Food Chemistry 134 (2012) 24732480

Contents lists available at SciVerse ScienceDirect

Food Chemistry
journal homepage: www.elsevier.com/locate/foodchem

Analytical Methods

Systematic optimisation of coupled microwave-assisted extraction-solid phase

extraction for the determination of pesticides in infant milk formula via LCMS/MS
Guihua Fang a, Hiu Fung Lau a, Wai Siang Law b, Sam Fong Yau Li a,
a
b

Department of Chemistry, National University of Singapore, Singapore, Republic of Singapore

Department of Chemistry and Applied Biosciences, Eidgenssische Technische Hochschule Zurich, Switzerland

a r t i c l e

i n f o

Article history:
Received 9 February 2011
Received in revised form 3 November 2011
Accepted 14 April 2012
Available online 21 April 2012
Keywords:
Infant milk formula
Pesticides
Microwave assisted extraction
LCMS/MS

a b s t r a c t
An extraction method consisting of both microwave-assisted extraction (MAE) and solid phase extraction
(SPE) was developed to extract pesticides of different polarities in infant milk formula. An experimental
central face composite design was established to evaluate the effects of several MAE variables simultaneously such as temperature and pH of extractant. With Derringer desirability function, all pesticides were
optimally extracted at 102 C within 20 min with 4.56 mL of extractant consisting of 0.1% of water in
methanol at pH 12. Extraction efciency was subsequently analysed with LCMS/MS. The extraction
recoveries ranging from 72% to 111% were obtained for all pesticides at fortication level of 5100 lg kg1
with relative standard deviations (RSDs) of less than 8%. Compared with Soxhlet/SPE, the MAE/SPE method
utilised less solvent and higher recoveries were achieved with a shorter extraction time.
2012 Elsevier Ltd. All rights reserved.

1. Introduction
Pesticides, which are also classied as persistent organic pollutants (POPs), are widely utilised for crop protection and pest control. The extensive application of pesticides is required to increase
the agricultural productivity such that the global food demand
can be met. Atrazine, for example, is one of the most frequently
used pesticides, for the pre- and post-emergence of grasses, weeds
and crops in United States. Other pesticides such as cyromazine are
utilised for y control in crop protection.
With the widespread usage of pesticides, residual amounts of
pesticides are being detected in both water and food cycles, posing
danger to human health. Epidemiologic studies were conducted to
evaluate the toxicity of pesticides. For example, US Environmental
Protection Agency (EPA) evaluated the chronic toxicity of atrazine
and its metabolites in Sprague Dawley rats to postulate the possible effects of persistent exposure to pesticides on humans (EPA,
1999). Atrazine was proposed to be classied as a likely human
carcinogen because of the elevated incidence of mammary tumours observed in female Sprague Dawley rats. In recent study,
the association between adverse birth outcomes and urinary biomarkers of prenatal atrazine exposure was assessed (Chevrier
et al., 2011). It was observed that the presence of quantiable levels of atrazine or a specic atrazine metabolite was associated with
fetal growth restriction and small head circumference for sex and
gestational age.
Corresponding author. Fax: +65 67791691.
E-mail address: chmlifys@nus.edu.sg (S.F.Y. Li).
0308-8146/$ - see front matter 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.foodchem.2012.04.076

Thus, the detection of pesticides at trace level is of vast importance to safeguard the wellbeing of the people. Analytical techniques have been developed to quantify the pesticides at trace
levels. One of the emerging areas of interest is the monitoring of
pesticides in infant products such as infant milk formula which is
the main source of food for infants. Pesticides can contaminate
the milk via three pathways: (i) the consumption of animal feed
or pastures, that have been sprayed with pesticides, by dairy cattle,
(ii) the spraying of pesticides on animals for ectoparasitic control
and (iii) the usage of pesticides in factory farms (Mezcua et al.,
2007). The concentration of pesticides present in milk is dependent
on its stability, the amount and time of exposure as well as the
metabolic rate of the animals. The triazine pesticides, for instance,
undergo degradation via three main reactions: dealkylation, oxidation and dechlorination in the environment. (Hapeman, Bilboulian,
Anderson, & Torrents, 1998) For example, cyromazine undergoes
dealklyation to form melamine which may result in renal diseases
with chronic consumption (Sun, Wang, Liu, Qiao, & Liang, 2009).
Since the presence of pesticides in food can also lead to acute
neurologic toxicity and chronic neuro-development impairment,
regulatory bodies such as US Food and Drug Administration (FDA,
1996) have set the maximum residue levels (MRLs) to monitor the
concentration of pesticides in a variety of food. Based on the Commission Directive 2006/125/EC (EU, 2006), the European committee
has regulated that not more than 10 lg kg1 of pesticides should be
present in infant food. In addition, Commission Regulation No.
1135/2009 (EU, 2009) has also stated that the concentration of
melamine should not exceed 2.5 mg kg1 in infant milk formula.
The MRL of pesticides are usually signicantly lower in infant food

