Available online at www.sciencedirect.

com

Journal of Chromatography A, 1187 (2008) 58–66

Matrix effects in pesticide multi-residue analysis by liquid

chromatography–mass spectrometry
Anneli Kruve, Allan Künnapas, Koit Herodes ∗ , Ivo Leito
Institute of Chemistry, University of Tartu, Jakobi 2, 51014 Tartu, Estonia
Received 12 October 2007; received in revised form 29 January 2008; accepted 31 January 2008
Available online 6 February 2008

Abstract
Three sample preparation methods: Luke method (AOAC 985.22), QuEChERS (quick, easy, cheap, effective, rugged and safe) and matrix
solid-phase dispersion (MSPD) were applied to different fruits and vegetables for analysis of 14 pesticide residues by high-performance liquid
chromatography with electrospray ionization–mass spectrometry (HPLC/ESI/MS). Matrix effect, recovery and process efficiency of the sample
preparation methods applied to different fruits and vegetables were compared. The Luke method was found to produce least matrix effect. On an
average the best recoveries were obtained with the QuEChERS method. MSPD gave unsatisfactory recoveries for some basic pesticide residues.
Comparison of matrix effects for different apple varieties showed high variability for some residues. It was demonstrated that the amount of
co-extracting compounds that cause ionization suppression of aldicarb depends on the apple variety as well as on the sample preparation method
employed.
© 2008 Elsevier B.V. All rights reserved.

Keywords: Fruits and vegetables; Multi-residue pesticide analysis; Ionization suppression; Electrospray ionization; Matrix effect; LC/MS

1. Introduction LC/MS is highly selective method in selected ion monitor-

Pesticides are widely used in agriculture to fight weeds,
moulds and pests thereby increasing productivity. Besides this
positive effect pesticides pose health-risk to consumers [1].
Therefore concentration of pesticide residues in many prod-
ucts, including fruits and vegetables, must be monitored and
regulations, e.g. SANCO/825/00 [2] have been developed. Gas
chromatography (GC) and high-performance liquid chromatog-
raphy (HPLC) with different detectors are widely used analytical
tools for pesticide residue analysis. The development of pesti-
cides has led to increasingly more polar and thermally labile
compounds [3] rendering their analysis with GC complicated.
HPLC is needed to overcome these difficulties [4]. In order
to achieve good sensitivity and selectivity HPLC coupled to a
tandem mass spectrometer via an electrospray ionization inter-
face (HPLC/ESI/MS/MS) has become the method of choice
for monitoring different pesticide residues in vegetables and
fruits.
∗ Corresponding author. Tel.: +372 737 5259; fax: +372 737 5264.
E-mail address: koit.herodes@ut.ee (K. Herodes).
ing and in multiple reaction monitoring mode. Only the signal of
interest is registered, leaving out the information about the occur-
rence of all the other compounds. This gives the illusion that the
other substances that co-elute with the analyte do not interfere
with the results. However, the other compounds – although invis-
ible in the LC/MS signal – may and very often do interfere. The
suppression or enhancement of ionization by the co-eluting com-
pounds occurs in the ESI source before any mass-spectrometric
detection and it is thus in principle impossible to compensate it
by mass spectrometry [5,6].
Change of ionization efficiency in the presence of other com-
pounds is called matrix effect. Matrix effect was first described
by Kebarle and Tang [7] who demonstrated that the response
of one organic base decreased as the concentration of other
bases increased. The exact mechanism of ion suppression is
not known. Still it has been found that matrix effect may be
caused by nonvolatile material [8] or by compounds of high
surface activity [9]. Matrix effect also strongly depends on
the chemical nature of the analyte. It has been observed that
the ionization efficiency of polar compounds is more influ-
enced by co-eluting compounds than the ionization of less polar
compounds [5]. Some instrumental parameters, e.g. ionization

