Analytica Chimica Acta 442 (2001) 207–219
Evaluation of an automated and integrated flow-throughimmunoanalysis system for the rapid determination of cephalexin
in raw milk
Z.-L. Zhi a, U.J. Meyer a, J.W. Van den Bedem b, M. Meusel a,∗a Institute of Chemical and Biochemical Sensor Research (ICB), Mendelstrasse 7, D-48149 Münster, Germany
b Central Orgaan voor Kwaliteitsaangelegenheden in de Zuivel (co*kZ), P.O. Box 250, 3830 AG Leusden, The Netherlands
Received 29 January 2001; received in revised form 28 May 2001; accepted 1 June 2001
Abstract
In the present work, an evaluation of an automated flow-through amperometric immunoanalysis system for the quantitationof the antibiotic drug cephalexin (Ce) in milk is described. The detection limit of the method was calculated as 1 �g/l (S/N =3), while the quantitation limit being 3 �g/l, well below the EU regulation mandated limit of 0.1 mg/kg for cephalexin in milk.Spiked milk samples with increased concentrations of cephalexin were investigated as blind coded duplicate samples. Total20 milk samples spiked with cephalexin between 3 and 30 �g/l were analyzed leading to a good correlation referred to thespiked values. The samples could be detected with a mean recovery of 97±23%, indicating a good agreement with the spikedconcentration. The precision and repeatability was determined using spiked samples with four different concentrations whichwere investigated on four different days. A mean within-day variation of 6.9% and a mean between-day variation of 10.9%were obtained. Correct classifications were achieved in false negative and false positive studies when the cut-off values wereset at 2.0 and 3.7 �g/l, respectively, further proving the detection and quantification capabilities of the method. The methodwas specific for cephalexin and free of interferences from other cephalosporins and penicillins at concentrations up to at least1000 �g/l. Milk matrix properties in terms of bacteriological quality, somatic cell content, and pH have no significant effectson the determination. The effect of the milk fat was eliminated by a simple-defatting step. For the investigation of sampleswith concentrations higher than 30 �g/l, a dilution step for the sample would be required. In addition, 17 natural contaminatedmilk samples of cephalexin-treated cows were also analyzed and the results obtained were confirmed by an enzyme-linkedimmunosorbent assay (ELISA) method. This intra-laboratory study demonstrated that the proposed method is suitable for therapid and reliable determination of cephalexin in raw milk samples. © 2001 Elsevier Science B.V. All rights reserved.
Keywords: Flow-through immunoanalysis; Cephalexin; �-Lactam antibiotic; Milk; Amperometric detection
1. Introduction
The public concern for the occurrence of antimi-crobial drug residues in food and the potential for
∗ Corresponding author. Tel.: +49-251-980-2879;fax: +49-251-980-2890.E-mail address: [emailprotected] (M. Meusel).
pathogenic organisms to develop resistance to thesedrugs have made it imperative to develop analyticalmethods with known performance for screening pur-poses. Typical primary screening tests based on mi-crobial inhibition and bacterial receptor assays [1–3]are normally qualitative or semiquantitative and notspecific due to various effects from dietary sources,animal diseases or other variables. Thus, confirmation
0003-2670/01/$ – see front matter © 2001 Elsevier Science B.V. All rights reserved.PII: S0003 -2670 (01 )01180 -1
208 Z.-L. Zhi et al. / Analytica Chimica Acta 442 (2001) 207–219
of the results of those tests is required. In recent years,there is a trend to develop immunochemical [3–12]and enzymatic assays [13–15] to detect �-lactamantibiotic residues. Moreover, HPLC methods cou-pled with immunochemical screening kits have alsobeen reported to determine the concentrations of theindividual �-lactam residues [16–19].