2474

G. Fang et al. / Food Chemistry 134 (2012) 24732480

compared to that for adults as the infants body weight are lighter.
Moreover, infants consume the same type of food such as infant
milk formula repetitively. Hence, they are more prone to the adverse effects of pesticides in food than the adults.
In order to monitor pesticides at trace level in complex food
matrices, sample preparation involving both extraction and
clean-up procedure is essential. Liquidliquid extraction and Soxhlet extraction are established as the extraction method of choice to
analyse pesticides in food and environmental samples (Andreu &
Pico, 2004; Srogi, 2006). Despite their good extraction efciencies,
these methods have several drawbacks which include the usage of
large volume of organic solvents, time consuming procedures and
difcultly in automating such method. Recently, new extraction
techniques such as supercritical uid extraction (SFE), pressurised
liquid extraction (PLE) and microwave-assisted extraction (MAE)
which utilised elevated pressure and temperature have been reported for the extraction of pesticides. However, the experimental
set-ups of both PLE and SFE are highly costly. Moreover, organic
modiers like methanol, which are added to increase the polarity
of the carbon dioxide extractant, may also introduce additional
matrix materials that may be co-extracted in the SFE. This will reduce the selectivity of the extraction. On the other hand, for PLE,
poor extraction efciencies were observed for wet samples using
non-polar extraction solvent (Schantz, 2006).
Unlike traditional heating in which heat is transferred from the
container to the samples, MAE involves the use of microwave energy to heat the polar solvents and samples directly via ionic conduction and dipole rotation. Consequently, MAE has a greater
extraction efciency compared to traditional heating. MAE also offers several advantages which include low solvent consumption
and high sample throughput. Besides, MAE has demonstrated
excellent capability in extracting POPs such as lipophilic pesticides
and polycyclic aromatic hydrocarbons from complex matrices
(Wang, Meng, Lu, Liu, & Tao, 2007).
Since numerous publications have focused on the effects of several variables such as the type of solvent, temperature, time and pH
on MAE efciency (Stout, daCunha, & Allardice, 1996), experimental
design is applied in this study to examine the effects of these variables on MAE efciency. The traditional optimisation approach of
examining one variable at a time, involve the investigation of one
response by changing one variable while maintaining all the other
factors at a constant level. It is commonly utilised to determine
the best extraction condition, for a range of compounds despite
their difference in polarities. Its major disadvantage is that it does
not include the additive or synergistic interaction among the variables (Bianchi & Careri, 2008). As a consequence, this technique
does not portray the complete effects of the parameters on the response. In addition, more experiments are usually conducted based
on trial and error or through numerous ne tuning of the variables
to achieve good extraction recoveries for most of the compounds of
interest. This results in an increase in the analysis time and use of
reagents. With the application of experiment design, a more systematic approach can be adopted. The minimum and maximum value of variables is xed from the start of the experiment, taking into
account of the limitation of the instruments or extraction methods.
The number of experiments required is dependent on the designs
chosen such as Box-Behnken, full factorial design and Doehlent.
These symmetrical designs differ from one another with respect
to their selection of experimental points, number of levels for variables, and number of runs and blocks. Unlike the traditional approach, these experiments involve the optimisation of several
extraction variables simultaneously. In addition, the optimal MAE
condition for the pesticides of different polarities can be achieved
with the aid of the Derringer desirability function.
After extracting the pesticides with optimal MAE condition,
clean-up procedures will be carried out to remove the co-extracted

matrix compounds. The presence of these matrix interferences can

result in generation of emulsions, sample turbidity and masking of
analyte signals which can decrease the sensitivity of the LCMS/MS
method (Pawliszyn, 2002). Clean-up methods such as gel permeation chromatography and SPE are often employed to remove the
co-extractives. In this work, SPE was selected to remove the fats
in the liquid extract of infant milk formula and to further pre-concentrate the pesticides.
Both gas chromatography (GC) and liquid chromatography (LC)
have been utilised for the analysis of pesticides. Unlike GC, laborious and costly derivatisation steps can be avoided in LC especially
for the analysis of polar compounds such as melamine (Ihunegbo,
Tesfalidet, & Jiang, 2010). LC has been coupled to convectional
detectors such photo diode array detector. However, mass spectrometry (MS) is preferred as it provides conrmatory evidence
of the identity of the compound. Thus liquid chromatography tandem mass spectrometry (LCMS/MS) was utilised in this study.
The aim of this work was to establish a reliable MAE and SPE
method to extract trace quantities of both hydrophilic and hydrophobic pesticides from infant milk formula. Experimental design
with the use of Derringer desirability function was applied to predict the optimal microwave extraction conditions. SPE was utilised
to clean-up and pre-concentrate the extracted pesticides. Subsequently, the supernatant was analysed by LCMS/MS. Comparison
studies involving Soxhlet/SPE and MAE/SPE were performed.
Method validation was eventually conducted to determine the various concentrations of fortied pesticides in infant milk formula.
2. Materials and methods
2.1. Chemicals and reagents
The common pesticides including prometryn, ametryn prometon, atrazine, simetryn, atraton, cyromazine and melamine were
bought from Riedel-de Haen (Seelze, Germany). Ammonium formate and formic acid were purchased from Sigma (Steinheim, Germany). Hydrochloric acid, potassium hydroxide, HPLC grade
acetonitrile and methanol were obtained from Merck (Darmstadt,
Germany). Ultrapure water with resistivity of 18.2 MX (Milli-Q
plus system, Millipore, Bedford, MA, USA) was used throughout
this work. 1000 mg L1 of pesticides stock solutions were prepared
in a mixture of methanol: water (7:3 v/v). Melamine standard solution was dissolved in 20 mM phosphate buffer of pH 1.5. The pesticides stock solutions were stored in the dark at 4 C. Matrixmatched standards were prepared by adding appropriate amounts
of standard solutions to the blank milk extracts that had undergone the described extraction and preconcentration with SPE.
The infant milk formula was purchased from a local supermarket
while the Oasis MCX solid phase cartridges (6 cc, 500 mg) were
supplied by Waters (Milford, MA).
2.2. LCMS/MS analysis
For quantitative analysis, a HP 1200 liquid chromatographic
system was interfaced with Agilent 6410 series triple quadrupole
mass spectrometer (Waldbronn, Germany). It was equipped with
an electrospray interface (ESI), an automatic injector, a degasser,
a quaternary pump and a column oven. Chromatographic separation was carried out on a Zorbax Rx-SIL (4.6 mm  250 mm id)
with 5 lm particle size (Agilent Technologies, Waldbronn, Germany). The gradient eluant system consisted of 5 mM ammonium
formate and 0.1% formic acid in water (solvent A) and 5 mM
ammonium formate and 0.1% formic acid in acetonitrile (solvent
B). Separation was achieved based on the following gradient: 98%
B decreased to 65% B in 7 min. The mobile phase ow rate and

2475

G. Fang et al. / Food Chemistry 134 (2012) 24732480

column temperature was set at 1 mL/min and 25 C respectively.

