Multi-residue screening of veterinary drugs in egg, fish and meat using high-resolution liquid chromatography accurate mass time-of-flight mass spectrometry

Journal of Chromatography A, 1216 (2009) 8206–8216

Contents lists available at ScienceDirect

Journal of Chromatography A journal homepage: www.elsevier.com/locate/chroma

Multi-residue screening of veterinary drugs in egg, fish and meat using high-resolution liquid chromatography accurate mass time-of-flight mass spectrometry R.J.B. Peters a,∗ , Y.J.C. Bolck a , P. Rutgers a , A.A.M. Stolker a , M.W.F. Nielen a,b a b

RIKILT – Institute of Food Safety, Wageningen UR, Bornsesteeg 45, P.O. Box 230, 6700 AE Wageningen, The Netherlands Wageningen University, Laboratory of Organic Chemistry, Dreijenplein 8, 6703 HB Wageningen, The Netherlands

a r t i c l e

i n f o

Article history: Available online 16 April 2009 Keywords: Multi-residue Multi-matrix Screening method Validation 2002/657/EC High-resolution liquid chromatography Time-of-flight mass spectrometry

a b s t r a c t The last 2 years multi-compound methods are gaining ground as screening methods. In this study a highresolution liquid chromatography combined with time-of-flight mass spectrometry (HRLC–ToF-MS) is tested for the screening of about 100 veterinary drugs in three matrices, meat, fish and egg. While the results are satisfactory for 70–90% of the veterinary drugs, a more efficient sample preparation or extract purification is required for quantitative analysis of all analytes in more difficult matrices like egg. The average mass measurement error of the ToF-MS for the veterinary drugs spiked at concentrations ranging from 4 to 400 ␮g/kg, is 3.0 ppm (median 2.5 ppm) with little difference between the three matrices, but slightly decreases with increasing concentration. The SigmaFit value, a new feature for isotope pattern matching, also decreases with increasing concentration and, in addition, shows an increase with increasing matrix complexity. While the average SigmaFit value is 0.04, the median is 0.01 indicating some high individual deviations. As with the mass measurement error, the highest deviations are found in those regions of the chromatogram where most compounds elute from the column, be it analytes or matrix compounds. The median repeatability of the method ranges from 8% to 15%, decreasing with increasing concentration, while the median reproducibility ranges from 15% to 20% with little difference between matrices and concentrations. The median accuracy is in between 70% and 100% with a few compounds showing higher values due to matrix interference. The squared regression coefficient is >0.99 for 92% of the compounds showing a good overall linearity for most compounds. The detection capability, CC␤, is within 2 times the associated validation level for >90% of the compounds studied. By changing a few conditions in the analyses protocol and analysing a number of blank samples, it was determined that the method is robust as well as specific. Finally, an alternative validation strategy is proposed and tested for screening methods. While the results calculated for repeatability, within-lab reproducibility and CC␤ show a good comparison for the matrices meat and fish, and a reasonable comparison for the matrix egg, only 27 analyses are required to obtain these results versus 63 analysis in the traditional, 2002/657/EC, approach. This alternative is suggested as a cost-effective validation procedure for screening methods. © 2009 Elsevier B.V. All rights reserved.

1. Introduction Veterinary drugs are applied in animal husbandry for different reasons. In principle, all preparations administered to production animals may lead to residues in milk, eggs and in edible tissues as well. These residues may include the non-altered parent compound as well as metabolites and/or conjugates, and may have direct toxic-effects on consumers, e.g. allergic reactions in hypersensitive individuals, or for example antibiotics may cause problems indirectly through induction of resistant strains of bacteria. For con-

∗ Corresponding author. E-mail address: [email protected] (R.J.B. Peters). 0021-9673/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.chroma.2009.04.027

trolling the residue problem the EU has set maximum residue limits (MRLs) for a variety of veterinary drugs in milk, eggs and tissues [1,2]. For monitoring of contaminants and veterinary drugs in food items, generally an analytical strategy is recommended using a two-step approach. At first a low-cost screening method is applied, optimised to prevent false negative results, with a high sample throughput and an acceptable percentage of false positive results. Secondly, a confirmation method is used to confirm any positive results from the screening thereby preventing false positive results. While the use of multi-component screening methods in fruit and vegetables is becoming widespread among routine laboratories, this trend is also developing in the analysis of veterinary drugs in foods from animal origin [3]. Traditionally, microbiologi-