0021-9673/$ – see front matter © 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.chroma.2008.01.077
 A. Kruve et al. / J. Chromatogr. A 1187 (2008) 58–66
source (APCI vs. ESI) [8] ionization mode [10] of flow rate
[11], have also been found to influence the extent of matrix
effect.
The compounds that cause matrix effect may appear as nor-
mal chromatographic peaks or as broad bands [5,6]. Interference
by a compound eluting as normal chromatographic peak usually
distorts the shape of the analyte peak. The influence of the broad
band may not be visually detectable, but is more easily accounted
for (by means of internal standard or echo-peak technique).
There are some common approaches to take matrix effect
into account. The most accurate one would be standard addi-
tion, which unfortunately increases the number of injections
[12]. Internal standards are not very often used in multi-residue
analyses because matrix effect is highly compound dependent.
An ideal internal standard would be isotopically labeled stan-
dard, but these are often not available or are very expensive [6].
Also it is known that sometimes isotopically labeled standard
substances may cause matrix effect as well [13]. Quite novel
approaches are echo-peak technique [14,15], post-column stan-
dard addition [16] and segmented post-column analyte addition
[17].
It has been recommended by SANCO [2] to use matrix-
matched calibration standards in order to account for matrix
effects. As preparation of standard solutions in all possible
matrixes is unrealistic, representative matrix for a group of
matrixes is being sought. Unfortunately, it has been shown
for pesticide fenbutatin oxide that its ionization is differently
affected by tomatoes, cucumbers and bananas extracts [18]. Also
it has been shown for other matrixes (e.g. human blood [19],
scallops [12]) that ionization suppression/enhancement varies
strongly from matrix to matrix. It has been demonstrated that
using predetermined correction factors to take matrix effect into
account can lead to erroneous results due to high variability
of matrix effects [20]. Also in reference [21] it was found that
matrix effects for fosetyl-aluminium differ from one lettuce vari-
ety to another, increasing relative standard deviation of response
to 57% over 10 varieties.
The abovementioned methods do not minimize matrix effect
but try to take it into account. In order to minimize matrix effects
sample preparation technique [22] and/or chromatographic con-
ditions [23] should be changed. Changing chromatographic
conditions is very time-consuming and may not always be effec-
tive enough.
For pesticide residue analyses in fruits and vegetables many
different sample pretreatment processes have been developed.
One of the earliest methods was the Luke method, which has
been adopted by AOAC as the Official Method 985.22 [24].
Luke method is based on liquid–liquid extraction and uses
a large amount of acetone, dichloromethane and petroleum
ether. A novel “quick, easy, cheap, effective, rugged and safe”
(QuEChERS) [25] method has been published recently and
is becoming the official AOAC 2007.01 method. A careful
validation has been carried out for this method [26]. Unfortu-
nately this study does not include estimation of matrix effect.
The QuEChERS method has also been adapted for base sensi-
tive pesticides [27] and fatty matrixes [28]. Various extraction
processes for pesticide residue analyses, for example pres-
59
surized liquid extraction, stir-bar sorptive extraction, matrix
solid-phase dispersion (MSPD) and solid-phase microextrac-
tion, have been compared by the group of Picó in different
matrixes [29–31]. However, matrix effect is not assessed in these
studies.
In this paper, three sample preparation methods are compared
in order to find method that produces least matrix effect and gives
high recoveries for 14 pesticide residues. Variability of matrix
effect under identical chromatographic conditions for 15 fruits
and vegetables are also studied. Also we compared matrix effects
in the case of five apple varieties.
2. Experimental
2.1. Reagents
Solvents, acetonitrile (J.T. Baker, Deventer, The Nether-
lands), methanol (J.T. Baker), acetone (J.T. Baker and Rathburn
Chemicals Ltd., Walkerburn, Scotland, UK), dichloromethane
(EM Science, Gibbstown, USA), petroleum ether 40–60 ◦ C boil-
ing range (Riedel-de Haën, Seelze, Germany), were of sufficient
purity. Used water was purified with Millipore Simplicity 185
(MILLIPORE GmbH, Molsheim, France). Salts, magnesium
sulfate, sodium sulfate sodium chloride and sodium acetate
were from Reakhim (Leningrad, Soviet Union). Before usage
the magnesium sulfate was baked for 5 h at 500 ◦ C in a muf-
fle furnace to remove possible phthalate impurities. Sodium
sulfate was freed from water and organic impurities by bak-
ing at above 400 ◦ C for 6 h. Glacial acetic acid (Lach-Ner,
Neratovice, Czech Republic) was used to improve stability
of base-sensitive pesticide residues in the final extract of the
QuEChERS method.
Pesticide standard substances were obtained from Dr. Ehren-
storfer (Augsburg, Germany). Stock solutions of 1000 mg/kg
in the appropriate solvent were prepared. Stock solution for
carbendazim was 80 mg/kg because of its poor solubility. The
working standard contained 14 pesticide residues at 40 mg/kg.
For spiking appropriate dilutions were made.
In the MSPD method C8 sorbent (Agilent) with average par-
ticle size 59 ␮m, average pore size 60 Å, surface area 546 m2 /g,
carbon loading 12% was used. The sorbent was not endcapped.
Primary Secondary Amine (PSA) (Supelco, Bellefonte, USA)
was used in the QuEChERS method.
Formic acid (Riedel-de-Haën) and ammonium acetate (Fluka
Chemie AG, Buchs, Germany) were used for preparing eluents
for LC.
2.2. Apparatus
The extracts were analyzed with Agilent Series 1100
LC/MSD Trap XCT (Santa-Clara, USA) instrument using elec-