Cephalexin (Ce) is a potent antibiotic commonlyused in veterinary medicine in some countries. Forexample, such a compound is available on the Spanishmarket under the name Relexine, Maxicef and Ce-porex [20]. We have previously reported an ampero-metric immunosensing method based on an automatedflow system for the quantification of this compound inmilk [21]. In the present report, the analytical perfor-mance of the proposed method was further evaluatedby applying it to a validation that was performedwithin the EU project FAIR CT96-1181. The valida-tion was performed in cooperation with partners atco*kZ (The Netherlands) and IKERTEK diagnosticsSI (Spain). Spiked and natural contaminated milk ofblind coded samples provided by the project part-ners were analyzed by using the proposed method.The analytical figures of the merit of the method interms of sensitivity, detection capability, repeatability,precision, specificity and susceptibility to interfer-ences, and trueness were tested in this intra-laboratorystudy.
2. Experimental
2.1. Chemicals
Cephalexin was obtained from duch*efa (Haarlem,The Netherlands). N′-(3-dimethylaminopropyl)-N-ethylcarbodiimide hydrochloride (EDC) was pur-chased from Sigma (Steiheim, Germany). Alkalinephosphatase (AP) from calf intestine was obtainedfrom Biozyme (Gwent, UK). Protein G was pur-chased from Pierce (KFM, Laborchemie GmbH, St.Augustin-Buisdorf, Germany). Epoxy-activated Eu-pergit C 250 L beads were obtained from Röhm(Darmstadt, Germany). The particles were 250 �min diameter with internal pores of approximately100 nm. Bovine serum albumin (BSA) and all buffersalts and other fine chemicals were purchased fromMerck (Darmstadt, Germany) or Sigma (Deisenhofen,
Germany). p-Aminophenyl phosphate was synthesizedat XanTec Analysensysteme (Münster, Germany).
2.2. Polyclonal anti-cephalexin antibody
Polyclonal cephalexin antiserum was supplied byDr. E. Loomans from ATO-DLO, Wageningen, TheNetherlands. The preparation of immunogen and im-munization of New Zealand white rabbit have beendescribed elsewhere [21]. Antibody purification fromthe serum was performed by affinity chromatographyusing protein G-sepharose, as described by Trau et al.[22]. The final IgG concentration was 1.9 mg/ml asdetermined by BCA protein assay reagent kit (Pierce).The antibody batch was then divided into 500 �laliquots and stored at −20◦C.
2.3. Cephalexin–alkaline phosphatase conjugate
Cephalexin–alkaline phosphatase conjugate (Ce–AP) was prepared by carbodiimide coupling pro-cedure. Cephalexin (20 mg) was first dissolved in1 ml 0.2 M phosphate buffer pH 7.0, and AP (2 mg)was dissolved in 2 ml of the same buffer. Activatedcephalexin and AP were mixed with an 850-fold mo-lar excess of cephalexin. The mixture was stirred for5 min at 4◦C. Then 1.6 mg EDC was added. The batchwas then incubated overnight at 4◦C under stirring.Unbound �-lactam was removed by dialysis through aSlide-A-Lyzer 10 K dialysis cassette (Pierce) against0.2 M phosphate buffer pH 7.0. The conjugate solutionwas then concentrated to 2.0 mg/ml by centrifugationusing a Centricon-10 device with a 12,000 Da cut-offmembrane (Millipore). The conjugate was finallyspiked with 0.05% NaN3 and kept at 4◦C for storage.
2.4. Milk samples
The sampling of raw milk and analyte spiking wereplanned and realized in close cooperation with co*kZ(Leusden, The Netherlands) on a blind coded studybase. The milk samples served as negative controlsamples or for the preparation of spiked samples andwere taken from 10 individual cows coded M1–M10with good clinical health and within at least 8 weekswithout any antimicrobial substance treatment. Assoon as the samples were collected from the farms at“Praktykondenzoek Rundvee, Schapen en Paarden”
Z.-L. Zhi et al. / Analytica Chimica Acta 442 (2001) 207–219 209
(Lelystad, The Netherlands), they were spiked withthe desired concentrations of cephalexin immediatelyand were stored at −20◦C. The samples were trans-ported to the laboratory for analysis on the next day.All the samples were then analyzed within 10 daysafter arrival. The concentrations of the analyte inthe samples were unknown for the analysts duringthe analysis.
For the different types of experiments, the follow-ing samples were investigated. For precision study,four sets of four samples of 25 ml with the additionof, respectively, 3, 5, 10, and 20 �g/l cephalexin wereprepared. On four different days, analyses of the sam-ples were performed each day in duplicate. On eachday, the analytical system had to be calibrated again.Standards containing 0, 1, 3, 5, 10, and 30 �g/l ofthe analyte in duplicate measurements were used forcalibration.