R.J.B. Peters et al. / J. Chromatogr. A 1216 (2009) 8206–8216

cal methods using sensitive microorganisms or bioassay techniques are widely used as screening methods since they offer a simple and low-cost method to provide semi-quantitative “total-residue” estimates. Physico-chemical methods have extended the possibilities for effective monitoring and liquid chromatography in combination with triple-quadrupole mass-spectrometric detection (LC–QqQ-MS) is used extensively for this purpose. However, most of these methods are limited to one class or related classes of compounds and only a few studies describe the simultaneous analysis of unrelated compound classes [4–8]. In addition, the number of analytes that can be monitored during one chromatographic run is limited to about 100 because of sensitivity or retention time window restrictions [9,10]. This paper describes the validation of a multi-component quantitative screening method based on high-resolution liquid chromatography combined with time-offlight mass spectrometry (HRLC–ToF-MS) for the analysis of about 100 veterinary drugs in eggs, fish and meat. The selected list of compounds, see Table 1, consists of the major classes of veterinary drugs that are commonly used to treat diseases in veterinary practice and represent benzimidazoles, anthelmintics, penicillines, macrolides, quinolones, sulfonamides, tetracyclines, amphenicols, nitroimidazoles, ionophores and tranquillizers and non-steroidal anti-inflammatory agents (NSAIDs). For true multi-component analyses a sensitive full mass scan MS technique like time-of-flight MS is required. These analysers provide high specificity due to both, high mass accuracy and high mass resolution and allow the reconstruction of highly selective accurate mass chromatograms of target residues in complex matrices. The advantage of a ToF-MS analyser is its ability to analyse a sample for a theoretically unlimited number of compounds and therefore the LC–ToF-MS approach is capable of screening for several hundreds of compounds with high sensitivity within one run. Furthermore, data can be acquired and reprocessed without any a priori knowledge about the presence of certain compounds; that is, no analyte-specific information is required before injecting a sample and the presence of newly identified compounds can be confirmed in previously analysed samples simply by reprocessing the data. The advantage of ToF-MS can be further improved by combining it with high-resolution LC. HRLC–ToF-MS provides significant advantages concerning selectivity, sensitivity and speed [11–14]. However, ToF-MS remains a single MS system and therefore (by definition) suitable only for screening analysis according to EU regulation 2002/657/EC [15]. In this study a Bruker micrOTOF system was used while in a previous study we used a Waters LCT Premier ToF-MS [11]. Because both studies are validations involving the same analytes the results of the new Bruker micrOTOF could to some extent be compared with those of the Waters LCT Premier. Newly developed methods have to be validated before they can be used in official control studies of foodstuffs. The validation procedure for the determination of residues in foodstuffs from animal origin is described in EU Commission Decision 2002/657/EC. Although the practical application and interpretation of the guideline still generates discussions, it is clear that a large number of sample analyses are required for a proper validation [16]. This is illustrated by studies of Hernando, Stolker and Kaufmann who all developed ToF-MS screening methods for veterinary drugs validated according to the EU guideline [11–13,17]. Typically, in the validation of a quantitative confirmation method 81 analyses are performed for the determination of CC␣, CC␤, recovery, repeatability and reproducibility. The strategy is to perform seven replicate analyses at three concentration levels on three different days. Additionally, a five-point calibration curve is included on each of these days. The number of additional samples for the determination of specificity and robustness depends on the method but a total of 26 will be the minimum, resulting in a total number of analyses of at least 119. Galarini prepared a simplified validation strategy