trospray ionization in the positive ion mode. The LC instrument
was equipped with a binary pump, autosampler, thermostated
column compartment and diode array detector. The mass spec-
trometer uses quadrupole ion trap mass analyzer. For instrument
control Agilent ChemStation for LC Rev. A. 10.02 and MSD
Trap Control version 5.2 were used. Data analyses were per-
60 A. Kruve et al. / J. Chromatogr. A 1187 (2008) 58–66
formed by Quant Analysis for LC/MSD Trap 1.6 and Data
Analysis for LC/MSD Trap 3.2.
2.3. Sample preparation
Fruits were obtained from local trade center and market. A
few of the fruits already contained some of the pesticide residues
that were under study. Sample preparation was carried out for
all of the matrixes and the obtained extracts were injected into
the LC/MS system. The fruits that already contained pesticide
residues were left out of the data treatment for these pesticide
residues. All of the fruits were under the scope of used sample
preparation methods by their water and fat content. The analyses
were carried out as in a routine analysis laboratory. No special
measures were taken to consider potential variability of physical
properties, e.g. pH, of fruits.
About 200 g portion of sample was weighted and there-
after chopped and homogenized for 1 min at 4500 rpm. All
three-sample preparation methods were carried out from same
homogenizate.
Three-sample preparation methods: Luke method [24],
buffered QuEChERS method [27] and MSPD method [29] were
used for sample preparation. In Luke method sample size and
solvent volumes were reduced two-fold. In MSPD method, the
amount of sample, sorbent and solvents was doubled. Procedures
are described in Appendix A.
2.4. LC/MS/MS analyses
Chromatographic separation was carried out on 250 mm
long Zorbax Eclipse XDB-C18 column, with internal diameter
4.6 mm, particle size 5 ␮m. Also an Eclipse XDB-C18 12.5 mm
long precolumn was used with internal diameter 4.6 mm and
particle size 5 ␮m. An autosampler was used to inject 10 ␮l of
probe solution. Gradient elution with methanol and buffer solu-
tion (pH 2.8) was used. Buffer as well as methanol contained
1 mM ammonium acetate and 0.1% formic acid. The linear gra-
dient started at 20% methanol and was raised to 100% within
15 min, then the column was eluted 17 min with methanol and
methanol’s content was lowered to 20% in 3 min. Flow rate of
eluent was 0.8 ml/min.
During the analysis nitrogen was used as the nebulizing gas
(40.0 psi = 276 kPa) and drying gas (10 l/min, 350 ◦ C). Target
pesticide residues with their chromatographic and MS param-
eters are presented in Appendix A. Mass spectrometer was
operated in selected reaction monitoring mode (SRM), maxi-
mum number of observed fragmentation reactions at any time
was three. No internal standard was used.
2.5. Matrix effect, recovery and repeatability studies
Matrix effect was used to describe the analyte ionization effi-
ciency. Recovery describes the efficiency of separating analyte
from the sample. Process efficiency summarizes the efficiency
of sample preparation (recovery) and analyte ionization (matrix
effect). Therefore, process efficiency is suitable quantity for
assessing the overall performance of an analysis method.
Matrix effect (%ME), recovery (%RE) and process efficiency
(%PE) were calculated as follows:
Area of post extraction spike
%ME = × 100% (1)
Area of standard
Area of pre extraction spike
%RE = × 100% (2)
Area of post extraction spike
Area of pre extraction spike
%PE = × 100%
Area of standard
(%ME × %RE)
= (3)
100%
No matrix effect is observed when %ME is equal to 100%, values
over 100% indicate ionization enhancement, and values lower
than 100%—ionization suppression.
Repeatability of matrix effect, recovery and process
efficiency was estimated at 0.10 mg/kg pesticide residue con-
centration in apple. For estimating matrix effect blank matrix
was extracted with three methods, six replicates each. There-
after 100 ␮l of appropriate standard solution was added to 900 ␮l
of extract. Standard solution with the same concentration in
methanol was prepared as well. For recovery estimation homog-
enized sample was spiked with (2.00 ml per 80.00 g of sample)
appropriate standard solution. Six replicates of spiked apple
samples were extracted and analyzed. Process efficiency was
evaluated by comparing the peak areas of pre-extraction addi-
tion extracts with peak areas of calibration solutions prepared in
the solvent.
Reproducibility of recovery and matrix effect over differ-
ent fruits were estimated at three concentrations 1.00, 0.10 and
0.01 mg/kg. In order to evaluate the possible differences between
fruit varieties five apple samples of different variety and origin
(Estonia and Poland) were used to estimate recovery and matrix
effect.
All statistical tests were carried out at 95% confidence level.
3. Results and discussion
3.1. Matrix effect
The matrix effects were determined for 14 pesticides in 15
fruits using three different sample preparation methods (Luke,
QuEChERS, MSPD) at three concentration levels (0.01, 0.10
and 1.00 mg/kg). While the chromatographic separation is not
altered the matrix effect depends on the analyte, matrix and
sample preparation procedure.
Thiodicarb showed strong ionization enhancement in all
matrixes independent of the sample preparation method. This
unpredictable result may be caused by the different properties
of thiodicarb in calibration solution in methanol and in sample
extracts. Potential differences in stability in pure methanol and
in extracts were studied, but this seemed not to be the reason for
enhancement effect. This anomaly must be studied further and
the results of thiodicarb have been left out of consideration in
this paper.
 A. Kruve et al. / J. Chromatogr. A 1187 (2008) 58–66
Fig. 1. Comparison of matrix effects of different sample preparation techniques