For trueness study, 20 samples of 25 ml with theaddition of 3–30 �g/l cephalexin (blind duplicatesof 10 samples) were obtained. Standards used forcalibration: 0, 1, 3, 6, 10, and 30 �g/l in duplicatemeasurements.
For the determination of quantitation limit, 10 neg-ative samples of 25 ml milk from each cow and 10positive samples of 25 ml from each cow with the ad-dition of 3 �g/l cephalexin were analyzed. Standardsused for calibration: 0, 1, 3, 6, and 10 �g/l in duplicatemeasurements.
In order to test the influence of the milk quality,eight samples of 10 ml each, with a high viable countor with a high amount of somatic cells were analyzed.The milk samples were obtained from The NetherlandsMilk Control Station (Zutphen, The Netherlands) and0 or 10 �g cephalexin/l of milk was added. Standardsused for calibration: 0, 1, 3, 10, and 30 �g/l in dupli-cate measurements.
For quality control purposes, each day two milksamples were analyzed. One milk with 10 portions of10 ml containing no cephalexin was used as a negativecontrol sample for the analysis. For positive qualitycontrol of the analysis and the stability study of theanalyte in the milk, a milk sample containing knownamount of the spiked analyte (10 �g/l) was also pre-pared and tested each day during the measurements.The spiked sample was divided into aliquots and storedat −20◦C. Each time only one vial was thawed andused for analysis.
For investigation of the effect of individual milkmatrix, 45 cephalexin-free samples were investigatedin duplicate measurements.
Milk samples derived from treated cows: 17 samplesfrom four cephalexin-administered cows which weresupplied by IKERTEK were analyzed in duplicate.The first three cows with mastitis were treated withthe antibiotic intramammarily for three times every12 h with 200 mg cephalexin. The milk was collectedbefore treatment (0 h) and after a certain time (12 h) ofeach administration. The fourth cow was treated onlyonce with 200 mg cephalexin.
2.5. Sample pretreatment
To minimize the risk that milk fat would depositin the immunoreactor or otherwise interfere withthe assay, all samples were defatted for 10 min at2000 × g. Moreover, the samples were heat-treatedat 70◦C for 3 min prior measurement in order to in-activate endogenic alkaline phosphatase in the rawmilk which would otherwise interfere the assay. Inthe previous study, a simple recovery study showedthat no cephalexin was lost with the milk defattingand heating treatments [21].
2.6. Instrumentation and the flow-through method
The main principle of the flow-through im-munosensing system has been described previously[21]. A brief description is given here. A columnreactor filled with protein G-coated porous polymerbeads was used as the affinity matrix for the antibodyimmobilization. In the assay, the assaying compo-nents were pumped sequentially to the immunoreac-tor. The cephalexin–alkaline phosphatase conjugatecompeted with the analyte in the sample for bindingto the immobilized antibody while flowing throughthe immunoreactor. Subsequently, p-aminophenol,generated from the enzymatic (AP) reaction withp-aminophenyl phosphate, was detected amperomet-rically. In the last step of the assay cycle, the complexof antibody and hapten–enzyme conjugate or the ana-lyte was dissociated from the protein G matrix in theimmunoreactor by a regeneration buffer. A completecycle for the fully automated assay procedure requiredapproximately 16 min including the regeneration. Dueto the excellent stability of the immunoreactor, the
210 Z.-L. Zhi et al. / Analytica Chimica Acta 442 (2001) 207–219
Fig. 1. Scheme of the flow-through immunosensing system for the determination of cephalexin in milk. Inserts represent the structure ofthe immunoreactor and amperometric detector, respectively.
system could be used for at least 2 weeks with morethan 250 measurements in total.