5 lL of sample was injected for analysis.
Analyses were performed in the multiple reaction monitoring
(MRM) in the positive ion mode. Supplementary Table 1 summarised the parameters employed in this study after optimisation
through ow injection analysis. Nitrogen was used as nebulising
gas, curtain gas, collision gas, and auxiliary gas. The optimised
MSMS parameters were as followed: a N2 ow rate of 10 L/min,
a spray potential of 4000 V, a nebulizer pressure of 40 psi (N2)
and a source temperature of 325 C.
2.3. Experimental design
Design of experiment was established with the software
(MODDE version 8) from Umetrics AB (Umea, Sweden). With the
mass of infant milk formula xed at 0.5 g, central face composite
design, which has axial points located at the centre of each face,
was developed. The variables such as extraction temperature, time,
volume of extraction solvent, pH and composition of water in solvent were optimised to investigate the effects on extraction recoveries. Three levels with coded values of 1, 0 and 1, representing
the lowest, mediocre and highest value of each variable was shown
in Table 1. The experiments were conducted randomly in order to
reduce bias from uncontrolled variables. Each analysis was performed in triplicate.
After completing the experiments, the results showed that the
optimal conditions for each individual pesticide did not coincide
with each other. (Supplementary Table 2) The MODDE software
utilised the Derringer desirability function to determine the ideal
extraction condition for all eight pesticides without having to perform additional experiments. The Derringer desirability function
involves creating a desirability function for each individual response (di), dened as a partial desirability function. This function
can range from zero (undesirable response) to one (optimal response). An overall desirability function (D) was subsequently
maximised to select the best extraction variable conditions. D is
dened as the weighted geometric mean of the partial desirability
functions. The function D is calculated using Eq. (1) (Derringer &
Suich, 1980):

"
#1=n
n
Y
pi
D
di

i1

where di is the partial desirability function of each response obtained from the transformation of the individual response of each
experiment, and Pi is the weight of the individual function, which
Pn
is normalised such that
i pi 1. For this study, the minimum
and the target value for the extraction recoveries were set at 80
110% respectively while the weight was xed as 1.
2.4. Extraction procedures
2.4.1. Microwave extraction
A Biotage Initiator microwave 2.5 (Uppsala, Sweden) was utilised for the extraction of pesticides from infant milk formula. During the optimisation process, xed concentration of pesticides

(2 mg kg1 for all analytes except 4 mg kg1 for melamine and

cyromazine) prepared in methanol, was spiked into 0.5 g of infant
milk formula and homogenised. The mixture was incubated for an
hour prior to extraction. 1 M hydrochloric acid or 2.5 M potassium
hydroxide was used to adjust the extraction solvent to the desired
pH. The appropriate volume of solvent was introduced into the
microwave vessel and the extraction was carried out at a xed
extraction temperature and time. The stirring speed was set at
600 rpm. After extraction, the samples were allowed to cool to
room temperature before opening the caps. The supernatant was
then centrifuged at 9000 rpm for 10 min at 5 C. During the optimisation study for the MAE conditions, higher concentration of pesticides was used to allow the direct LCMS/MS analysis of the
supernatant. After obtaining the optimal MAE condition, the supernatant was subjected to SPE for preconcentration of pesticides
prior to analysis by LCMS/MS.
2.4.2. Soxhlet extraction
The extraction condition was modied from the procedure reported by Song et al. (You et al., 2007). Approximately 1.0 g of
spiked sample was directly weighed in the cellulose thimble before
introducing into the Soxhlet extraction chamber. 20 mL of 0.1% of
water in methanol solvent was added and reuxed for 1 h. The
sample mixture was cooled to room temperature before passing
though the MCX SPE cartridge. It was then blown dried with argon
and reconstituted to 200 lL with a mixture of methanol and 0.1%
formic acid in water (6:4 v/v) through LCMS/MS analysis.
2.4.3. Solid phase extraction
The MAE was scaled up to three times (1.5 g of infant milk formula and 13.56 mL of extraction solvent) before passing through
an Oasis MCX SPE cartridge for preconcentration. The modied
procedure for the SPE was necessary to attain good extraction
recoveries (Papadopoulos, Gikas, Zalidis, & Tsarbopoulos,
2007).The sorbent was initially equilibrated with 5 mL of water before rinsing with 5 mL of methanol. 11.5 mL of supernatant collected after MAE was acidied with 1 mL of 1 M hydrochloric
acid. Subsequently, the sorbent was washed with 2 mL of 0.1 M
hydrochloric acid and methanol respectively. 10 mL of 4% ammonium hydroxide in methanol was utilised to elute the pesticides.
It was blow dried under a gentle stream of argon and reconstituted
to 200 lL with a mixture of methanol and 0.1% formic acid in water
(6:4 v/v).
2.5. Method validation for MAE/SPE
Matrix-matched standard calibration curves were prepared by
adding appropriate amounts of standards to the blank milk extracts after MAE/SPE. Ten standards with varying concentration
of 0.5600 ppb were prepared and triplicate analyses of the standards were conducted. The Masshunter quantication analysis
software (Waldbronn, Germany) integrated the peak area of the
MRM transitions for each analyte and plotted the calibration curve.
The linearity of the calibration curve was assessed with Mandels
tting test value while the correlation coefcients (r2) described