for apple at 0.10 mg/kg spiked concentration level (pesticide residues ordered by
increasing retention time). The error bars are given at the level of one standard
deviation over six replicates.
3.1.1. Matrix effect and its repeatability in apple
In order to evaluate differences in matrix effect caused by
different sample preparation methods changes of ionization
efficiency in apple extracts were analyzed by six replicates at
0.10 mg/kg level. The results are presented in Fig. 1 (%ME val-
ues and respective RSD values are presented in Appendix A
Table S-3). It is evident that matrix effects of the Luke method are
generally less pronounced (%ME closer to 100%) than matrix
effects for the QuEChERS method. Statistically significant dif-
ferences (t-test values less than 0.05) between these two methods
were found for demeton-S-methyl sulphoxide, carbendazim,
methomyl, aldicarb, phorate sulphone and methiocarb. Ioniza-
tion suppression for the MSPD method is also smaller than for
the QuEChERS method. Between these two methods differences
are statistically significant for aldicarb sulphoxide, aldicarb sul-
phone, demeton-S-methyl sulphoxide, carbendazim, methomyl,
thiabendazole, methiocarb sulphone and aldicarb. For three pes-
ticides (methomyl, thiabendazole and aldicarb) matrix effect
in MSPD method was smaller than in the case of the Luke
method.
The importance of matrix effect is clearly demonstrated by
the results of aldicarb. Matrix effects of aldicarb in the apple
matrix are 67, 52 and 94% for the Luke, QuEChERS and MSPD
method, respectively. This means that in the case of the Luke and
QuEChERS methods strong ionization suppression of aldicarb
occurs in the obtained extract.
The variability of results, caused by repeatability of sample
preparation, inconsistency of matrix effect and performance of
the chromatographic system, is low for all three-sample prepara-
tion methods. RSD values are lower than 15% for all pesticides
in a single variety of apples.
3.1.2. Reproducibility of matrix effect over different fruits
Reproducibility of matrix effect was determined in 15 dif-
ferent fruits and vegetables (tomato, sweet pepper, orange,
raspberries, banana, cucumber, lemon, blackcurrant, peach,
grape, apple, grapefruit, pear, red currant and leek) at three
concentration levels – 1.00, 0.10 and 0.01 mg/kg. Each fruit
was analyzed once at one of the spiking levels. The variability
of these results is caused by the reproducibility of the LC/MS
61
system performance and sample preparation as well as by dif-
ferences between fruits.
The reproducibility of matrix effects over different fruits
tends to be higher than repeatability in one fruit (one variety
of apple in this case) for all pesticides irrespective of the sample
preparation method. Based on the F-test the Luke method is more
variable over different fruits than in one fruit (apple) for five
pesticides (demeton-S-methyl sulphoxide, methiocarb sulphox-
ide, aldicarb, phorate sulphone and methiocarb), the QuEChERS
method for seven pesticides (demeton-S-methyl sulphoxide,
carbendazim, methomyl, thiabendazole, methiocarb sulphone,
aldicarb and phorate sulphone) and the MSPD method for eight
pesticides (aldicarb sulphoxide, aldicarb sulphone, demeton-
S-methyl sulphoxide, carbendazim, thiabendazole, methiocarb
sulphoxide, methiocarb sulphone and methiocarb). The RSD
values over different fruits were as high as 50% for some
pesticides (see Appendix A Tables S-4, S-5 and S-6). Lower
variability is observed for pesticides eluting at the beginning
of the chromatogram. The high between-fruit variability of
matrix effect shows that different fruits give quite different
matrix effects, which obviously is due to different interfer-
ing compounds in different fruits. For pesticides, that showed
statistically significant differences between reproducibility and
repeatability of matrix effect, the variability of matrix effect
between fruits was calculated sbetween fruits = sreproducibility −
2 2
srepeatability . The results are presented in Table 1.
2
Because of the high variability of the matrix effect over
different fruits statistically significant differences between the
Luke and QuEChERS methods can be observed for aldicarb
sulphoxide, methomyl and methiocarb sulphone. The differ-
ences between the Luke and MSPD methods were significant
only for methomyl and the differences between QuEChERS
and MSPD were significant for aldicarb sulphone, methomyl
and methiocarb sulphone.
In pesticide residue analysis it is a common practice to pre-
pare calibration solutions in an extract of one fruit and use it to
quantitate extracts of other fruits as well. The high variability of
matrix effects over different fruits for many pesticides leads to
the conclusion that matrix-matched calibration solutions should
not be used for the analysis of extracts of other fruits.
Also high variability of matrix effect over different fruits
shows the importance of carrying out (sample preparation and
chromatography) method validation not only at different con-
centrations but also using different fruits and vegetables.
3.1.3. Matrix effect at different analyte concentrations
The estimation of matrix effects over different fruits was
carried out at three concentration levels. Although the average
values of matrix effects do not exhibit statistically significant
differences between the used concentration levels, standard devi-
ations of %ME do (based on F-test)—at 0.01 mg/kg level matrix
has stronger influence on the ionization efficiency of the analyte.
The variability of results is statistically significant for five pesti-
cides (aldicarb sulphoxide, aldicarb sulphone, aldicarb, phorate
sulphoxide and methiocarb) for Luke and QuEChERS meth-
ods. Differences in MSPD extracts of different concentrations
occurred for four pesticides (aldicarb sulphone, thiabendazole,
62 A. Kruve et al. / J. Chromatogr. A 1187 (2008) 58–66