2.7. Flow-through amperometricimmunosensing system components
The detecting system, as shown in Fig. 1, consistsof two connected modules: the control/power supplymodule including micro-controller (V40 card), key-board, drivers, potentiostat, LCD-display and printerand the flow-through module including a selector(four two-way valves), three three-way valves (Ander-son, Dortmund, Germany), a peristaltic pump (KBL,Karlstein, Germany), a home-made immunoreactorand a flow-through amperometric detector. The wholesystem is designed and constructed as a compactstand-alone device. The system can also be controlledexternally via a computer with windows compati-ble in-house made software FIABOLO. In addition,the system is connected to an external autosamplerwhich controls the delivery of sample/conjugate orthe washing buffer into the immunoreactor through
a one-channel peristaltic pump. The immunochemi-cal reaction takes place in the immunoreactor whichis fitted directly to the two three-way injection/loadvalves. The amperometric detector consists of ascreen-printed carbon working electrode and anAg/AgCl reference electrode. The electrodes wereconnected to an integrated potentiostat (MP/3, Tech-nical University Vienna, Austria) and polarized at aworking potential of +150 mV versus Ag/AgCl.
2.8. Flow-through immunoreactor
The immunoreactor used in this study, as shown inFig. 1, was made of plexiglass material. Details of thedesign of the reactor have been described in previouswork [21].
2.9. Immobilization of protein G to the beadsin the reactor
Before the beads were filled into the immunoreactor,protein G was immobilized covalently on the surface
Z.-L. Zhi et al. / Analytica Chimica Acta 442 (2001) 207–219 211
of the epoxy-activated polymeric material (EupergitC 250 L) according to a protocol described at [21].Quantitation of protein G contents in the supernatantbefore and after immobilization showed that the pro-tein G loading on the porous beads was 16 mg/g.
2.10. Measurement procedure
After defatting and heat treatment, a 1.0 ml volumeof undiluted sample that was mixed with 0.5 ml vol-ume of Ce–AP conjugate was introduced into a 3.0 mlpolypropylene vial and placed on the sample rack ofthe autosampler. The automatic analyzing procedurewas then started in the following way.
In the first step, the immunoreactor was loadedwith 9.5 �g/ml polyclonal anti-cephalexin antibod-ies in 0.1 M phosphate buffer, pH 7.2, containing1.5 M NaCl (PBS buffer) and 1% BSA for 2 min ata dispensing flow rate of 0.3 ml/min. In a secondstep, the Ce–AP conjugate (0.2 �g/ml in 0.1 M PBSbuffer containing 3% BSA) was added to the sam-ple (two parts of sample plus one part of conjugate,2 + 1, v/v). The mixture was then passed throughthe immunoreactor at a flow rate of 0.3 ml/min. Sub-sequently, 2 mM p-aminophenyl phosphate substratesolution — kept in a brown bottle for light protec-tion — in 50 mM carbonate buffer, pH 10, containing1 mM MgCl2 was filled into the immunoreactor ata flow rate of 0.6 ml/min in 1 min, after another2 min of stopped-flow, the generated electroactivep-aminophenol from the AP-catalyzed reaction withp-aminophenyl phosphate was flushed out of theimmunoreactor by re-starting the pump and thendetected amperometrically. Finally, the immunocom-plex was removed from the reactor by applying abuffer containing 0.1 M glycine, pH 2.0, 50% ethy-lene glycol and 0.05% Tween-20 for 3 min at a flowrate of 0.6 ml/min, leading to a dissociation of theantibody–protein G interaction. The immunoreactorwas then flushed by the washing/conditioning buffer(50 mM Tris buffer with saline pH 7.2 and 0.05%Tween, supplemented with 1% BSA) and was readyfor a new measuring cycle. Between each step of addi-tion of immunochemical components, a 1 min washingstep with washing buffer was included at a flow rateof 0.6 ml/min. Occasionally, the immunoreactor hadto be back-flushed with the aid of a syringe (washingbuffer) in order to maintain a constant flow rate.
2.11. Construction of calibration curvesand data analysis
A stock standard solution containing 1000 �g/ml ofcephalexin was used to prepare the working standardscontaining known concentrations of the analyte in anegative control milk. The stock was divided into dif-ferent batches and stored at −20◦C when not in use.Calibration curves with six concentration levels (0, 1,3, 10, 30, and 100 �g/l) were prepared and were usedsubsequently to determine the concentration of ana-lyte in the sample. The x-axis (concentration in �g/l)was in logarithmic scale, while the y-axis representedthe peak height of the amperometric signal in nA. Thedata points for the calibration were fitted using thefour-parameter logistic model (sigmoidal shape). Forcomparison of different calibration curves, the peakheights were normalized against signal of the negativecontrol milk sample containing no analyte.