Table 1
Range of MAE parameters selected for optimisation for different solvent system.
Variables

Temperature (C)
Time (min)
Volume of extraction solvent (mL)
pH
Percentage of water in organic solvent (%)

Methanol/water mixture

Acetonitrile/water mixture

Optimised level (1, 0, 1)

Selected level

Optimised level (1, 0, 1)

Selected level

60, 90,120
10,20,30
2, 3.5, 5
2, 7, 12
0, 15, 30

102
20
4.56
12
0.1

60, 90,120
10,20,30
2, 3.5, 5
2, 7, 12
20, 25, 30

61
12
4.92
12
20.1

2476

G. Fang et al. / Food Chemistry 134 (2012) 24732480

the linear relationship between concentration and area. The standards of higher concentration which deviated from linearity were
removed, determining the working range of the calibration curve.
The limit of detection (LOD) and limit of quantication (LOQ) were
calculated based on signal-to-noise ratio of 3 and 10, respectively.
Method accuracy and precision were evaluated by recovery studies
using fortied samples at three concentration levels (5, 25, and
100 mg kg1).
3. Results and discussion
3.1. LCMS/MS optimisation
In order to achieve good sensitivity, 1 mg kg1 of individual pesticide standard was prepared in water for MS optimisation through
ow injection analysis. The [M + H]+ precursor ions and daughter
ions were determined with the use of full scan and MS/MS spectrum. Subsequently, the fragmentator (20200 V), collision energy
(540 V) and dwell time (50200 s) were optimised in order to obtain high signal to noise and sufcient data points during multiple
reactions monitoring (MRM). The quantiers are assigned to the
most abundant daughter ions employed for quantication whereas
the qualiers are usually the next most abundant daughter ions for
conrmatory purposes.
For the quantication of pesticides in infant milk formula, the
chromatographic separation was rst optimised. It has been reported that highly polar molecules, specically melamine and its
analogues, had insufcient retention on traditional octadecylsilane
(C18) column. However, longer retention time of approximately
13 min was observed for melamine when two C18 columns were
used in series (Ehling, Tefera, & Ho, 2007). Alternatively, ion pairing
agent such as sodium n-heptanesulfate was added to the mobile
phase to improve the retention of melamine on C18 column (Sun,
Wang, Ai, Liang, & Wu, 2010). Nevertheless, the use of non-volatile
ion pairing agent as aforementioned is not recommended for LC
MS/MS system as the presence of these ion pairing agents might
contaminate the electrospray source leading to ion suppression
(Hernandez, Sancho, & Pozo, 2002).
To solve this problem, FDA proposed the use of hydrophilic
interaction liquid chromatography (HILIC) (FDA, 2008). This technique utilises hydrophilic stationary phase such as silica and relatively hydrophobic mobile phases consisting of a mixture of 540%
water or a volatile buffer and acetonitrile. The hydrophilic stationary phase was able to retain the highly polar compounds, such as
melamine and cyromazine with a retention time of approximately
5.8 and 6.3 min respectively (Fig. 1). Conversely, the more hydrophobic pesticides eluted earlier in the order of prometryn, ametryn, prometon, simetryn, atrazine and atraton.
3.2. MAE optimisation
3.2.1. Selection and optimisation of experimental design
A central face composite design which is a type of central composite design, in which the axial points are placed on the face centres
of the cube, was chosen. Each variable has only three levels instead
of ve when compared to the central composite design. Thus, the total number of experiments required (N) = 2(k  1)+2k + C0 where k
represented the number of factors and C0 referred to the total number of centre points (Ferreira et al., 2007). During optimisation, the
ve variables including extraction temperature, time, volume of
extraction solvent, pH, composition of water in solvent and the
nature of solvent for the MAE were investigated. With this design,
3 centre runs were conducted to produce a reasonable stability of
prediction variance. Hence, a total of 29 randomised experiments
were executed.

The experimental domains of the variables were dened for the

experiment design taking into the account of the instrumental and
operative limits. For example, the extraction temperature utilised
is restricted by the conguration of the microwave system and
the solvent boiling point. With the use of closed vessel microwave
system, the solvent boiling point increased under pressure which
permitted the use of higher extraction temperature. Thus, the
extraction temperature was xed at 120 C. Higher temperature
was not selected to avoid excessive build-up of pressure in the
MAE vessel. High extraction temperature would be advantageous
to as it decreased the solvents surface tension and viscosity,
resulting in an increase in the solvents penetration in the infant
milk formula. Such solvent penetration disrupted the intermolecular interactions between the pesticides and infant milk and thereby
improved in the extraction efciency of pesticides (Eskilsson &
Bjorklund, 2000).
Besides temperature, the extraction time can affect the recoveries of the pesticides. With shorter irradiation time, lower extraction recoveries might occur due to the insufcient absorption of
microwave energy by the infant milk formula to disrupt the intermolecular interaction between the analytes and matrix (Shen &
Lee, 2003). In contrast, excessive extraction time might result in
degradation of pesticides.
The complete immersions of samples in extraction solvents
were also necessary for efcient extraction. Therefore, the volume
of extraction solvent was controlled within a range of 25 mL. Besides, varying pH from acidic to basic condition was essential for
high extraction recoveries as it affected the protonation of the
functional groups on the pesticides.
The nature of the solvent is a critical variable in MAE. A good
extraction solvent should selectively solubilise the pesticides, and
at the same time, has high dielectric constant to permit efcient absorption of microwave energy (Vivekananda, Yogesh, &
Hemalatha, 2007). Non-polar solvent such as hexane were not
selected as it has poor microwave absorptivity. Two sets of polar
binary solvent mixture methanol/water and acetonitrile/water
were thus chosen for the optimisation process. These solvent
systems were postulated to result in good extraction efciencies
with the relatively polar pesticides on the basis of like attract like
principles. The percentage of water in organic solvent was limited
to 30% as high percentage of organic solvent was required to
precipitate the protein present in infant milk formula.
3.2.2. Comparison of the extraction recoveries between different
solvent systems
The effects of the ve variables mentioned above were investigated with a total of 29 experiments conducted for each solvent
system methanol/water and acetonitrile/water solvents. Derringer desirability function was subsequently employed to calculate
the optimal extraction condition of each solvent system. As presented in Table 1, lower temperature, shorter extraction time and
higher percentage of water were favoured for the acetonitrile/
water mixture when compared to the optimal extraction condition
for the methanol/water mixture. However, improved recoveries of
the pesticides, especially for hydrophilic melamine was observed
for the methanol/water solvent mixture when compared to the
acetonitrile/water solvent mixture (Fig. 2). Hence, the following
section will focus on the effect of variables on the extraction recoveries of the methanol/water solvent mixture.
3.2.3. Effects of the variables on extraction recoveries for methanol/
water solvent mixture
The extraction recoveries data obtained were tted with a polynomial equation as followed which aided in relating the effects of
the variables to the recoveries of each pesticide. In addition, this

