T. Cairns,
R.D. Gibbons, J.D. Henion, M.A. Kaiser, D.L. Lewis, J.E. Matusik,
Further scientific study of the degree of selectivity of
J.A. Sphon, R.W. Stephany, R.K. Trubey, J. Am. Soc. Mass
MS techniques coupled with analytical separations is
Spectrom. 8 (1997) 1180.
needed. Ideally, a systematic approach based on sound
[23] R.A. Bethem, R.K. Boyd, J. Am. Soc. Mass Spectrom. 9 (1998)
theory can be devised to assess accurately the degree of
accuracy in qualitative MS analysis, but, until that time,
[24] European Commission, Off. J. EU L221 (2002) 8.
[25] A. Amirav, Org. Mass. Spectrom. 26 (1991) 1.
we believe that the proposed general approach is more
[26] A. Amirav, A. Gordin, M. Poliak, A.B. Fialkov, J. Mass Spectrom.
scientific and better fits the purposes of chemical-residue
43 (2008) 141.
analysis than the arbitrary criteria currently in place.
[27] A.B. Fialkov, A. Amirav, Rapid Comm. Mass Spectrom. 17 (2003)
[28] M.W.F. Nielen, M.C. van Engelen, R. Zuiderent, R. Ramaker, Anal.
Chim. Acta 586 (2007) 122.
We thank Lutz Alder, David Heller, Eugenia Soboleva,
[30] U.S. Environmental Protection Agency, Tera- through Octa-
Charles Stafford and Stephen Stein for their participation
Chlorinated Dioxins and Furans by Isotope Dilution HRGC/HRMS,
in discussions. This research was supported in part by
the Israel Science Foundation (Grant No. 1172/07). This
[31] R.M. Silverstein, G.C. Bassler, T.C. Morrill, Spectrometric Identi-
research was further supported by the James Franck
fication of Organic Compounds, 4th edition., John Wiley & Sons,
Center for Laser Matter Interaction Research, Chicago,
New York, NY, USA, 1981.
Illinois, USA, and a grant from the Ministry of Science,
[32] T. Alon, A. Amirav, Rapid Commun. Mass Spectrom. 20 (2006)
Culture, Sport of the State of Israel, and Karlsruhe Re-
[33] A.B. Fialkov, U. Steiner, S.J. Lehotay, A. Amirav, Int. J. Mass
search Center (FZK), Germany.
Spectrom. 260 (2007) 31.
[34] J. Cochran, J. Chromatogr. A 1186 (2008) 202.
[35] M. Kochman, A. Gordin, P. Goldshlag, S.J. Lehotay, A. Amirav,
J. Chromatogr. A 974 (2002) 185.
[1] R.A. Bethem, J. Boison, J. Gale, D. Heller, S. Lehotay, J. Loo,
[36] U.S. Dept. of Agriculture Agricultural Marketing Service, Pesticide
S. Musser, P. Price, S. Stein, J. Am. Soc. Mass Spectrom. 14 (2003)
Data Program, Laboratory Quality Control SOPs QC-15, Manas-
sas, VA, USA, 2003.
[2] European Commission, Method Validation and Quality Control
[37] S. Stein, D. Heller, J. Am. Soc. Mass Spectrom. 17 (2006) 823.
Procedures for Pesticides Residues Analysis in Food and Feed,
[38] M. Poliak, A.B. Fialkov, A. Amirav, J. Chromatogr. A 1210
Document No. SANCO/2007/3131, 31 October 2007
[39] U.S. Code of Federal Regulations, Title 40, Chapter 1, Part 180
[3] European Commission, Off. J. EU L71 (2003) 17.
[4] National Academy of Sciences, The Evaluation of Forensic DNA
Washington, DC, USA, July 1, 2002. (
Evidence, National Academy Press, Washington, DC, USA, 1996.
[5] F. Andre´, K.K.G. De Wasch, H.F. De Brabander, S.R. Impens,
L.A.M. Stolker, L. van Ginkel, R.W. Stephany, R. Schilt,
M. Danaher, A. Furey, Anal. Chim. Acta (in press).
D. Courtheyn, Y. Bonnaire, P. Fu
¨ rst, P. Gowik, G. Kennedy,
[41] I. Ferrer, M. Thurman, J. Chromatogr. A 1175 (2007) 24.
T. Kuhn, J.-P. Moretain, Trends Anal. Chem. 20 (2001) 435.
[42] C.R. Borges, Anal. Chem. 79 (2007) 4805.
[6] R.A. de Zeeuw, J. Chromatogr., B 811 (2004) 3.
[43] E.L. Schymanski, C. Meinert, M. Meringer, W. Brack, Anal. Chim.