Table 1
Variability of matrix effect in apple (srepeatability ), over different fruits (sreproducibility ) and between fruits(sbetween fruits )

Pesticide residue srepeatability sreproducibility sbetween fruits

Luke QueChERS MSPD Luke QueChERS MSPD Luke QueChERS MSPD

Aldicarb sulphoxide 9% 6% 5% 9% 6% 27% – – 27%

Aldicarb sulphone 12% 4% 5% 12% 8% 15% – – 14%
Demeton-S-methyl sulphoxide 1% 4% 7% 13% 14% 29% 13% 14% 29%
Carbendazim 8% 5% 9% 12% 14% 33% – 13% 32%
Methomyl 17% 10% 10% 33% 27% 23% – 25% –
Thiabendazole 8% 6% 4% 15% 23% 15% – 22% 14%
Methiocarb sulphoxide 6% 11% 8% 25% 20% 21% 24% – 20%
Methoicarb sulphone 17% 9% 12% 28% 27% 62% – 25% 61%
Aldicarb 5% 2% 12% 15% 18% 19% 14% 18% –
Imazalil 14% 11% 7% 18% 18% 12% – – –
Phorate sulphoxide 13% 6% 15% 6% 9% 30% – – –
Phorate sulphone 9% 10% 19% 26% 28% 40% 24% 26% –
Methiocarb 3% 4% 5% 9% 6% 16% 8% – 15%

