Contents lists available at SciVerse ScienceDirect
Journal of Chromatography A
jou rn al h om epage: www.elsev ier .com/ locat e/chroma
ultiresidue determination of 375 organic contaminants including pesticides,olychlorinated biphenyls and polyaromatic hydrocarbons in fruits andegetables by gas chromatography–triple quadrupole mass spectrometry withntroduction of semi-quantification approach
A residue analysis method for the simultaneous estimation of 349 pesticides, 11 PCBs and 15 PAHsextracted from grape, pomegranate, okra, tomato and onion matrices, was established by using a gaschromatograph coupled to an electron impact ionization triple quadrupole mass spectrometer (GC–EI-MS/MS). The samples were extracted by ethyl acetate and cleaned by dispersive solid phase extractionwith PSA and/or GCB/C18 by the methods reported earlier. The GC–EI-MS/MS parameters were optimizedfor analysis of all the 375 compounds within a 40 min run time with limit of quantification for most of thecompounds at <10 �g/L, which is well below their respective European Union-Maximum Residue Levels.The coefficient of determination (r2) was >0.99 within the calibration linearity range of <5–250 ng/mLfor compounds with LOQs < 5 ng/mL. While for the compounds with LOQs within 5–10 �g/kg, the low-est calibration level was 5 and 10 �g/kg as applicable. The recoveries at 10, 25 and 50 ng/mL werewithin 70–110% (n = 6) with associated RSDs < 20% indicating satisfactory precision. The information
generated from the single laboratory validation was further utilized for building a semi-quantitativeapproach. The accuracies in quantification obtained via individual calibration standards vis-à-vis semi-quantification approach were comparable. For incurred samples, the concentrations estimated by thesemi-quantification approach were within ±10% of the values obtained by direct quantification. Thisapproach complements the existing GC–EI-MS/MS methods by offering targeted screening and quantifi-
cation capabilities.
. Introduction
India is a habitat of plant genetic diversity. With its currentroduction of around 32 million MT, India accounts for about 8%f the world’s total fruit production. India also has the credit ofeing the second largest producer of vegetables in the world andccounts for about 15% of the world’s total production. Consideringhe high pest and disease pressure, the multitude of agrochemicalssed for plant protection in India is diverse. Currently, 230 plantrotection products (PPP) are registered for agricultural use [1] in
ndia with more than 820 compounds being in schedule for intro-uction into Indian market in due course of time. Moreover, everyear the agrochemical industries keep introducing newer PPPs in
the Indian market targeting management of various crop and pestcombinations. Although a limited number of pesticides might berecommended for use in any specific crop, there are possibilities oftransmission of non-recommended pesticide residues from adjoin-ing farms where other crops are cultivated with a different setof recommended pesticides being sprayed on them. Additionally,the residues of persistent organic pollutants like polychlorinatedbiphenyls (PCB) and polyaromatic hydrocarbons (PAH) could findtheir ways into the food chain through various sources, e.g. sur-face deposition, etc. necessitating simultaneous monitoring of theresidues of pesticides, PCBs and PAHs in crops for holistic riskassessment.
A preliminary assessment reveals that among the approxi-mately 450 pesticides for which maximum residue limits (MRLs)
are currently set [2] in the European Union (EU) on various agri-cultural commodities, more than 300 compounds are amenablefor analysis by GC–EI-MS/MS. Several studies have been reportedfor targeted analysis of multiclass, multiresidue compounds in a
ariety of fruits and vegetables by GC using single quadrupole [3],on trap [4], and triple quadrupole mass analyzers [5,6].
In general, low-energy collision induced dissociation tandemass spectrometry analysis (CID-MS/MS) using the multiple reac-
ion monitoring (MRM) scan mode is used for the identificationnd quantification of a target list of compound residues. The appli-ation, scope and success of such methods essentially require thevailability of certified reference standards. To obtain a compre-ensive knowledge on the food safety status of any sample withnknown history of contamination, a full scan analysis based onlemental composition and accurate mass (as offered by time-of-ight mass spectrometry) could be required. However, high costsnd the complexity of data processing related to application of highesolution GC–MS limits its usage in routine residue analysis. Multi-le benefits could be accrued from a high throughput multi-residueethod targeting a large number of analytes by a single GC–EI-S/MS run covering all probable compounds that could appear in
ruits and vegetables from direct as well as indirect sources. Datacquisition methods comprising a large number of MRM transi-ions as described in this paper can be applied for the detectionnd quantification of a target list of analytes for which the referencetandards are available. In addition, it can also offer the benefits ofualitative analysis and semi-quantification of those compoundsor which reference standards are not available, on the basis ofheir compound-specific quantitative and qualitative MRM tran-itions, their abundance ratio and application of the calibration ofompounds with similar GC–MS/MS responses.