2.12. Enzyme-linked immunosorbent assay (ELISA)
An antibody-coating format was used to performthe ELISA using Maxisorb microplates from Nunc(Wiesbaden-Biebrich, Germany). The detailed proto-col of the ELISA has been described elsewhere [21].
3. Results and discussion
3.1. Cephalexin calibration curves anddetection limit
The calibration curve was first constructed in du-plicate measurements for each concentration usingcephalexin standards prepared in 0.01 M PBS buffer.It showed a sigmoidal shape for the cephalexin con-centration range of 0–30 �g/l and was fitted by thefour-parameter semi-logarithmic model. The detec-tion limit (S/N = 3) was estimated to be 1.0 �g/l.
3.2. Specificity/susceptibility to interferences
Cephalexin calibration curves were further com-pared using the standards prepared in 0.01 M PBSbuffer, defatted control milk and untreated controlmilk. The results are presented in Fig. 2. It can beseen that the calibration curve using untreated raw
212 Z.-L. Zhi et al. / Analytica Chimica Acta 442 (2001) 207–219
Fig. 2. Effect of the matrix on the cephalexin determination. Calibration curves performed in 0.01 M PBS buffer, pH 7.2, in defattednegative control milk and in a negative control raw milk. Mean values and errors of duplicate determinations.
milk was shifted in parallel to higher signals in com-parison to the calibrations in buffer and defatted milk(10 min centrifugation at 2000 × g). An effect ofendogenous alkaline phosphatase activity could beexcluded, because the milk was heat-treated at 70◦Cprior to use. Thus, only the fat content was respon-sible for the matrix effects observed in Fig. 2. Thisobservation also indicated that a sample-defatting stepwas essential. The absolute signal (peak height) fora defatted milk was only slightly higher compared tothat in PBS buffer (Fig. 2). In order to make the cali-bration as precise as possible, the working standardsfor the subsequent study were prepared in the controlmilk rather than in PBS buffer.
To study cross-reactivity, competitive calibrationcurves were performed for some other �-lactam an-tibiotics, including penicillins (penicillin G, amox-cillin, cloxacillin, ampiciline) and cephalosporins(ceftiofur and cephapirin). The results showed thatpenicillin G, cloxacillin, amoxicillin, and ceftiofurdid not interfere in the cephalexin determination upto concentrations of 10 mg/ml. Only cephapirin wasfound to show a competition with the Ce–AP conju-gate in binding to the antibody. In Fig. 3, calibrationcurves for cephalexin and cephapirin are presented.Comparing both calibrations, it could be concluded
that cephapirin did not interfere with cephalexin deter-mination up to a concentration of 1 mg/l cephalexin.The cross-reactivity of cephapirin was calculated tobe 0.13%. Considering the cephalexin determinationrange of 3–30 �g/l, it can be concluded that the sen-sitivity to other �-lactam antibiotics is at least 1000times lower than to cephalexin. Thus, the interferenceis negligible and quantitative cephalexin results canbe expected for milk samples containing �-lactams atconcentrations lower than 1 mg/l.
The influence of the milk quality to the analysis wasalso tested by analyzing two samples with high viablecount and another two with a high number of somaticcells. In each case, one sample was spiked with the an-alyte at a level of 10 �g/l, and another was non-spiked.Results are presented in Table 1. It is obvious that nei-ther a high viable count nor a high number of somaticcells may interfere in the determination of cephalexinin milk samples.
Another parameter that may affect the cephalexindetermination is the pH value of the milk samples. Dueto the high buffering capacity of the PBS buffer usedin preparing the Ce–AP conjugate, the pH values of amixture of milk sample and the conjugate was foundhighly stable, the variation of pH from a number of 20different samples were measured to be 6.90±0.02 after
Z.-L. Zhi et al. / Analytica Chimica Acta 442 (2001) 207–219 213
Fig. 3. Comparison of calibration curves for cephalexin and cephapirin. Mean values and errors of duplicate determinations.
the samples were mixed with the Ce–AP conjugate.Thus, it can be concluded that the effect of the pHvalue of the individual milk samples on the cephalexindetermination is negligible.