“–” Indicates that for this pesticide no statistically significant difference of srepeatability and sreproducibility was observed.

imazalil and methiocarb). This high variability of matrix effect
at 0.01 mg/kg level is not caused by system instability, because
reproducibility of calibration solutions of 0.01 mg/kg does not
show higher variability than calibration solution of 1.00 and
0.10 mg/kg do (no statistically significant differences based on
F-test).
For the pesticides of this study the lowest maximum residue
limit is 0.05 mg/kg (e.g. aldicarb in apples and pears). High vari-
ability of matrix effect at lower concentrations may cause false
negative or positive results therefore precautions must be taken
to minimize the matrix effect.
3.1.4. Matrix effect from variety to variety
Assessment of the matrix effect variability between vari-
eties of the same fruit was carried out on the example of
apples. Apples of five different varieties and origin were stud-
ied. Statistically significant variabilities, compared to the ones
found in repeatability studies, were obtained for six pesticide
residues (demeton-S-methyl sulphoxide, methomyl, methiocarb
sulphone, aldicarb, phorate sulphone and methiocarb) in the
Luke method, seven pesticides (aldicarb sulphoxide, carben-
dazim, methomyl, thiabendazole, methiocarb sulphone, aldicarb
and phorate sulphone) in the QuEChERS method and for seven
pesticides (aldicarb sulphoxide, demeton-S-methyl sulphoxide,
methomyl, thiabendazole, methiocarb sulphone, imazalil and
phorate sulphone) in the MSPD method. The matrix effect
variability for aldicarb over different apple varieties in the
QuEChERS method was as high as RSD = 42%, compared to
the RSD = 4% obtained from the repeatability study in a sin-
gle apple variety. Also the RSD values for aldicarb sulphoxide,
methomyl and thiabendazole over different apple varieties were
much higher (44, 39 and 34%, respectively) than in a sin-
gle apple matrix (7, 17, and 7%, respectively) in QuEChERS
extract.
From the large differences in the RSD values within and
between varieties it can be concluded, that matrix effect depends
not only on species but also on its variety. To rationalize this
result the dependence of matrix effect on variety and sample
preparation method in the case of aldicarb in apple was taken
under deeper investigation.
Studying the UV chromatogram and MS/MS chromatogram
for aldicarb (Fig. 2) it can be seen, that an interfering compound
(retention time 13.2 min) elutes next to aldicarb (retention time
12.9 min). This compound is active on the UV chromatogram.
It is not a common case to be able to detect the compound that
causes matrix effect. In this case the compound that interferes
with aldicarb absorbs UV radiation. The UV chromatogram also
shows us that the cause of matrix effect may be present as a
usual chromatographic peak. As it can be seen from Fig. 2 inter-
fering compound distorts the shape of aldicarb’s peak in the
MS/MS chromatogram of the extract compared to that of the
standard solution. A well-defined intensity minimum of aldicarb
in MS/MS occurs in the extract at the point of the highest UV
absorption of the interfering compound. Thus, in addition to the
change in peak shape the area of the peak is remarkably reduced.
These results were confirmed by six replicates.
Careful study of the UV chromatograms shows that the inten-
sities of the compound that interferes with the aldicarb’s peak
vary from one apple variety to another just as do the matrix
effects (see Table 2). As can be seen the matrix effect varies
remarkably: from 36 to 92%. Though changes in chromato-
Fig. 2. Superimposed UV (254 nm) and MS2 chromatograms of aldicarb peak
in QuEChERS extract, spiked at 0.10 mg/kg level.
 A. Kruve et al. / J. Chromatogr. A 1187 (2008) 58–66 63

Table 2
Matrix effect of aldicarb in different varieties of apple, area and height of the
interfering compound (QuEChERS method)
Apple variety Matrix effect (%) Area (mAU s) Height (mAU)

Sibulõun (Estonia) 82 11 1
Sügisjoonik (Estonia) 36 36 2
Kuldrenet (Estonia) 46 28 2
Jonagored (Poland) 92 7 1
Talvenauding (Estonia) 52 87 7

graphic conditions may change the extent of matrix effect it

can be concluded that matrix-matched calibration within one
fruit will not automatically lead to correct results and must thus
be validated before the use over different varieties of the same
fruit.
UV chromatograms of interfering compound for one apple
variety using different sample preparation methods demonstrate
(Fig. 3) that in the extracts of the Luke and QuEChERS methods
the amount of the interfering compound is almost the same. In the
MSPD extract the amount of the interfering compound is much
smaller. The %ME values of aldicarb also vary from method to
method. In the Luke and QuEChERS methods they are 67% and
52%, respectively. In the MSPD method of one apple variety the
intensity of the interfering compound is smaller and the %ME
value is higher: 94%.
3.2. Recovery
Fig. 4. Comparison of recoveries of different sample preparation techniques for
apple at 0.10 mg/kg spiked concentration level (pesticide residues ordered by
increasing retention time). The error bars are given at the level of one standard
deviation over six replicates.
method the MSPD method gives very low recoveries for aldicarb
sulphoxide, demeton-S-methyl sulphoxide, carbendazim, thi-
abendazole, methiocarb sulphoxide and imazalil. The reasons
for this kind of unsatisfactory recoveries were studied sepa-
rately and are discussed below. The MSPD method gives higher
recovery for methomyl than the other two methods.
Repeatability RSD values of recovery are less then 20%
for most of the pesticides. Somewhat higher RSD values were
recorded for pesticides with very low recoveries in the MSPD
method.