G. Fang et al. / Food Chemistry 134 (2012) 24732480

2477

Fig. 1. The chromatograms from the analysis of the pesticides in milk matrices.

equation can be used to investigate the second order interactions

between variables (Ferreira et al., 2007).

! b0

5
X
i1

bi xi

4 X
5
X
bij xi xj

i1 j1

where ! is the extraction recovery of pesticide, b0 is the constant, bi

and bij represents the individual and quadratic model coefcients,
and vi and vj the variables studied such as extraction time and
temperature.
Some of the quadratic model coefcients such as temperature  time, time  volume of solvent, time  pH and pH  % of
water were excluded to improve the tness of the models. R2
and Q2 were employed to provide the test of the t of the models.
R2 was the fraction of variation of the response explained by the
model and Q2 referred to the fraction of variation of the response
that could be predicted by the model. R2 referred to an overestimated while Q2 referred to an underestimated measure of the
goodness of t. A value of 0.5 or larger for Q2 indicated that the
model had good predictive ability and would have small prediction
errors (Eriksson, Johansson, Kettaneh-Wold, Wikstrm, & Wold,

2008). With the removal of some of the quadratic model coefcients as aforementioned, the R2 and the Q2 attained were greater
than 0.5, indicating good predictability. In addition, excellent
reproducibility was achieved based on the triplicate of the centre
points.
The normalised b coefcient values for Eq. (2) were displayed in
Table 2. The volume of extraction solvent and the pH of the extractant were observed to have more inuential effect on the extraction
recoveries for most compounds in this study at 95% condence level. The quadratic coefcient of volume of solvent  pH has positive
effect to the extraction recoveries of most of the pesticides at 5%
signicance level. This indicated that the extraction recoveries increased further for most pesticides when the pH was basic and
with large volume of extraction solvent as compared to basic pH
with small volume of extractant.
The positive b coefcient values of the volume of extraction solvent indicated that increasing the volume of extraction solvent
would improve the extraction recoveries for most pesticides. In
addition, extraction temperature and time had a positive inuence
on the recoveries of most analytes except melamine and cyromazine. At high temperature and longer extraction time, the

2478

G. Fang et al. / Food Chemistry 134 (2012) 24732480

Fig. 2. Comparison of the extraction recoveries attained from MAE from different solvent mixture.

Table 2
Normalised b coefcient plots for the pesticides which were studied with the experimental design during MAE optimisation.
Prometryn
Constant
Temperature (temp)
Time
Volume of solvent (vol)
pH
% water
Temp  Temp
Time  Time
vol  vol
pH  pH
% water  % water
Temp  vol
Temp  pH
Temp  % water
Time  % water
vol  pH
vol  % water
R2
Q2
Reproducibility
*

3.773
0.161
0.088
0.431*
0.457*
0.068
0.124
0.289
0.595*
0.102
0.310
0.072
0.144
0.113
0.143
0.228*
0.122
0.942
0.681
0.96

Ametryn
*

4.156
0.185*
0.082
0.345*
0.343*
0.092
0.150
0.247
0.600*
0.334*
0.323*
0.108
0.113
0.127*
0.164*
0.229*
0.079
0.96
0.73
1

Prometon
*

4.017
0.216*
0.064
0.326*
0.330*
0.142
0.104
0.271
0.761*
0.301
0.519*
0.138
0.120
0.115
0.188*
0.247*
0.047
0.926
0.602
0.99

Atrazine
*

3.441
0.084
0.124
0.245*
0.201
0.066
0.247
0.014
0.091
0.601
0.311
0.038
0.042
0.157
0.082
0.129
0.127
0.886
0.698
0.953

Simetryn
*

4.303
0.165*
0.069
0.521*
0.449*
0.114
0.114
0.275*
0.592*
0.113
0.385*
0.083
0.147
0.104
0.159*
0.247*
0.078
0.963
0.721
0.983

Atraton
*

4.355
0.236*
0.063
0.396*
0.318*
0.167
0.109
0.290
0.791*
0.260
0.570*
0.150*
0.133
0.073
0.191*
0.231*
0.038
0.939
0.599
0.987

Cyromazine
*

1.171
0.054
0.066
0.011*
0.356
0.146
0.248
0.267
0.895*
0.512
0.851*
0.035
0.088
0.096
0.107
0.223
0.028
0.661
0.506
1

Melamine
0.652*
0.023
0.092
0.456*
0.446*
0.188
0.165
0.250
0.566*
0.603*
0.640*
0.026
0.110
0.055
0.082
0.233*
0.082
0.866
0.564
1

Signicant at 95% condence interval.