3.3. Estimation of the quantitation limit
To further demonstrate the detection and quantita-tion capabilities of the proposed method, 10 sampleswithout cephalexin and 10 samples spiked with 3 �g/lcephalexin (at the level of limit of quantitation) wereanalyzed in single measurements. The mean valuesof the negative and positive samples were 2.0 ± 1.4and 3.4 ± 1.0 �g/l, respectively. There is a significant
Table 1Influence of the viable count and somatic cells on the determination of cephalexin in milk
Milk sample Code Milk quality Cephalexinadded (�g/l)
Cephalexindetermined (�g/l)
Recovery(%)
M14 R707 Viable count 18,000,000 per ml 0 0 –R707 Viable count 18,000,000 per ml 0 0 –R708 Viable count 18,000,000 per ml 10 10.1 101R708 Viable count 18,000,000 per ml 10 11.2 112
M16 R715 Somatic cells 670,000 per ml 0 0 –R715 Somatic cells 670,000 per ml 0 0 –R716 Somatic cells 670,000 per ml 10 10.6 106R716 Somatic cells 670,000 per ml 10 9.4 94
difference between both groups of samples. Thenumber of measurements, however, is too low tocalculate a cut-off value so that samples with concen-trations higher than the cut-off could be considered“positive”.
To prove that the individual sample matrix doesnot lead to false positive results of the method, 45milk samples containing no cephalexin (controlled byELISA experiments) were analyzed in duplicate mea-surements by the proposed method. Normalized signalintensities are presented in Fig. 4 referring to the spe-cific working range (between the signal of the negativecontrol sample B0 and signal of the unspecific Ce–APconjugate binding B). The normalized intensities were
214 Z.-L. Zhi et al. / Analytica Chimica Acta 442 (2001) 207–219
Fig. 4. Influence of the sample matrix on the cephalexin determination. Normalized signal intensities referring to the specific workingrange (between the signal of the negative control sample B0 and signal of the unspecific Ce–AP conjugate binding B). The normalizedintensities were then subtracted by 1 to define the signals of the blank milk as 0. The measurements were performed during 5 days. Theline for 1 �g/l is defined as the detection limit of the method. The dotted line represents the standard deviation of the normalized valuefor 1 �g/l of independent determinations on five different days. Mean values and errors of duplicate determinations.
then subtracted by 1 to define the signals of the blankmilk as 0. The measurements were performed during5 days. The line for 1 �g/l is defined as the detectionlimit of the method. The dotted line represents thestandard deviation of the normalized value for 1 �g/lof independent determinations on five different days.As can be seen in Fig. 4, all values are ≤1 �g/l takingthe standard deviation of each value and the 1 �g/l lineinto account. It can thus be concluded that a cut-offvalue of 3 �g/l can be defined. With this cut-off value,all samples tested can be considered as negative andno false positives were detected.
3.4. Repeatability
The repeatability was evaluated by determining themean within-day and between-day variations accord-ing to the outline of collaborative study procedureof the International Dairy Federation [23]. Four sam-ples spiked each with cephalexin levels of 3, 5, 10,and 20 �g/l were analyzed in duplicate within 4 days.Calibrations were performed independently each day.The raw data of these experiments and the within-dayand between-day variation results are combined inTable 2. The mean within-day relative variation was
6.9%, while the average of between-day relative vari-ation coefficient being 10.9%.