To evaluate the practical applicability of the above discussionver a range of compounds, a fast and sensitive method based onthyl acetate extraction and estimation by GC–EI-MS/MS was vali-ated for analysis of 375 compounds including pesticides, PAHs andCBs in fruits viz., grapes, pomegranate and vegetables viz., onion,kra and tomato. The method was employed to generate a databaseonsisting of target compound name, quantifier and qualifier MRMransitions, and the slopes of calibration curves from which rel-tive ratios were calculated and applied for semi-quantificationf the detected residues. Our aim was to evaluate the efficiencyf the semi-quantitative approach with reasonable accuracy andonsistency.
. Experimental
.1. Chemicals
The solvents, viz. ethyl acetate and acetonitrile, were of residuenalysis grade and purchased from Thomas Baker (Mumbai, India).eagent-grade anhydrous sodium sulfate was purchased fromerck (Mumbai, India). The QuEChERS extraction tubes containing
g magnesium sulfate and 1 g sodium chloride were procured fromgilent Technologies (Bangalore, India). The bulk sorbents, PSA
primary secondary amine) bonded silica (C18, 100 g) and graphi-ized carbon black (GCB) were supplied by Agilent TechnologiesBangalore, India).
The standards of all the test compounds (Table 1) werebtained from Dr. Ehrenstorfer GmbH (Augsburg, Germany) andigma–Aldrich (Saint Louis, USA).
.2. Apparatus
The analysis of samples was performed using an AgilentC (7890A) equipped with a CTC Combipal (CTC Analytics,
witzerland) autosampler, connected to a triple quadrupole masspectrometer (7000B, Agilent Technologies, Santa Clara, USA).he system was controlled using MassHunter software (ver.05.00.412). The analytical separation was performed using two
. A 1270 (2012) 283– 295
HP-5MS (15 m × 0.25 mm, 0.25 �m) capillary columns with mid-point backflush set up. During backflush the inlet pressure wasmaintained at 2 psi whereas the backflush pressure was 35.322 psiand backflush flow to the inlet was 3.6 mL/min for which additionalhelium flow was supplied through a purged ultimate union. Thebackflush was carried out for 2.5 min after the completion of theanalytical run. The column oven temperature during this periodwas maintained at 300 ◦C. A gooseneck liner (78.5 mm × 6.5 mm,4 mm) from Agilent Technologies (Santa Clara, USA) was used withhelium as carrier gas set at constant flow rate of 1.2 mL/min. Theoven temperature program was set as follows: initial temperatureof 70 ◦C (1 min hold), ramped to 150 ◦C at 25 ◦C/min (0 min hold),then at 3 ◦C/min up to 200 ◦C (hold 0 min) and finally to 285 ◦C at8 ◦C/min (8 min hold) resulting in a total run time of 39.49 min. Thetransfer line temperature was maintained at 285 ◦C.
The multi-mode inlet (MMI) was operated in solvent vent modefor large volume injection and 5 �L of sample was injected. Theprogrammable temperature vaporizer (PTV) was set at the initialtemperature of 70 ◦C (0.07 min hold), raised to 87 ◦C at 50 ◦C/min(0.1 min hold) followed by rapid heating at 700 ◦C/min up to 280 ◦C(3 min hold). The purge flow to solvent vent was set at 50 mL/min,2.7 min after injection and vent flow was maintained at 50 mL/minuntil 0.17 min.
The mass spectrometer was operated in MRM mode with acqui-sition starting from 4.4 min. The electron impact ionization (EI+)was achieved at 70 eV and the ion source temperature was set at280 ◦C. The specific MRM transitions for all the test compounds andother parameters are given in Table 1.
2.3. Standard preparation and calibration
Stock standard solutions of each compound were prepared byweighing 10 ± 0.1 mg and dissolution in 10 mL ethyl acetate andstored in amber colored glass vials at −20 ◦C. A total of seven inter-mediate mixtures (containing 50–60 compounds each) of 10 mg/Lconcentration were prepared by diluting adequate quantity ofeach compound in ethyl acetate. A working standard solution(1 mg/L) was prepared by mixing adequate quantity of interme-diate standard solution and dilution with ethyl acetate and storedat −20 ◦C. The calibration standards at 2.5, 5, 10, 20, 40, 80 and160 �g/L were freshly prepared for construction of the calibrationcurves.