hydrolysis of melamine and cyromazine might occur, hence, leading to lower extraction recoveries (Ono et al., 1998).
The pH of the extractant is also one of the most inuential variables for MAE. With the b coefcient values tted to the Eq. (2), response contours plots could be generated to clearly visualise the
inuence of each variable on the extraction recoveries (Supplementary Fig. 1). Since the pKa of pesticides ranged from 2.35 to
5.27, the amine group on the pesticides would remain in its native
state at high pH, thereby, resulting in higher preference for solvent
mixture consisting of less percentage of water. Therefore, basic pH
condition was preferred for the extraction of most pesticides except atrazine. Since atrazine hydrolysed easily in acidic and basic
aquatic environment (Kamrin, 1997), higher extraction recovery
of atrazine could only be achieved at neutral pH with methanol
consisting of 222% of water.
The negative b coefcient values implied that the addition of
water to methanol resulted in poor recoveries for most of the pesticides. One possible explanation is due to the dissipation factor of
the solvent which comprised of dielectric loss and dielectric

constant. Methanol has low dielectric constant but higher dielectric loss compared to water. Thus, it absorbs microwave energy
poorly but dissipates microwave energy to thermal energy preferentially. Consequently, rapid heating can occur instantaneously,
resulting in higher extraction efciency for methanol as compared
to water (Eskilsson & Bjorklund, 2000).
3.2.4. Solid phase extraction
With the optimal MAE condition determined for methanol/
water solvent mixture, the MAE supernatant was subjected to
SPE for pre-concentration of pesticides and removal of interferences. The supernatant obtained from MAE was rst acidied with
hydrochloric acid to protonate the amine moieties and to enhance
its afnity for the anions on the sorbent. Oasis MCX sorbent which
consisted of sulphonate functional groups was selected instead of
octadecylsilane as it had more selectivity towards the protonated
amine moieties on the pesticides. The cartridge was pre-conditioned with water and methanol. After loading the sample, it was
washed with hydrochloric acid to further promote the binding

2479

G. Fang et al. / Food Chemistry 134 (2012) 24732480

for the extraction of pesticides as less solvent consumption and

extraction time was required compared to Soxhlet extraction.

between the pesticides and the negatively charged sorbent. Methanol was subsequently added to remove any hydrophobic interference. Basic eluent was added to deprotonate the ionised amine
group and reduce the pesticides afnity for the sorbent such that
high recoveries were achieved. Hence, 10 mL of 4% ammonium
hydroxide in methanol was added to ensure complete elution of
the pesticides from the sorbent.

3.4. Method validation for MAE/SPE

For quantitative analysis via LCMS/MS, the presence of matrix
components can affect the ionisation efciency of the target compounds by reducing or enhancing the response compared with the
standards. In this work, analyte responses were suppressed compared to signal of standards prepared in water. To resolve this issue, matrix-matched standards were analysed. Triplicates
analyses of the pesticides were performed over a concentration
range of 0.5600 lg kg1 (Table 4).
Mandel test was implemented to determine the linearity of the
calibration curves. All Mandel test values did not exceed the critical F values at 99% condence level. These demonstrated that the
linear regression analyses provided a better description of the relationship between area and concentration of all the analytes. The
slope and the intercepts were calculated for each curve by applying
linear regression analysis. The responses of all compounds were
correlated with r2 values of close to 1. The limits of detection
(LOD) and limits of quantication (LOQ) were extrapolated based
on S/N = 3 and S/N = 10 correspondingly. The LOD and LOQ of pesticides ranged from 0.12 to 2.53 lg kg1 and 0.41 to 8.42 lg kg1,
respectively.
Precision was assessed by repeated injections of a spiked MAE
extract containing between 100 and 200 lg kg1 of the pesticides
within the same day (intra-day precision, n = 3) and on different
days (inter-day precision, n = 3). The RSD values across the various
concentrations were less than 4.5% for intra-day precision analysis
whereas the inter-day precision ranged from 0.56% to 13%.
In order to evaluate the performance of the proposed method, a
series of experiments were conducted with varying concentrations
of pesticides spiked into infant milk formula. The MAE process was
scaled up three times, utilising 1.5 g of milk powder and 13.68 mL
of solvent such that trace levels of pesticides could be detected.
Subsequently, the supernatant obtained from MAE was passed

3.3. Comparison between MAE and Soxhlet extraction when coupled to

SPE
The extraction recoveries of the MAE/SPE and the Soxhlet/SPE
were evaluated statistically using the Students t-test at the 95%
condence level to evaluate the applicability of the proposed
method for the determination of pesticides in infant milk formula.
The extraction recoveries data were compared to those obtained
for Soxhlet/SPE as shown in Table 3. The extraction recoveries for
the two methods were signicantly different since the calculated
t value (5.10) was greater than the tabulated t values (2.37). At
the fortication level of 25 lg kg1 of pesticides, the recoveries
were between 8798% and 3575% for MAE/SPE and Soxhlet/SPE
respectively. This clearly demonstrated the suitability and the feasibility of MAE/SPE as an improvised extraction method for these
selected pesticides from infant milk formula.
It was possible to attain higher extraction recoveries of pesticides within a shorter period of time when MAE was coupled to
SPE. This might be due to different heating mechanism between
the MAE and Soxhlet extraction. Soxhlet extraction is a continuous
solidliquid extraction such that the solid are repeatedly extracted
with fresh solvents. The vessel was heated rst followed by the heat
transfer to the sample and solvent (conduction and convection
phenomena) (Vivekananda et al., 2007). However, MAE utilises
microwave energy to heat the polar matrix and solvent rst instead
of the vessel. Therefore, this accelerates the heating rate of the
extraction temperature which in turn affects the extraction efciency and time required. Besides, better precision was also
attained for MAE. Thus the proposed method was more suitable