3.5. Trueness
The trueness/accuracy of the analytical resultsprovided by this method was evaluated by analyz-ing 20 spiked milk samples (blind duplicate of 10different samples). Each sample was analyzed in asingle measurement. Due to the nature of the blindpair study, no repetition was possible for all mea-surements. A comparison of the determined concen-trations with the spiked values are lised in Table 3,together with other milk properties determined, suchas the fat content, pH, somatic cell count, and bac-terial count. It appeared that the difference betweenthe determined values and the values were not af-fected by the milk composition or quality. Fig. 5shows a correlation between the determined values(average of the blind pair) and the spiked values.A slight over-determination was observed at lowerconcentrations and under-determination was found atthe higher concentrations. In general, the determinedvalues were significantly correlated with the spikedvalues (r = 0.9402, n = 10, P < 0.0001), a slope of
Z.-L
.Z
hiet
al./Analytica
Chim
icaA
cta442
(2001)207–219
215
216 Z.-L. Zhi et al. / Analytica Chimica Acta 442 (2001) 207–219
Table 3Analysis of milk samples for cephalexin concentration and the milk properties of the samples
Milksample
Code Cephalexin (�g/l) Recovery(%)
Fat contenta
(%)pH Somatic cell
countb (×1000 per ml)
Bacterialcountc
Spiked Determined Corrected for fat
M1 R674 3 3.8 3.6 120.0 5.42 6.60 20 193R688 3 7.5 7.1 236.7
M2 R676 5 4.8 4.7 94.0 1.18 6.65 11 446R689 5 7.8 7.7 154.0
M3 R687 10 10.9 10.7 107.0 1.92 6.46 14 >999R690 10 7.6 7.5 75.0
M4 R677 10 13.2 12.6 126.0 4.63 6.51 269 258R686 10 12.2 11.6 116.0
M5 R675 15 17.6 17.2 114.7 2.15 6.66 16 224R685 15 15.9 15.6 104.0
M6 R684 20 13.2 12.5 62.5 5.44 6.64 80 136R691 20 21.1 20.0 100.0
M7 R678 20 18.8 18.0 90.0 4.31 6.76 38 5R682 20 15.9 15.2 76.0
M8 R679 25 17.7 16.8 67.2 4.81 6.64 46 7R683 25 18.7 17.8 71.2
M9 R681 25 27.5 27.0 108.0 1.85 6.66 28 8R692 25 25.0 24.5 98.0
M10 R673 30 23.0 22.4 74.7 2.61 6.76 18 3R680 30 27.0 26.3 87.7
a Determined by infrared (Milcoscan).b Determined by fosfomatic.c Determined by bactoscan.
0.721 ± 0.092 testifying that the determined valueswere somewhat lower than that of the spiked values.Taking into account that only one single measurementfor each sample (blind duplicate) was performed, thecorrelation observed can be considered reasonablygood.
3.6. Analysis of milk from cephalexin-treated animals
Finally, milk samples from four cephalexin-treatedanimals (represented by I, II, III, and IV) were suppliedby Ikertek Diagnostics SL (Spain). For those sampleswith cephalexin concentrations higher than 30 �g/l, anappropriate dilution of the sample was performed. InFig. 6, the determined concentrations of cephalexin
by the proposed method and the ELISA are comparedin a column graph. A good agreement between bothmethods is observed.
3.7. Quality control and analyte stabilityduring storage
Routine analysis quality control was performedthrough the analysis of a milk sample spiked byknown amount (10 �g/l) of cephalexin using the pro-posed method. Fig. 7 shows the determined values forsuch a sample on seven consecutive days. As can beseen, no drift or systematical changes in the concen-trations determined was observed. Moreover, thesefindings indicate that the analyte was stable duringthe storage at −20◦C.
Z.-L. Zhi et al. / Analytica Chimica Acta 442 (2001) 207–219 217
Fig. 5. Correlation between the determined cephalexin concentrations and the spiked true values for 10 milk samples. Each data pointis a mean value of two blind single determinations. Error bars represent the error of both measurements, the dotted lines indicate theconfidence interval of 95%.
Fig. 6. Comparison of the concentrations in natural contaminated milk samples determined by the proposed method and ELISA. Thesamples with concentrations higher than 30 �g/l were appropriately diluted prior to analysis.
218 Z.-L. Zhi et al. / Analytica Chimica Acta 442 (2001) 207–219
Fig. 7. Influence of storage stability on the cephalexin determination (quality control milk sample containing spiked 10 �g/l of the analyte)on several days of storage at −20◦C.
4. Conclusions
The proposed method is a fast and reliable methodfor the quantitative determination of cephalexin in con-centrations from 3 to 30 �g/l milk. Sample pretreat-ment is simple.