3.2.1. Repeatability of recovery 3.2.2. Reproducibility of recovery over different fruits

Similarly to matrix effect, repeatability of recovery was
studied in one variety of apples by six replicates for all three-
sample preparation methods. For most of the pesticides slightly
higher recoveries are obtained with the QuEChERS method
than with the other two methods (Fig. 4). Statistically signifi-
cant differences are observed between the Luke and QuEChERS
methods for aldicarb sulphoxide, aldicarb sulphone, demeton-S-
methyl sulphoxide, methomyl, methiocarb sulphoxide, aldicarb,
imazalil and methiocarb. Compared to the Luke and QuEChERS
Fig. 3. Comparison of UV chromatograms (254 nm) of the extracts showing the
peak of the interfering compound (retention time 13.1 min). The extracts are
obtained from the same apple variety (Sügisjoonik) by different sample prepa-
ration techniques. The UV chromatogram of standard solution demonstrates
absence of peak in that region.
The reproducibility of recovery was studied through repeated
recovery tests in 15 fruits at three concentrations. Over dif-
ferent fruits statistically significant differences of recovery
between Luke and QuEChERS methods were found for aldicarb
sulphoxide, aldicarb sulphone, demeton-S-methyl sulphoxide,
carbendazim, aldicarb and methiocarb. Similarly to the results
in apple recoveries of MSPD method were very low. Recoveries
of 11 pesticides were lower (statistically significant by t-test) for
MSPD method than for the Luke or QuEChERS method.
The variability of recovery values over different fruits is
higher than observed within one fruit. RSD values for repeata-
bility of recovery were less than 20% for all pesticides. With
the Luke method the highest reproducibility RSD over different
fruits (over 1.00 and 0.10 mg/kg) is observed for methiocarb
sulphone 26% and with the QuEChERS method for phorate
sulphone 39%. Recoveries of the MSPD method are low and
vary strongly over different fruits. The RSD values are up to
117% (thiabendazole), which is caused by low recoveries. Sta-
tistically significant (F-test) differences between within-fruit
and between-fruit variabilities were found for two pesticides
(methomyl and methiocarb sulphone) with the Luke method,
for five pesticides (carbendazim, thiabendazole, methiocarb sul-
phone, methiocarb sulphoxide and phorate sulphone) with the
QuEChERS and for seven pesticides (carbendazim, methomyl,
thiabendazole, methiocarb sulphone, methiocarb sulphoxide,
imazalil and phorate sulphone) with the MSPD method. This
means that not only matrix effect but also recovery depends on
fruit species (e.g. pH, water content, fat content). Therefore, esti-
64 A. Kruve et al. / J. Chromatogr. A 1187 (2008) 58–66

mation of recovery during method validation has to be carried
out using different fruits.
Between-fruit RSD values of recoveries for different concen-
trations were compared. At lower concentrations (0.01 mg/kg)
RSD of recovery is higher (statistically significant differences
based on F-test values) for three pesticides (aldicarb sulphone,
aldicarb and phorate sulphone) with the Luke method, for four
pesticides (aldicarb sulphoxide, aldicarb sulphone, methiocarb
sulphoxide and aldicarb) with the QuEChERS method and for
five pesticides (carbendazim, thiabendazole, aldicarb, imazalil
and methiocarb) with the MSPD method.
Overall variabilities of the matrix effect and recovery over
all fruits and pesticides can be compared by comparing the stan-
dard deviation values over all pesticides at one concentration
(for example 0.10 mg/kg) using the F-test. For the QuEChERS
method matrix effect is considerably more variable than recov-
ery. This means that matrix effect depends more on pesticide
as well on matrix analyzed than recovery does. For the MSPD
method this is not so because some pesticides give very low
recoveries and the variability of recovery over all pesticides is
also very high. For the Luke method standard deviation of matrix
effect is higher than the standard deviation of recovery but this
difference is not statistically significant.
3.2.3. Recovery study for MSPD sample preparation
method
From recovery plots obtained by analyzing apple samples
(Fig. 4) it is evident that MSPD method gives for some pesticide
residues very low recoveries. Aldicarb sulphoxide, demeton-
S-methyl sulphoxide, carbendazim, thiabendazole and imazalil
are almost totally lost during sample preparation with the MSPD
procedure. Reasons for these unsatisfactory recoveries may orig-
inate form several steps of sample preparation by MSPD. First of
all, these pesticides may be very strongly retained by the C8 solid
phase. Secondly, as pesticide residues with poor recoveries are
quite polar, these residues may be retained by the free silanol
groups of the stationary phase. These are possibly present in
the C8 solid phase and may be additionally introduced when the
solid phase is mixed with the homogenizate using the mortar and
pestle. This mixing may break the solid phase C8 particles, which
exposes additional silanol groups. Another possible explanation
is that poor recoveries arise from the solution concentration step.
Some pesticide residues might be volatile enough to volatilize
under the stream of nitrogen (especially ones that are liquid at
room temperature).
In order to determine the reason for poor recoveries sample
preparation steps of the MSPD method were separately vali-
dated. Unprocessed C8 solid phase was introduced into one
column and the same amount of solid phase processed with
mortar and pestle was introduced into another column. The
same amount of pesticide standard solution was added to both
columns. Further sample preparation steps were carried out in
the usual way. In order to evaluate the concentration step 1 ml
of pesticide solution in methanol with the appropriate concen-
tration was added to 10 ml of dichloromethane and concentrated
under the stream of nitrogen. The results are presented in Table 3.
It is evident that carbendazim, thiabendazole and imazalil are
strongly retained by both solid phases. So, no differences are
brought about by processing the solid phase with mortar and
pestle. It is important to point out that these pesticides have rela-
tively high pKa values (see Table 3). Free silanol groups present
in the C8 sorbent can adsorb basic compounds. Under the con-
ditions of this sample preparation procedure the adsorption was
irreversible.
Also it can be concluded from these recovery values that
some pesticide residues, for example thiodicarb, are partially
retained by solid phase. Reasons for the very low recoveries
of aldicarb sulphoxide and demeton-S-methyl sulphoxide can-
not be explained by these findings. Still one potential reason
may be that observations of pesticide retention were carried
out using standard solution in methanol. Real samples always
contain some water that can alter the retention of the pesticide
residues on the C8 solid phase. Enzymatic decomposition cannot
be responsible for the poor recoveries, because the recoveries for
these two pesticides are quite good for the Luke and QuEChERS
sample preparation methods. As the recoveries for the concen-
tration step demonstrate, some amount of pesticide residues is
lost during concentration under the stream of nitrogen. This may