´ .J. Pozo, J.V. Sancho, M. Iba´n˜ez, F. Herna´ndez, W.M.A. Niessen,
Acta 615 (2008) 136.
Trends Anal. Chem. 25 (2006) 1030.
[44] J.W. Cochran, J. Chromatogr. Sci. 40 (2002) 254.
[8] I. Ferrer, J.F. Garcı´a-Reyes, A. Fernandez-Alba, Trends Anal.
[45] V. Furtula, G. Derksen, A. Colodey, J. Environ. Sci. Health, Part B
Chem. 24 (2005) 671.
41 (2006) 1259.
Trends in Analytical Chemistry, Vol. 27, No. 11, 2008
[46] R.A. de Zeeuw, J. Forensic Sci. 50 (2005) 745.
[57] R.C. Perz, H. van Lishaut, W. Schwack, J. Agric. Food Chem. 48
[47] D.S. Mottram, B.I. Wedzicha, A.T. Dodson, Nature (London) 419
[58] G. Crnogorac, S. Schmauder, W. Schwack, Rapid Commun. Mass
[48] K. Mastovska, S.J. Lehotay, J. Agric. Food Chem. 54 (2006) 7001.
Spectrom. 22 (2008) 2539.
[49] L. Castle, J. Eriksson, J. AOAC Int. 88 (2005) 274.
[59] K. Burns, J. Am. Vet. Med. Assoc. 230 (2007) 1600.
[50] S.J. Lehotay, J.A. Harman-Fetcho, L.L. McConnell, Mar. Pollut.
Bull. 37 (1998) 32 Corrigendum 38 (1999) 1265.
[51] A. Leitner, P. Zollner, W.J. Lindner, J. Chromatogr. A 939 (2001)
[61] J.-P. Antignac, B. Le Bizec, F. Monteau, F. Andre, Anal. Chim.
Acta 483 (2003) 325.
[52] C. Bock, C. Stachel, P. Gowik, Anal. Chim. Acta 586 (2007) 348.
[62] W.J. de Boer, J. van der Voet, W.G. de Ruig, J.A. van Rhijn,
[53] European Food Safety Authority (EFSA), Statement of the
K.M. Cooper, D.G. Kennedy, R.K.P. Patel, S. Porter, T. Reuvers,
Scientific Panel on Food Additives, Flavourings, Processing Aids
V. Marcos, P. Mun
˜ oz, J. Bosch, P. Rodrı´guez, J.M. Grases, Analyst
and Materials in Contact with Food (updating advice available on
(Cambridge, UK) 124 (1999) 109.
semicarbazide in packaged foods), EFSA, Brussels, Belgium, EFSA/
[63] I. Bobeldijk, ‘‘Identification criteria for the GC-MS analysis of
AFC/FCM/17-final, October 1, 2003.
environmental contaminants in various matrices,'' Report KOA
[54] S. Phongvivat, ‘‘Nitrofurans Case Study: ThailandÕs Experience'',
00.100 (in Dutch), Kiwa NV, Nieuwegein, The Netherlands,
Joint FAO/WHO Technical Workshop on Residues of Substances
November 2000.
[64] R. Baigorri, A.M. Zamarren
˜ o, M. Fuentes, G. Gonza´lez-Gaitano,
J.M. Garcı´a-Mina, G. Almendros, F.J. Gonza´lez-Vila, J. Agric. Food
[55] K. Hoenicke, R. Gaterman, Accred. Qual. Assur. 11 (2006) 29.
Chem. 56 (2008) 5480.
[56] K.M. Cooper, R.J. McCracken, D.G. Kennedy, Analyst (Cambridge,
[65] M.J. Schneider, D.J. Donoghue, J. Chromatogr. B 780 (2002)
UK) 130 (2005) 824.
Source: http://www.tau.ac.il/~polaris/Papers/2008%20-%20Ident-TrAC.pdf
Road Traversability Analysis Using Network Properties of Roadmaps* Muhammad Mudassir Khan1, Haider Ali2, Karsten Berns3 and Abubakr Muhammad1 Abstract— Traversability analysis is an important aspect of its traversability which is sometimes not possible. To avoid autonomous navigation in robotics. In this paper, we relate traversing the terrain to find its traversability, exterioceptive
Gebrauchsinformation: Information für Patienten Senshio® 60 mg Filmtabletten Ospemifen Dieses Arzneimittel unterliegt einer zusätzlichen Überwachung. Dies ermöglicht eine schnel e Identifizierung neuer Erkenntnisse über die Sicherheit. Sie können dabei helfen, indem Sie jede auftretende Nebenwirkung melden. Hinweise zur Meldung von Nebenwirkungen, siehe Ende Abschnitt 4.