According to cross-reactivity testing and the mea-surements of real milk samples, the method is robustto interferences of other antibiotics and variations inthe milk matrix, even when the sample contains highviable counts or a high number of somatic cells.
The method can well discriminate suspectedfrom unsuspected samples with low possibility offalse-negative or false-positive results when a con-centration cut-off value range of 2.0–3.7 �g/l wasfixed. Recovery values obtained for as well spikedsamples as natural contaminated samples are good.The repeatability looks well (<10%).
Acknowledgements
The authors acknowledge the European Union forthe financial support of this work within the projectFAIR CT96-1181. Z.L. Zhi gratefully appreciates a
DAAD-K.C. Wong fellowship from the German Aca-demic Exchange Service (DAAD).
References
[1] G.O. Korsrud, C.D.C. Salisbury, A.E.C. Fesser, J.D. MacNeil,in: A.K. Agarwal (Ed.), Analysis of Antibiotic Residues inFood Products of Animal Origin, Plenum Press, New York,1992, pp. 75–79.
[2] G. Suhren, M. Luitz, Milchwissenschaft 50 (1995) 467.[3] E. Maertlbauer, Milchwirtsch. Ber. Bundesanst. Wolfpassing
Rotholz 124 (1995) 115.[4] L.H. Stanker, S. Buckley, M. Muldoon, W.A. Moats, C.
Braswell, Food Agric. Immunol. 10 (1998) 121.[5] A. Sternesjo, G. Johnsson, J. Food Prot. 61 (1998) 808.[6] G. Chiappetta, L. Gandino, Ind. Aliment. (Pinerolo, Italy) 37
(1998) 455.[7] A. Kumar, R.M. Rocco, D.K. Leung, L.S. Jang, S. Kharadia,
C. Yu, K.K. Hara-Mikami, G.M. Jang, M. Piani, ACSSymp. Ser. 621 (1996) 450.
[8] A. Sternesjo, G. Johnsson, Int. Dairy Fed. Spec. Issue 9505(1995) 264.
[9] M. Mitchell, B. Bodkin, J. Martin, J. Food Prot. 58 (1995)577.
[10] E. Usleber, E. Hensler, K. Doetsch, E. Maertlbauer, G.Terplan, Arch. Lebensmittelhyg 45 (1994) 80.
[11] E. Usleber, M. Lorber, M. Straka, G. Terplan, E. Maertlbauer,Analyst (Cambridge, UK) 119 (1994) 2765.
Z.-L. Zhi et al. / Analytica Chimica Acta 442 (2001) 207–219 219
[12] A. Grether, P. Hammer, W. Heeschen, Kiel. Milchwirtsch.Forschungsber. 46 (1994) 101.
[13] E. Lochbihler, E. Usleber, G. Terplan, G. Engelhardt, P.Wallnoefer, Arch. Lebensmittelhyg 46 (1995) 60.
[14] G. Suhren, J. Reichmuth, H.G. Walte, Milchwissenschaft 51(1996) 269.
[15] C. Bell, J.R. Rhoades, P. Neaves, D. Scannella, Neth. MilkDairy J. 49 (1995) 15.
[16] H.M. Chandler, PCT Int. Appl., WO 9638720.[17] R. Harik-Khan, W.A. Moats, ACS Symp. Ser. 636 (1996)
96.
[18] R. Harik-Khan, W.A. Moats, J. AOAC Int. 78 (1995) 978.[19] S.A. Barker, A.R. Long, J. AOAC Int. 77 (1994) 848.[20] IKERTEK Diagnostics SI (Spain), Fair Project Report
CT96-1181, Norwich, March 2000.[21] U.J. Meyer, Z.L. Zhi, E. Loomans, F. Spener, M. Meusel,
Analyst 124 (1999) 1605.[22] D. Trau, T. Theuerl, M. Wilmer, M. Meusel, F. Spener,
Biosens. Bioelectron. 12 (1997) 499.[23] International Dairy Federation Standard 135A, Milk and Milk
Products. Precision Characteristics of Analytical Methods —Outline of Collaborative Study Procedure, 1988, pp. 1–10.