Table 3
The extraction recoveries of three different concentrations of spiked pesticides in infant milk formula.
MAE/SPE

Soxhlet/SPE

5 lg kg1

Prometryn
Ametryn
Prometon
Atrazine
Simetryn
Atraton
Cyromazine
Melamine

25 lg kg1

100 lg kg1

25 lg kg1

Average n = 3

% RSD

Average n = 3

% RSD

Average n = 3

% RSD

Average n = 3

% RSD

88
90
88
73
91
91
78
72

5
5
2
5
8
3
3
4

98
95
87
90
92
95
76
87

4
2
3
7
0
5
7
6

101
101
111
85
101
106
94
91

4
4
4
3
4
4
6
7

75
80
66
57
76
82
62
35

2
2
4
2
2
1
7
13

Table 4
Calibration data, which included precisions and limit of detection and quantication.

Prometryn
Ametryn
Prometon
Atrazine
Simetryn
Atraton
Cyromazine
Melamine

Number of
standards

Mandel
test

Fcritical
(P = 99%)

r2

Linear
range (ppb)

Slope

Intercept

Intraday precision,
RSD% n = 3

Interday precision,
RSD% n = 6

LOD
(ppb)

LOQ
(ppb)

7
9
8
10
7
9
7
6

1.24
0.19
0.50
0.07
1.16
3.89
8.33
9.69

15.52
8.26
10.67
6.84
15.52
8.26
15.52
28.71

0.998
0.999
0.999
0.999
0.998
0.999
0.999
0.999

0.5250
0.1250
2.5600
0.5600
1.0250
0.5600
2.5500
7.5600

10,686
2424.3
12,477
1268.3
4068.4
17,726
548.3
120

3324.1
140.12
25,863
2486.9
2169.4
15,256
2387.9
527.58

0.26
0.37
0.93
0.44
0.28
0.30
0.17
0.82

6.73
2.20
12.52
0.66
1.23
11.95
5.08
0.56

0.14
0.12
0.69
1.23
0.28
0.55
2.53
1.61

0.46
0.41
2.30
4.11
0.93
1.85
8.42
5.36

2480

G. Fang et al. / Food Chemistry 134 (2012) 24732480

through SPE for pre-concentration. The extraction recoveries of the

pesticides with the optimal MAE condition were shown in Table 3.
The recovery values for the pesticides were calculated by comparison of analyte responses obtained from the samples spiked before
and after the extraction at the same concentration level. Triplicate
analyses of each recovery test were carried out. The recoveries ranged from 72% to 111% with RSD of less than 8%. The EU stated that
the acceptability of analytical method performance is that the
analytical method developed for residue pesticides in food should
be demonstrated at validation as being capable of providing mean
recovery values at each spiking (fortication) level and for each
representative commodity within the range 70120%, with a
repeatability RSD 6 20%, for all compounds to be sought using
the method (EU, 2007). Since the recoveries values were well
within the acceptable criteria for residues analysis, the proposed
method would be suitable for the determination of pesticides in infant milk formula.
4. Conclusion
MAE coupled with SPE was successfully developed to extract
pesticides of different polarities from infant milk formula. Experimental design together with Derringer desirability function was
rst utilised to determine the optimal MAE conditions before
pre-concentration of the samples with Oasis MCX SPE cartridges.
During the quantication with LCMS/MS, the column was selected to ensure that the analytes had sufcient retention on column. Recoveries ranging from 72% to 111% were obtained for all
compounds at fortication levels (5100 lg/kg) with RSD 6 8%.
Better extraction efciencies were achieved with the proposed
method compared to Soxhlet extraction coupled to SPE. Moreover,
the coupled MAE/SPE used less time and solvent compared to the
coupled Soxhlet/SPE method.
Acknowledgements
We acknowledge nancial support from the National University
of Singapore (NUS), NUS Environmental Research Institute (NERI),
National Research Foundation (NRF) and Economic Development
Board (EDB) (SPORE, COY-15-EWI-RCFSA/N197-1), Ministry of
Education (R-143-000-382-112, and R-143-000-416-232, and R143-000-441-112), A-STAR (SERC PSF 052 101 0044), Ministry of
Defence (R-143-000-416-646) and Environment and Water Industry Development Council (0601-IRIS-093-08).
Appendix A. Supplementary material
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.foodchem.2012.
04.076.
References
Andreu, V., & Pico, Y. (2004). Determination of pesticides and their degradation
products in soil: Critical review and comparison of methods. TRAC Trends in
Analytical Chemistry, 23(1011), 772789.
Bianchi, F., & Careri, M. (2008). Experimental design techniques for optimization of
analytical methods. Part I: Separation and sample preparation techniques.
Current Analytical Chemistry, 4(1), 5574.
Chevrier, C., Limon, G., Monfort, C., Rouget, F., Garlantezec, R., Petit, C., et al. (2011).
Urinary biomarkers of prenatal atrazine exposure and adverse birth outcomes
in the PELAGIE birth cohort. Environmental Health Perspectives, 119(7),
10341041.
Derringer, G., & Suich, R. (1980). Simultaneous-optimization of several response
variables. Journal of Quality Technology, 12(4), 214219.
Ehling, S., Tefera, S., & Ho, I. P. (2007). High-performance liquid chromatographic
method for the simultaneous detection of the adulteration of cereal ours with
melamine and related triazine by-products ammeline, ammelide, and cyanuric
acid. Food Additives and Contaminants, 24(12), 13191325.