The calibration graphs (seven points in triplicates) for all thecompounds were obtained by plotting the individual peak areasagainst the concentration of the corresponding calibration stan-dards. Matrix-matched standards at the same concentrations weresimultaneously prepared using pre-tested, residue free, organicallygrown matrix of grape, pomegranate, okra, tomato and onion. Toevaluate the matrix influence in terms of suppression or enhance-ment of analyte signals, the slopes of the matrix calibration graphfor each analyte was divided by its corresponding solvent standardand the ratios were compared.
2.4. Sample preparation
The samples (2 kg each) of grape, onion, okra and tomato wereblended directly in a mixer-grinder while pomegranate sampleswere blended after adding water (1:1, v/v) using the proceduredescribed in earlier publications [7]. From the crushed material,10 ± 0.1 g of the sample (15 ± 0.1 g for crushed pomegranate) wastransferred to 50 mL centrifuge tubes and extracted with 10 mLethyl acetate in the presence of 10 g sodium sulfate, followed
by homogenization at 10,000 rpm for 2 min using high speedhomogenizer (Heidolph, Germany) and centrifugation (3000 rpm,5 min). Dispersive solid phase extraction (DSPE) cleanup of thesupernatant (1 mL) was performed using 25 mg PSA and 7 mg GCB
K. Banerjee et al. / J. Chromatogr. A 1270 (2012) 283– 295 285
Table 1MRM transitions and other parameters for the test compounds.
a RT, retention time.b TS, time segment.c Q1, quantifier mass transition.d CE1, collision energy corresponding to Q1.e Q2, qualifier mass transition.f CE2, collision energy corresponding to Q2.
2 atogr
afouw
2
gwqntq
2
i0taciiotna
RpPRfsam
f
R
2
pstp(ae(
U
aist
2
mw
90 K. Banerjee et al. / J. Chrom
nd centrifuged (10,000 rpm, 5 min) to obtain a clear supernatantrom which 5 �L was injected into GC–EI-MS/MS. In the case ofnion, 800 �L of the supernatant was evaporated to near drynessnder gentle flow of nitrogen (5 psi) and reconstituted up to 800 �Lith ethyl acetate and 5 �L was injected into GC–EI-MS/MS.
.5. Validation data analysis and statistical calculations
The analytical method validation was carried out using SANCOuidelines (SANCO/12495/2011) [8]. The sensitivity of the methodas evaluated in terms of limit of detection (LOD) and limit of
uantification (LOQ). LOD is the concentration at which the sig-al to noise ratio (S/N) for the quantifier ion is ≥3, whereas, LOQ ishe concentration at which S/N of the quantifier MRM is ≥10 andualifier MRM ≥3.
.5.1. Precision and accuracyThe recovery experiment was carried out in replicates (n = 6)
n all the tested matrices at three different concentration levels of.005, 0.01 and 0.025 mg/kg. The samples were fortified with mix-ure of all the compounds and extracted by the method describedbove. The quantification was carried out using matrix matchedalibration standards. The precision in the conditions of repeatabil-ty (three analysts prepared six samples each on a single day) andntermediate precision (three analysts prepared six samples eachn six different days) were determined separately at the fortifica-ion level of 0.01 mg/kg. Since Horwitz ratio (HorRat) [9,10] wasot applicable at this concentration the Thompson equation waspplied [9].
Precision RSDR (reproducibility) for 1 to 120 ng/g is expressed bySDR = 22.0 (for C ≤ 120 �g/kg or c ≤ 120 × 10−9), and the maximumermitted value of observed RSD for reproducibility is 2 × RSDR.recision RSDr (repeatability) for 1–120 ng/g is expressed as 0.66SDR = 0.66 × 22. The maximum permitted value of observed RSD
or repeatability is 2 × RSDr. These equations are generalized preci-ion equations, which have been found to be independent of analytend matrix but solely dependent on concentration for most routineethods of analysis.The accuracy in terms of percent recovery was calculated by the
ollowing equation:
ecovery (%) = peak area of pre-extraction spikepeak area of postextraction spike
× 100
.5.2. Assessment of uncertaintyThe combined uncertainty was assessed as per the statistical
rocedure described in EURACHEM/CITAC Guide CG 4 [11] in theame way as reported earlier [12,13]. Uncertainty associated withhe calibration graph (U1), day-wise uncertainty associated withrecision (U2), analyst-wise uncertainty associated with precisionU3), day-wise uncertainty associated with accuracy/bias (U4), andnalyst-wise uncertainty associated with accuracy/bias (U5) wasvaluated for all the test compounds. The combined uncertaintyU) was calculated as
=√
U21 + U2
2 + U23 + U2
4 + U25
nd reported in relative measures as expanded uncertainty whichs twice the value of the combined uncertainty. Relative uncertaintytands for the ratio of uncertainty value at a given concentration tohe concentration at which the uncertainty is calculated.