Table 3
Recovery evaluation of MSPD steps and pKa values (pKa values of the protonated forms) of analyzed pesticides
Residue pKa [32] Concentration Elution from unprocessed Elution from C8 previously Average recovery in
under N2 flow (%) C8 sorbent (%) processed with mortar and pestle (%) fruits (%)

Aldicarb sulphoxide 1.41 67 54 42 14

Aldicarb sulphone 0.1 66 39 43 51
Demeton-S-methyl sulphoxide – 72 42 48 9
Carbendazim 7.08 50 0 0 5
Methomyl 0.51 97 106 96 81
Thiabendazole 8.19 47 0 0 10
Methiocarb sulphoxide −1.09 52 33 35 41
Methoicarb sulphone −0.33 47 35 24 68
Aldicarb 2.93 53 38 28 61
Imazalil 4.82 53 0 0 2
Thiodicarb 0.51 111 61 50 69
Phorate sulphoxide – 86 56 53 57
Phorate sulphone – 83 38 47 64
Methiocarb 0.21 79 50 48 58
 A. Kruve et al. / J. Chromatogr. A 1187 (2008) 58–66
be due to the volatility of some pesticides. Also it is possible that
the pesticides are adsorbed on the glass walls of the test tube used
for the concentration step.
Better recoveries for basic pesticides by MSPD method have
been reported. Blasco et al. [31] determined 10 pesticide residues
in oranges by MSPD, the lowest recovery 47–52% was obtained
for carbendazim. In ref. [33] C18 is used as solid phase and
recovery of another basic pesticide, imazalil, is up to 95%.
These differences may be brought about by differences in C8
solid phase properties. The C8 used in our experiment was not
endcapped, which might be responsible for high silanol activity.
3.3. Process efficiency
Process efficiency is the overall performance characteristic of
the method. %PE values near 100% generally indicate that both
%ME and %RE are near 100% (Eq. (3)). At 0.1 mg/kg level the
best overall process efficiency is displayed by the QuEChERS
method: the average %PE over different pesticides and different
fruits is 104%, unfortunately corresponding RSD value is very
high 57%. The same for the Luke method is 83% with RSD 27%.
Both methods have on an average acceptable matrix effects and
recoveries. The %PE of the MSPD method is clearly inferior,
while the %ME values are good the recoveries are unaccept-
able. The method yields sufficiently clean extracts but analyte
losses are serious. Corresponding %PE over all the pesticides
and different fruits is 41% and RSD 81%.
The %ME values often markedly differ from 100% leading
to significant differences between %RE and %PE meaning that
these two quantities cannot be used interchangeably. Instead pre-
extraction addition results must be compared to post-extraction
addition results in order to determine recovery. As was shown
previously matrix effects differ significantly between fruits. This
means that post-extraction addition must be carried out using
extract of the same fruit and same variety in which recovery is
estimated. Unfortunately this is not a common practice yet.
4. Conclusion
The aim of this work was to compare three sample preparation
methods and to investigate potential differences of matrix effects
between fruits and varieties.
As the observation showed Luke and QuEChERS method
both give good and adequate overall results. Although the ion-
ization suppression for Luke is a bit smaller than for QuEChERS
for many pesticides, the latter may become a method of choice.
QuEChERS is more economic in terms of time, labor and sol-
vents. QuEChERS has also an advantage of slightly higher
recoveries for many pesticides. Furthermore, as no chlorinated
solvents are used QuEChERS is more environmentally friendly.
Considering the MSPD sample preparation it is obvious that
although the matrix effects on ionization are minimal, unaccept-
ably low recoveries for some pesticide residues are obtained (e.g.
carbendazim, imazalil, thiabendazole). The main reason seems
to be the irreversible adsorption of the more basic pesticides on
the free silanol groups of the sorbent.
65
Matrix effect and recovery are both dependent on the matrix.
For obtaining reliable results during validation both matrix effect
and recovery should be studied for all the fruits and vegetables
that are to be analyzed.
Also it was demonstrated in the case of apples that matrix
effect for a given sample preparation method is dependent on
the apple variety. For example, RSD of matrix effect for aldicarb
over different varieties of apple is 42% compared to RSD 4%
found in a single variety of apple. This finding has serious conse-
quences. In particular, in the case of matrix-matched calibration
one cannot expect that all matrix effects are automatically taken
into account if apples of different variety are analyzed.
It was shown, that process efficiency is a complex quantity,
because it is influenced by matrix effect and recovery. It is much
more useful to record matrix effect and recovery separately.
Acknowledgements
This work was supported by the grant 7127 from the Estonian
Science Foundation and by Estonian Ministry of Agriculture
(under the state project “Agricultural applied research and devel-
opment in 2004–2008”).
Appendix A. Supplementary data
Supplementary data associated with this article can be found,
in the online version, at doi:10.1016/j.chroma.2008.01.077.
References
[1] C. Bolognesi, G. Morasso, Trends Food Sci. Technol. 11 (2000) 182.
[2] Guidance document on residue analytical methods, SANCO/825/00 rev. 7
17/03/2004.
[3] R. Bossi, K.V. Vejrup, B.B. Mogensen, W.A.H. Asman, J. Chromatogr. A
957 (2002) 27.
[4] C.M. Torres, Y. Picó, J. Mañes, J. Chromatogr. A 754 (1996) 301.
[5] P.J. Taylor, Clin. Biochem. 38 (2005) 328.
[6] W.M.A. Niessen, P. Manini, R. Andreoli, Mass Spectrom. Rev. 25 (2006)
881.
[7] P. Kebarle, L. Tang, Anal. Chem. 65 (1993) 972A.
[8] R. King, R. Bonfiglio, C. Fernandez-Metzler, C. Miller-Stein, T. Olah, J.
Am. Soc. Mass. Spectrom. 11 (2000) 942.
[9] N.B. Cech, C.G. Enke, Anal. Chem. 72 (2000) 2717.
[10] E.M. Thurman, I. Ferrer, D. Barcelo, Anal. Chem. 73 (2001) 5441.
[11] A. Kloepfer, J.B. Quintana, T. Reemtsma, J. Chromatogr. A 1065 (2005)
153.
[12] S. Ito, K. Tsukada, J. Chromatogr. A 943 (2001) 39.
[13] H.R. Liang, R.L. Foltz, M. Meng, P. Bennett, Rapid Commun. Mass Spec-
trom. 17 (2003) 2815.
[14] J. Zrostlı́ková, J. Hajšlová, J. Poustka, P. Begany, J. Chromatogr. A 973
(2002) 13.
[15] L. Alder, S. Lüderitz, K. Lindtner, H.J. Stan, J. Chromatogr. A 1058 (2004)
67.
[16] B.K. Choi, A.I. Gusev, D.M. Hercules, Anal. Chem. 71 (1999) 4107.
[17] A. Kaufmann, P. Butcher, Rapid Commun. Mass Spectrom. 19 (2005)
611.
[18] K.A. Barnes, R.J. Fussell, J.R. Startin, H.J. Mobbs, R. James, S.L.
Reynolds, Rapid Commun. Mass Spectrom. 11 (1997) 159.
[19] B.K. Matuszewski, M.L. Constanzer, C.M. Chavez-Eng, Anal. Chem. 75
(2003) 3019.
[20] C. Jansson, T. Pihlström, B.G. Österdahl, K.E. Markides, J. Chromatogr.
A 1023 (2004) 93.
66 A. Kruve et al. / J. Chromatogr. A 1187 (2008) 58–66
[21] F. Hernandez, J.V. Sancho, O.J. Pozo, C. Villaplana, M. Ibáñez, S. Grimalt,
J. AOAC Int. 86 (2003) 832.
[22] K. Bester, G. Bordin, A. Rodriquez, H. Schimmel, J. Pauwels, G. Van-
Vyncht, Fresenius J. Anal. Chem. 371 (2001) 550.
[23] M.D. Nelson, J.W. Dolan, LC-GC Europe (2002 February) 73.
[24] P. Cunniff (Ed.), Official Methods of Analyses of AOAC International,
AOAC International, Gaithersburg, 1997, p. 10, Chapter 10, method
985.22.
[25] M. Anastassiades, S.J. Lehotay, D. Štajnbaher, F.J. Schenck, J. AOAC Int.
86 (2003) 412.
[26] S.J. Lehotay, A. de Kok, M. Hiemstra, P. van Bodegraven, J. AOAC Int. 88
(2005) 595.
[27] S.J. Lehotay, K. Maštovska, A.R. Lightfield, J. AOAC Int. 88 (2005)
615.
[28] S.J. Lehotay, K. Maštovska, S.J. Yun, J. AOAC Int. 88 (2005) 630.
[29] C. Blasco, G. Font, Y. Picó, J. Chromatogr. A 1098 (2005) 37.
[30] C. Blasco, M. Fernández, Y. Picó, G. Font, J. Chromatogr. A 1030 (2004)
77.
[31] C. Blasco, G. Font, Y. Picó, J. Chromatogr. A 970 (2002) 201.
[32] The pKa values were calculated using the SPARC on-line property esti-
mation system: University of Georgia (UGA), Department of Chemistry,
Athens, GA, USA (2007) Available from: (http://ibmlc2.chem.uga.edu/
sparc/).
[33] C.M. Torres, Y. Picó, J. Manes, J. Chromatogr. A 778 (1997) 127.