EPA (1999). The grouping of a series of triazine pesticides based on a common

mechanism of toxicity. Available from http://www.regulations.gov (accessed
30.04.12).
Eriksson, L., Johansson, E., Kettaneh-Wold, N., Wikstrm, C., Wold, S. (2008). Design
of experiments: Principles and applications. Umterics.
Eskilsson, C. S., & Bjorklund, E. (2000). Analytical-scale microwave-assisted
extraction. Journal of Chromatography A, 902(1), 227250.
EU (2006). Commission Directive 2006/125/EC of 5 December 2006 limiting the
concentration of pesticides in processed cereal-based foods and baby foods for
infants and young children.
EU (2007). Document No. SANCO/2007/3131 method validation and quality control
procedures for pesticide residues analysis in food and feed.
EU (2009). Commission regulation (EC) No. 1135/2009 Of 26 November 2009
laying down the concentration of melamine permitted in infant milk products.
FDA (1996). Food quality act which state the tolerance levels enforced for pesticides in
food products. Available from http://www.epa.gov/pesticides/regulating/laws/
fqpa/backgrnd.htm (accessed 30.04.12).
FDA (2008). Determination of melamine and cyanuric acid residues in infant formula
using LCMS/MS. Available from http://www.fda.gov/Food/ScienceResearch/
LaboratoryMethods/DrugChemicalResiduesMethodology/ucm071637.htm
(accessed 30.04.12).
Ferreira, S. L. C., Bruns, R. E., da Silva, E. G. P., dos Santos, W. N. L., Quintella, C. M.,
David, J. M., et al. (2007). Statistical designs and response surface techniques for
the optimization of chromatographic systems. Journal of Chromatography A,
1158(12), 214.
Hapeman, C. J., Bilboulian, S., Anderson, B. G., & Torrents, A. (1998). Structural
inuences of low-molecular-weight dissolved organic carbon mimics on the
photolytic fate of atrazine. Environmental Toxicology and Chemistry, 17(6),
975981.
Hernandez, F., Sancho, J. V., & Pozo, O. J. (2002). Direct determination of alkyl
phosphates in human urine by liquid chromatography/electrospray tandem
mass spectrometry. Rapid Communications in Mass Spectrometry, 16(18),
17661773.
Ihunegbo, F. N., Tesfalidet, S., & Jiang, W. (2010). Determination of melamine in milk
powder using zwitterionic HILIC stationary phase with UV detection. Journal of
Separation Science, 33(67), 988995.
Kamrin, M. A. (1997). Triazines and triazoles. Pesticide proles toxicity, environmental
impact, and fate. CRC Press.
Mezcua, M., Repetti, M. R., Aguera, A., Ferrer, C., Garcia-Reyes, J. F., & FernandezAlba, A. R. (2007). Determination of pesticides in milk-based infant formulas by
pressurized liquid extraction followed by gas chromatography tandem mass
spectrometry. Analytical and Bioanalytical Chemistry, 389(6), 18331840.
Ono, S., Funato, T., Inoue, Y., Munechika, T., Yoshimura, T., Morita, H., et al. (1998).
Determination of melamine derivatives, melame, meleme, ammeline and
ammelide by high-performance cation-exchange chromatography. Journal of
Chromatography A, 815(2), 197204.
Papadopoulos, N., Gikas, E., Zalidis, G., & Tsarbopoulos, A. (2007). Simultaneous
determination of terbuthylazine and its major hydroxy and dealkylated
metabolites in wetland water samples using solid-phase extraction and highperformance liquid chromatography with diode-array detection. Journal of
Agricultural and Food Chemistry, 55(18), 72707277.
Pawliszyn, J. (2002). Sampling and sample preparation for eld and laboratory.
Elsevier Science.
Schantz, M. M. (2006). Pressurized liquid extraction in environmental analysis.
Analytical and Bioanalytical Chemistry, 386(4), 10431047.
Shen, G., & Lee, H. K. (2003). Determination of triazines in soil by microwaveassisted extraction followed by solid-phase microextraction and gas
chromatographymass spectrometry. Journal of Chromatography A, 985(12),
167174.
Srogi, K. (2006). A review: Application of microwave techniques for environmental
analytical chemistry. Analytical Letters, 39(7), 12611288.
Stout, S. J., daCunha, A. R., & Allardice, D. G. (1996). Microwave-assisted extraction
coupled with gas chromatography electron capture negative chemical
ionization mass spectrometry for the simplied determination of
imidazolinone herbicides in soil at the ppb level. Analytical Chemistry, 68(4),
653658.
Sun, H. W., Wang, L. X., Ai, L. F., Liang, S. X., & Wu, H. (2010). A sensitive and
validated method for determination of melamine residue in liquid milk by
reversed phase high-performance liquid chromatography with solid-phase
extraction. Food Control, 21(5), 686691.
Sun, H. W., Wang, L. X., Liu, N., Qiao, F. X., & Liang, S. X. (2009). SPE then RP-LC for
simultaneous analysis of cyromazine and its metabolite melamine in liquid
milk and egg. Chromatographia, 70(1112), 16851689.
Vivekananda, M., Yogesh, M., & Hemalatha, S. (2007). Microwave assisted extraction
An innovative and promising extraction tool for medicinal plant research.
Pharmacognosy Reviews, 1(1), 718.
Wang, W. T., Meng, B. J., Lu, X. X., Liu, Y., & Tao, S. (2007). Extraction of polycyclic
aromatic hydrocarbons and organochlorine pesticides from soils: A comparison
between Soxhlet extraction, microwave-assisted extraction and accelerated
solvent extraction techniques. Analytica Chimica Acta, 602(2), 211222.
You, J., Zhang, H., Yu, A., Xiao, T., Wang, Y., & Song, D. (2007). Determination of
triazines in infant nutrient cereal-based foods by pressurized microwaveassisted extraction coupled with high-performance liquid chromatography
mass spectrometry. Journal of Chromatography B-Analytical Technologies in the
Biomedical and Life Sciences, 856(12), 278284.