.5.3. Data analysisThe validation carried out for 375 compounds in 5 different
atrixes resulted in a huge volume of data. An MS Excel macroas developed and applied for analysis of data.
. A 1270 (2012) 283– 295
2.6. Semi-quantitative approach for determination of residues
The developed method was employed to generate a databaseconsisting of the compound name, MRM transitions, and the peakareas of the quantifier ion of each compound. For developmentof the database repetitive injections (n = 20) of solvent based andmatrix matched calibration standards were performed. The peakareas obtained for each analyte from a specific set of transitionswere noted and the peak area ratios obtained along with therespective standard deviations. The mean ratio from the set of20 matrix matched standards was then applied for the quantifi-cation of residues in recovery samples from the same and differentbatches. The precision and accuracy in quantification of the residuesof any compound using the calibrations of other compounds vis-à-vis its own calibration were evaluated. Initially, the dataset wasgenerated for around 95 compounds routinely monitored in Indiangrape samples. Based on the success of the conversion factors gen-erated for 95 compounds, a database comprising of 375 analyteswas subsequently generated.
2.6.1. Approach for calculation of conversion factor forsemi-quantification
Assuming that the multiresidue mixture consists of the chem-icals (1, 2, 3, . . ., n) having peak areas of P1, P2, P3, . . ., Pn, ata particular concentration level, the ratios were calculated as:P2−1 = P1/P2, P3−1 = P1/P3, P3−2 = P2/P3, for the (n(n − 1))/2 numberof combinations, where “n” is the total number of analytes. Fromthe replicate ratios (20 replicates) generated for each combination,the average and the RSDs were calculated. For compound ‘1’ and‘2’, at a concentration of ‘C’ with peak areas of P1 and P2,
P1 = m1C + A1 (1)
and
P2 = m2C + A2 (2)
where m1 and m2 are the slopes of each calibration curve withintercepts A1 and A2. The ratio thus would be
P2−1 = P1
P2= m1C + A1
m2C + A2(3)
Assuming a real situation where the compound ‘2’ has peak areaof P′
2 and the calibration for compound ‘2’ is unavailable, the actualpeak area from ‘2’ is converted to the equivalent peak area obtainedfrom the compound ‘1’ (say P′
1) with the help of Eq. (3). Thus, P ′1 =
P2−1 × P ′2. Applying this to Eq. (1), the equivalelnt concentration =
((P2−1 × P′2) − A1)/m1. For most practical situations, the intercept
(A1) � the slope of the calibration curve (m1). Therefore, ((P2−1 ×P ′
2) − A1)/m1 ∼= (P2−1 × P ′2)/m1. Also, Eq. (3) could be expressed as
P2−1 = P1/P2 ∼= m1/m2. Thus, the equivalent concentration is approx-imately equal to (P2−1 × P ′
2)/m1 = P ′2/m2 = C ′
2, which is the actualconcentration. Thus the ratio of peak areas was used as the con-version factor for semi-quantification (examples demonstrated inSupplementary material S1).
2.7. Application of method for analysis of incurred samples
The reproducibility of the method was confirmed by analyzingthe incurred samples at two laboratories (National Research Cen-tre for Grapes, Pune and Agilent Technologies, Bangalore). Around10 incurred samples of each commodity were analyzed using the
validated method described above and quantified by both thequantitative and semi-quantitative approach. The samples werecollected from the local markets and supermarkets in the city ofPune and Bangalore.
atogr
3
3
essmaqrtwmos(t<eactcgfngtsiacaw
3
efshc((aIanousrowoieaopGat
K. Banerjee et al. / J. Chrom
. Results and discussion
.1. Optimization of instrumental conditions
Since a large numbers of analytes (375 numbers) were consid-red in this study, the chromatographic separation and the masspectrometric conditions played a vital role in determining theelectivity and sensitivity of the analysis. Now-a-days most instru-ent vendors supply a database of MRM transitions that could be
pplied to analyze a large list of compounds. The new generationuadrupole instruments are supported with fast data acquisitionates or scan speeds. Moreover, because of the fast detector elec-ronics it is possible to run the instrument at shorter dwell timeshich helps in acquiring hundreds of compounds in a single chro-atographic run, provided the instrument parameters are properly
ptimized. With the current generation triple quadrupole masspectrometers, acquisition of a large number of MRM transitions≈10,000 for the instrument used) is possible. But, for a large mix-ure of molecules, as data is acquired at dwell times typically of10 ms, the sensitivity of the analysis is adversely affected [14],specially for compounds known to have lower response suchs synthetic pyrethroids (e.g. cyfluthrin, cypermethrin). Therefore,hromatographic separation and the dwell time have to be simul-aneously adjusted so that sufficient sensitivity is attained. In theurrent endeavor, multiresidue analysis of 375 compounds by a sin-le method involved screening of at least 750 MRM transitions (oneor quantifier and one for qualifier). Accommodating such a largeumber of MRM transitions requires segmenting of the chromato-raphic run time into appropriate sections in such a manner thathe dwell times and number of data points (to attain proper peakhapes, sensitivity and quantification) together facilitate achiev-ng required selectivity, specificity and sensitivity. Besides, therere other factors such as chromatographic separation and injectiononditions that need to be optimized to attain required selectivitynd sensitivity. Therefore, a thorough instrumental optimizationas necessary, as presented in Supplementary material S2.