8207

for routine laboratories that complies with 2002/657/EC [18]. Their strategy requires 106 analyses for the determination of the performance characteristics for a quantitative confirmation method while an alternative validation strategy proposed by Antignac requires a minimum of 55 samples [19]. Since method validation generally takes a lot of time and resources, simpler validation strategies are helpful. In this study an alternative validation procedure requiring a lower number of samples to be analysed is suggested and tested. Although this alternative method could also be used for quantitative confirmation methods, it is in first instance thought of to be applied in the validation of screening methods. Results of this alternative validation are compared with those of the full validation that was in accordance with the 2002/657/EC document. 2. Materials and methods 2.1. Chemicals, reagents and solutions Veterinary drug analytical standards were purchased from Sigma–Aldrich Chemie b.v (Zwijndrecht, the Netherlands) Smith Kline Beecham (Zeist, the Netherlands), Riedel de Haen (Seelze, Germany), or Fluka (Buchs, Switzerland). Oxfendazole and oxfendazole sulfon were purchased from Syntex (Clare Castle, Co Clare, Ireland), valnemulin from Novartis (Basel, Switerzerland), marbofloxacin from Vetoquinol (New Jersey, US). Albendazole (sulfoxide), oxfendazole (sulfone), hydroxyl-ipronidazol, carazolol, piroxycam, propyphenazoe and piroxicam were obtained from Bundesamt fur Verbraucherschutz und Lebensmittelsicherheit, BVL-CRL (Berlin, Germany). 5-Hydroxy thiabendazole was obtained from The National Institute for Public Health and the Environment, RIVM-CRL (Bilthoven, the Netherlands). Individual veterinary drug stock solutions were prepared in pure methanol or acetonitrile and were stored at −20 ◦ C. The mixed standard solution is stored at −80 ◦ C and is used for 1 year. The penicillines are not very stable drugs and these compounds are added just before analysis to the mixed standard solutions. The stock standards of penicillines were stored at −80 ◦ C. HPLC-grade acetonitrile, methanol and LC–MS water (used as eluent in the HRLC–ToF-MS analysis) were obtained from Biosolve (Valkenswaard, The Netherlands). A Milli-Q-Plus ultrapure water system from Millipore (Amsterdam, The Netherlands) was used throughout the study to obtain the HPLC-grade water used during the sample preparation. Formic acid was obtained from Merck (Darmstadt, Germany). Solid phase extraction columns, type Strata-X 33 ␮m Polymeric Reversed Phase 60 mg/3 ml, were obtained from Phenomenex (Torrance, CA, USA). 2.2. Samples All samples tested were collected by the Dutch Food and Consumer Product Safety Authority, Laboratory Region East (Wageningen, The Netherlands) during 2008. All egg samples were chicken eggs, fish samples consisted of salmon, trout, mackerel, eel and catfish, and meat samples of chicken, porcine and bovine animals. The samples were received in frozen conditions and were kept frozen (−20 ◦ C) until analysis. 2.3. Instrumentation Chromatographic separation of the veterinary drugs in sample extracts was achieved using a Waters Acquity UPLC system, consisting of vacuum degasser, autosampler and binary pump, and equipped with a reversed phase Waters Acquity UPLC BEH C18 analytical column of 100 mm × 2.1 mm and 1.7 ␮m particle size. Elution solvents used were 0.1% formic acid (A) and acetonitrile/0.1% formic acid, 9:1, v/v (B) with the following gradient: 0–1 min, 0% B; 1–4 min, linear increase to 40% B; 4–10 min, linear increase to 100%

8208

R.J.B. Peters et al. / J. Chromatogr. A 1216 (2009) 8206–8216

Table 1 Veterinary drugs and performance characteristics as determined in this study. With the exception of VL, RT and CC␤, the parameters are expressed as the range of the average parameters for the three tested matrices (meat, fish and egg) at the three concentrations (0.5, 1.0 and 1.5 times VL). CC␤ is given as the range for the three tested matrices and 1.0 times VL. RTb (min)

RTc (min) Massd (ppm)

Elemental composition

VLa (␮g/kg)

Benzimidazoles Albendazole Albendazole sulfoxide Albendazole sulfone Albendazole amino sulfone Fenbendazole Oxfendazole Oxfendazole Sulfone Mebendazole Amino mebendazole Hydroxy mebendazole Levamisole Thiabendazole 5-Hydroxy thiabendazole Flubendazole Amino flubendazole Hydroxy flubendazole Oxibendazole