.2. Sample preparation
The ethyl acetate based sample preparation method reportedarlier [12] resulted in satisfactory recovery of the test compoundsrom grapes, okra and tomato with minor modifications in cleanuptrategy. Since okra contains chlorophyll pigments in considerablyigher concentrations, cleanup with only PSA could not removeolor from the extracts. Upon injection of this dark green extract5 �L), deposition of matrix on the GC liner was observed after few≈20) injections. This resulted in variable responses (RSDs > 20%)s observed while doing repeatable injections of the same extract.n addition, degradation of some compounds such as iprodionend carbaryl was also observed when the GC liner got contami-ated with the matrix components. The cleanup strategy was thusptimized by recovery experiments and the matrix effects eval-ated. Introduction of 7 mg of GCB along with PSA (25 mg) wasufficient in attaining the required cleanup resulting in repeatableesponses. Comparison of RSDs from repeatable injections (n = 20)f the extracts showed that RSDs in case of the extract treatedith GCB and PSA were lower than the extracts treated with PSA
nly. An increase in the quantity of GCB above 7 mg/mL resultedn lower recoveries for chlorothalonil which is also reported inarlier studies [15,16]. Addition of 7 mg GCB also did not requireny additional step of recovering adsorbed pesticides by additionf toluene as reported in literature [16]. Recoveries of most com-
ounds did not change significantly with increase in the amount ofCB up to 15 mg. The overall recoveries of PCBs and PAHs were notffected till 10 mg GCB. However, further addition of GCB reducedhe recoveries significantly to <67%.
. A 1270 (2012) 283– 295 291
In case of pomegranate and onion, the same method had limita-tions as evidenced by the interfering matrix peaks that affected thequantification of the target compounds. Modification in the cleanupstrategy was therefore essential. The ethyl acetate extract of oniontreated with PSA alone resulted in tR shifts up to 1–2 min for mostof the early eluting compounds and the chromatographic resolu-tion between the closely eluting compounds was severely affected(Fig. 1). The shift in tR could be explained by the overloading effectsthat are strongly related to the sample capacity of stationary phases.During PTV injection, time given for removal of the solvent or lowboiling matrix components through evaporation is short and it failsto remove many of the co-extracted interfering matrix componentswhen an onion extract is injected. The screening application in suchcases also appears difficult due to change in retention times, sensi-tivity, etc. Such shifts in tR could be avoided when the same ethylacetate extract obtained after cleanup of the onion extract with PSAwas evaporated under gentle stream of nitrogen (to vaporize off thevolatile matrix compounds), reconstituted in ethyl acetate and sub-sequently injected into the GC–EI-MS/MS. In case of pomegranate,the matrix induced signal suppressions were noted for most of thecompounds. Satisfactory results could be obtained by cleanup using25 mg PSA and 25 mg C18 per mL of the extract as described earlier[7,17].
3.3. Method validation
Linearity of the calibration curves of all the test compoundsin each of the five matrices could be established with r2 > 0.98.Detection of false positives in the control sample extracts foreach matrix was <1% indicating the specificity and sensitivity ofthe method. The method had sufficient sensitivity as indicatedby the MDLs in all the five tested matrices which were within1–2 �g/kg and below the prescribed EU-MRLs. However, due tothe fact that the method linearity is not adequate at these lowconcentrations the practical LOQ was considered as the standardconcentration corresponding to the first calibration point. The LOQsfor most of the compounds were <5 �g/kg whereas for few com-pounds the LOQs ranged between 5 and 10 �g/kg (Fig. 2A). Inmost cases, the LOQ of individual compounds followed the ordergrape < okra ≈ tomato < onion < pomegranate. Although LOQs weresomewhat higher in certain compound-matrix combinations suchas onion and pomegranate, in every case these were below the MRLsfor all the tested matrices. Examples of compounds having higher,but still adequate LOQs are carbaryl, dicofol, fenvalerate, esfen-valerate, and prallethrin. The evaporation step used in case of onionreduced the matrix co-extractives. However, the same was notapplicable in case of pomegranate and evaporation of the sampleextracts had negligible effect on removal of matrix coextractives.In case of pomegranate, the matrix induced signal suppressionsresulted in higher LOQs as compared to onion.
Negligible matrix effect was noted for most of the test analytes ingrape samples. The application of CID-MS/MS is also of high signif-icance in this respect, since the sensitivity and selectivity achievedare due to the possibility of monitoring compound specific set ofprecursor and product ions, which could discriminate the targetcompounds from matrix co-extractives. When using calibrationstandards prepared in solvent, significant matrix enhancement wasnoted for samples of pomegranate and onion, particularly for theearly eluting compounds, such as dichlorvos, fenobucarb, propoxur,monocrotophos, etc. This results from the relatively higher con-centration of co-extractives in these matrices which compete for
active sites in the flow path [18]. Moderate enhancement in signalswas observed for tomato and okra. However, in order to obtainaccurate quantifications, the matrix matched calibration standardswere preferred.
292 K. Banerjee et al. / J. Chromatogr. A 1270 (2012) 283– 295
A-Ma trix matched
standard af ter drying
B-Ma trix matched
standard before
drying
C-Solvent standard
F recone hout d
wa1ipts(caoacScccwfaba
tt
tI
ig. 1. Partial separation of �-HCH and lindane could be obtained after drying andluting compounds, �-HCH and lindane was severely affected in onion extracts wit
The recovery for the test compounds at 5, 10 and 25 �g/kgas within 70–120% with the associated relative standard devi-
tions <20% in all the test matrices. Recoveries in grapes at0 �g/kg were >90% (Fig. 2B) for most of the compounds whereas
n okra, tomato and pomegranate the recovery values were com-aratively lower than the observed values in grapes for most ofhe compounds which could be attributed to the matrix inducedignal suppressions. Similar trend of relatively lower recoveries<90%) were observed for okra and tomato at 25 �g/kg. In onion,hlorothalonil disappeared rapidly and was not detectable in ethylcetate extracts. Chlorothalonil added to ethyl acetate extracts ofnion also disappeared due to reaction with matrix co-extractivesnd conversion to more polar compounds [19]. For other test matri-es the recovery of chlorothalonil was >70% with RSDs below 20%.imilarly, due to the interaction of carbosulfan with the matrixomponents [20] carbosulfan disappeared in all the tested matri-es with recovery of <10%. Recoveries of polar organophosphorousompounds viz. acephate, methamidophos, monocrotophos, etc.ere >75% at all the tested concentrations. The ratio of the RSD
or reproducibility to RSDR and RSD for repeatability to RSDr ofll the analytes calculated at 10 ng/mL level of fortification wereelow 2, indicating satisfactory level of intra-laboratory precisionnd accuracy.
The measurement uncertainty of the analytes was estimated atheir respective LOQs. Based on the expanded uncertainty valueshe analytes could be broadly classified into three groups.
Group I: Expanded uncertainty up to 10%Group II: Expanded uncertainty 10–20%Group III: Expanded uncertainty 20–50%
Most analytes could therefore be estimated with ≤20% uncer-ainties in all the commodities. Analytes belonging to GroupII were carbosulfan, cyfluthrin isomers, cypermethrin isomers,
stituting onion extracts (A) while chromatographic resolution between the closelyrying (B) as compared to solvent standards (C).
dimethomorph, azoxystrobin, difenoconazole, and propanil whilethose belonging to Group II were 4-bromo-2-chlorophenol(metabolite of profenophos), alachlor, carbaril, carbofuran-3-OH,chlorothalonil, demeton-S-methyl, dichlorvos, dicofol, difluben-zuron, dimethoate, fenchlorphos-oxon, fluchloralin, malathion,metribuzin, oxadiazon, oxycarboxin, phenothrin, phorate, pro-cymidone, profenophos, pyremethanil. Examination of the indi-vidual uncertainty components indicated that in Group II thecomponent U1 had maximum contribution towards the combineduncertainty (>30% as opposed to <20% in Group I) which was theresult of poor peak shapes with considerable tailing. This resulted inquantification losses during automated peak processing. However,it could be resolved by manual integration of the peaks of these ana-lytes. For analytes belonging to Group III, the contribution of U1 wasconsiderably higher (>50%) as compared to the other two groups.The other components of uncertainty corresponding to precisionand accuracy were within 10–15% of the combined uncertainty.When the individual matrices were compared, it was observed thatanalytes in general had higher uncertainties in pomegranate matrixfollowed by onion, okra, tomato and grape. This was in confor-mity with the decreasing trend of matrix effects observed in thesesamples.