C12 H15 N3 O2 S C12 H15 N3 O3 S C12 H15 N3 O4 S C10 H13 N3 O2 S C15 H13 N3 O2 S C15 H13 N3 O3 S C15 H13 N3 O4 S C16 H13 N3 O3 C14 H11 N3 O C16 H15 N3 O3 C11 H12 N2 S C10 H7 N3 S C10 H7 N3 OS C16 H12 FN3 O3 C14 H10 FN3 O C16 H14 FN3 O3 C12 H15 N3 O3

100 100 100 100 20 20 20 4 4 4 4 100 100 4 4 4 4

5.2 4.0 4.6 3.5 6.0 4.5 5.2 5.2 4.4 4.2 3.5 3.5 3.2 5.5 4.2 4.4 4.5

0.04–0.06 0.02–0.07 0.01–0.04 0.04–0.07 0.05–0.05 0.09–0.12 0.03–0.09 0.05–0.07 0.11–0.12 0.04–0.05 0.03–0.06 0.03–0.05 0.03–0.05 0.03–0.08 0.04–0.06 0.06–0.13 0.04–0.06

1.2–1.8 1.5–1.8 2.3–4.5 1.7–3.1 2.6–3.1 3.9–4.4 2.3–2.6 1.6–2.1 2.5–3.4 2.1–5.8 2.5–3.9 2.4–3.0 3.8–4.6 2.0–4.3 3.1–4.6 5.7–8.1 2.3–3.7

0.004–0.006 0.003–0.004 0.003–0.026 0.012–0.056 0.007–0.029 0.005–0.010 0.004–0.009 0.016–0.064 0.030–0.154 0.011–0.028 0.007–0.015 0.004–0.010 0.003–0.005 0.008–0.077 0.025–0.071 0.007–0.019 0.008–0.061

0.992–0.994 0.996–0.997 0.996–0.998 0.992–0.997 0.993–0.997 0.997–0.998 0.998–0.999 0.995–0.999 0.994–0.996 0.996–0.998 0.993–0.999 0.987–0.997 0.990–0.998 0.993–0.998 0.994–0.995 0.993–0.997 0.994–0.995

93–99 95–110 95–114 96–112 99–113 91–110 96–101 93–97 87–96 89–106 86–109 97–115 94–106 93–110 86–92 91–101 90–96

12–26 5–11 4–14 7–8 11–30 5–14 11–14 13–17 12–26 6–14 9–13 7–9 8–8 13–25 13–17 10–16 9–16

19–50 5–11 5–17 8–14 26–43 4–17 14–15 16–21 12–51 5–14 11–16 10–21 9–12 17–25 15–29 15–19 16–20

162–263 117–135 115–157 127–146 37–48 22–31 28–30 6.1–6.0 5.5–10 4.7–5.9 5.5–6.1 132–169 131–140 6.3–7.3 5.9–7.8 6–6.5 6.1–6.7

Macrolides Erythromycin Spiramycin Tilmicosin Tylosin Josamycin Tiamulin Valnemulin Lincomycine Pirlimycine Tulatromycine

C37 H67 NO13 C43 H74 N2 O14 C46 H81 N2 O13 C46 H77 NO17 C42 H70 NO15 C28 H47 NO4 S C31 H52 N2 O5 S C18 H34 N2 O6 S C17 H31 ClN2 O5 S C41 H79 N3 O12

160 100 100 100 100 100 100 100 100 100

5.0 4.2 4.6 5.2 5.8 5.4 5.7 3.4 4.3 3.7

0.07–0.07 0.08–0.08 0.05–0.10 0.02–0.06 0.05–0.09 0.05–0.10 0.06–0.08 0.04–0.06 0.04–0.05 0.04–0.10

1.5–2.0 1.4–2.5 2.2–3.0 2.3–2.9 1.8–2.5 0.9–1.8 0.9–1.1 2.0–3.6 2.8–3.5 2.0–3.3

0.083–0.107 0.025–0.113 0.010–0.020 0.012–0.067 0.014–0.040 0.008–0.013 0.007–0.008 0.064–0.254 0.009–0.010 0.055–0.125