The validation set for each of the 5 matrices consisted of32 sample runs (7 solvent based calibration standards, 7 matrixbased calibration standards, 6 recovery samples for each of the3 levels) with 375 compounds each resulting in a total of 60,000data values. Analysis of validation data (for LOQ, matrix effects,recovery and RSD/RSDr calculations) was therefore a time con-suming and tedious job. An in house developed MS Excel basedmacro was thus developed to process the data and found effec-
tive in processing such large amount of data. The excel tableexported from the quantitative file of the MassHunter softwarecontains compound-wise recovery data, S/N ratios, etc. Data anal-ysis conventionally takes huge amount of time since that required
K. Banerjee et al. / J. Chromatogr. A 1270 (2012) 283– 295 293
0.00
0.50
1.00
1.50
2.00
2.50
3.00
3.50
4.00
4.50
5.00
0 50 100 150 200 250 300 350
LOQ
(µg/
Kg)
Analyte Numb er
A) LOQ
Grape Okra Onion Pome granate Tomato
70
75
80
85
90
95
100
105
110
115
120
0 50 100 150 200 250 300 350
Aver
age
Reco
very
(%)
Analyte Numb er
B) Average Recovery (%)Grape Ok ra Onion Po megranate Tomato
F nds haR es for
roaocLmtsmtcd
(
(
set)
ig. 2. (A) LOQ of the test compounds in five tested matrices. Most of the compouelatively higher LOQs were observed for onion and pomegranate. (B) The recoveri
earranging the data for all 375 compounds. A macro was devel-ped specifically to rearrange the data for calculation of recoveriest different fortification levels. The macro was initially devel-ped for one compound only and repeated for the set of 375ompounds. Similarly macros were developed for calculation ofOQs, and summarization of data for identification of the analyteseeting the recovery criteria of 70–120%. The same macros were
hen applied on the other four commodities. The compilation andummarization of data for 375 compounds in five different com-odities could be completed quickly using macros. It was observed
hat the time required for processing of validation data of each
ommodity could be accomplished within 2 h as opposed to 2ays.
d LOQs < 5 ppb. In general, lower LOQs were observed for grape, okra and tomato.the test compounds were within 70–120% for all the test matrices.
3.4. Application of semi-quantification method
The data files obtained during the validation study and real sam-ple analysis was divided into three sets:
a) Set I: consisting of runs from the matrix matched standards(validation set)
b) Set II: consisting of runs from the recovery samples (test set)(c) Set III: consisting of runs from the incurred samples (application
The slope ratios from each of the matrix matched stan-dards from the validation set were calculated against each other
294 K. Banerjee et al. / J. Chromatogr. A 1270 (2012) 283– 295
ig. 3. Incurred residues of methamidophos (A), acephate (B) and dimethoate (C) womegranate and onion.
Supplementary information). As discussed in Supplementarynformation, semi-quantification of an analyte by calibration stan-ards with conversion factors (slope ratio) ≈ 1 lead to minimumrror (%) in quantification. A preliminary study indicated thator analytes with similar response such as dichlorvos, �-HCH,cephate, pyremethanil, triphenylphosphate and pentoxazone thatad conversion factors in the range of factors 0.8–1.2 resulted inemi-quantification with <10% error in quantification. The errorsn quantification increased to ≈20% when the analytes with con-ersion factors in the range of 1.2–1.8 or 0.6–0.8 were used, asbserved for etridazole and dichlorvos (example demonstrated inupplementary material).
The values of the slope ratios obtained from the validationet were examined on the “test set”. As for example, consideringhe absence of calibration curve of an analyte, e.g. trifloxys-robin, the calibrations from the other compounds with conversionactor ≈1 was employed to quantify the residue content of tri-oxystrobin. For a recovery sample fortified with trifloxystrobinesidues at0.025 mg/kg concentration, the average concentrationn = 6) calculated from the calibration curves of dichlorvos, tri-uralin, carbofuran, ethion, propiconazole, and etofenprox were.025 (±4%), 0.022 (±3%), 0.024 (±4%), 0.024 (±3%), 0.025 (±3%)
nd 0.031 (±3%) mg/kg, respectively. Quantification of the sameample through the calibration curve of trifloxystrobin itselfesulted in concentration of 0.024 (±3%) mg/kg. Thus, the calibra-ion equation of dichlorvos, carbofuran, ethion and propiconazole
und in grape, while residues of carbofuran (D) and dichlorvos (E) were detected in
could be well applicable for the quantification of trifloxystrobinresidues, each providing more than 96% accuracy in quantifica-tion.
After examining the applicability of semi-quantification on the“test set”, the real world samples comprising the “application set”were quantified in a similar way and the results obtained werewithin ±5% of the concentration derived from the respective cali-bration curves with RSDs < 10%.
3.5. Application for analysis of incurred samples
The optimized method was applied for the analysis of incurredsamples (10 samples of each matrix) obtained from local mar-kets of Pune and Bangalore. Incurred residues of methamidophos,acephate and dimethoate (Fig. 3) were found in grape, whileresidues of dichlorvos and carbofuran were detected in onion andpomegranate, respectively. The other samples were free from anyresidues of the test chemicals. However, in all cases the residueconcentrations were below the respective EU-MRLs. The incurredresidues of these identified chemicals were also quantified by thesemi-quantification approach and the concentrations estimatedwere within ±15% of the values calculated through the calibra-
tion graph of methamidophos, acephate, dimethoate, dichlorvosand carbofuran.
Grape samples in three different sets spiked at different con-centrations with chlorpyriphos methyl, �-cyhalothrin and �-HCH
K. Banerjee et al. / J. Chromatogr. A 1270 (2012) 283– 295 295
Table 2Application of the semi-quantification approach on inter-laboratory test samples.
Name of compound Laboratory 1 Laboratory 2 Laboratory 3
ere distributed among three commercial testing laboratories inndia and analyzed using the validated method. The quantifica-ion of the positive findings was carried out with the calibration ofheir own standards and also by the semi-quantification approach.he results obtained with the two approaches are summarizedn Table 2. From the results it could be concluded that theemi-quantification approach could be used for large scale tar-et screening of pesticide residues in routine residue monitoringrograms.
. Conclusions
The multiresidue method was successful for the analysis of 375ompounds in five different commodities with satisfactory preci-ion and accuracy, demonstrating the suitability of the method fornalysis of contaminants from various fruits and vegetables bothor regulatory as well as routine residue monitoring purposes. Inddition to the relative simplicity of the extraction method, theide scope of the analytes as well as the matrices tested offer theotential of its application as a readymade method. In addition, theethod has the potential of being employed for screening residues
eyond the target list and attaining a semi-quantified result. As aesult of the wide scope of the method, the acquired data couldurther be used to mine the data for non-targeted compoundsithin the scope of the MRM data base and thereby aide surveil-
ance studies. In the future, an inter-laboratory collaborative studys proposed to examine reproducibility of the semi-quantificationpproach and its application under different sets of GC–EI-MS/MSonditions.
cknowledgments
The authors acknowledge funding support from the ICAR
ational Fellow project and the National Referral Laboratoryroject of APEDA. Thanks are also due to Paul Zavitsanos, WW Busi-ess Development Manager, Agilent Technologies, for support and
Supplementary data associated with this article can befound, in the online version, at http://dx.doi.org/10.1016/j.chroma.2012.10.066.
References
[1] Insecticides Registered under section 9 (3) of the Insecticides Act, 1968 ason 20/01/2012, New Delhi, India. http://www.cibrc.nic.in/reg products.htm(accessed 13.03.12).
[3] J.W. Wong, K. Zhang, K. Tech, D.G. Hayward, A.J. Krynitsky, I. Cassias, F.J.Schenck, K. Banerjee, S. Dasgupta, D. Brown, J. Agric. Food Chem. 58 (2010)5884.
[4] R. Savant, K. Banerjee, S.C. Utture, S.H. Patil, M.S. Ghaste, P.G. Adsule, J. Agric.Food Chem. 58 (2010) 1447.
[5] S. Walorczyk, J. Chromatogr. A 1208 (2008) 202.[6] J.L.F. Moreno, A.G. Frenich, P.P. Bolanos, J.L.M. Vidal, J. Mass Spectrom. 43 (2008)
1235.[7] S.C. Utture, K. Banerjee, S. Dasgupta, S.H. Patil, M.R. Jadhav, S.S. Wagh, S.S.
Kolekar, M.A. Anuse, P.G. Adsule, J. Agric. Food Chem. 59 (2011) 7866.[8] Method validation & quality control procedures for pesticide residues analysis
in food & feed, Document No. SANCO/12495/2011.[9] W. Horwitz, R. Albert, J. AOAC Int. 89 (2006) 1095.10] W. Horwitz, L.R. Kamps, K.W. Boyer, J. Assoc. Off. Anal. Chem. 63 (1980) 1344.11] Guide CG 4, Quantifying Uncertainty in Analytical Measurement, 3rd