0.993–0.998 0.992–0.997 0.990–0.993 0.971–0.995 0.671–0.998 0.820–0.998 0.993–0.997 0.985–0.997 0.998–0.999 0.918–0.996

69–105 86–102 94–128 85–99 101–203 86–261 84–97 86–120 93–100 83–123

5–21 14–17 14–36 15–34 8–16 9–75 9–22 7–15 5–11 19–58

9–32 17–21 18–29 15–42 13–55 14–142 12–34 12–20 6–15 29–66

205–326 155–169 160–194 150–238 142–281 145–566 140–210 138–164 118–148 193–316

Penicillines Penicilline V Penicilline G Ampicillin Oxacillin Nafcillin Cloxacillin Dicloxacillin

C16 H18 N2 O5 S C16 H18 N2 O4 S C16 H19 N3 O4 S C19 H19 N3 O5 S C21 H23 N2 O5 S C19 H18 ClN3 O5 S C19 H18 Cl2 N3 O5 S

100 100 100 400 400 400 400

5.7 3.8 3.6 6 6.4 6.3 6.7

0.02–0.05 0.06–0.08 0.03–0.06 0.02–0.07 0.05–0.07 0.02–0.08 0.06–0.11

3.4–4.1 1.4–2.3 5.1–7.8 1.7–2.8 1.2–2.3 0.9–1.6 1.2–1.5

0.014–0.021 0.006–0.034 0.095–0.187 0.004–0.007 0.004–0.020 0.006–0.007 0.011–0.018

0.971–0.998 0.996–0.999 0.963–0.999 0.994–0.996 0.994–0.998 0.991–0.997 0.990–0.999

95–143 102–125 100–112 91–123 70–92 88–126 91–114

8–29 7–11 11–12 8–14 11–16 10–13 9–10

10–40 10–20 17–17 12–16 14–18 14–21 9–13

133–230 134–164 155–156 555–607 577–631 585–679 518–576

Quinolones Nalidixic acid Oxolinic acid Flumequine Norfloxacin Ciprofloxacin Lomefloxacin Enrofloxacin Marbofloxacin Difloxacin Danofloxacin Sarafloxacin

C12 H12 N2 O3 C13 H11 NO5 C14 H12 FNO3 C16 H19 FN3 O3 C17 H18 FN3 O3 C17 H20 F2 N3 O3 C19 H22 FN3 O3 C17 H20 FN4 O4 C21 H20 F2 N3 O3 C19 H21 FN3 O3 C20 H17 F2 N3 O3

4 4 100 50 10 10 100 100 4 60 10

5.5 4.8 5.6 3.6 3.7 3.8 3.8 3.6 4.1 3.8 4

0.03–0.08 0.04–0.08 0.03–0.07 0.07–0.09 0.05–0.10 0.05–0.07 0.06–0.11 0.03–0.09 0.05–0.13 0.03–0.08 0.05–0.06

2.9–10 1.7–2.6 1.4–2.7 2.1–3.3 1.7–3.7 3.1–4.6 2.4–3.5 1.7–2.5 3.1–4.7 2.2–3.5 1.7–3.2

0.014–0.027 0.043–0.183 0.003–0.010 0.009–0.105 0.056–0.111 0.027–0.033 0.018–0.055 0.005–0.013 0.019–0.033 0.108–0.117 0.017–0.038

0.996–0.999 0.996–0.999 0.994–0.999 0.997–0.999 0.997–0.999 0.997–0.999 0.993–0.997 0.997–1.000 0.994–0.998 0.996–0.998 0.996–0.999

96–109 96–107 100–119 97–112 97–106 96–105 89–107 95–110 88–112 93–103 94–113

8–10 6–11 6–8 8–11 8–10 7–11 6–21 8–10 7–13 8–9 7–12

9–15 6–15 8–11 9–13 9–16 10–14 7–24 8–12 9–14 9–11 6–16

5.2–6.0 4.8–5.9 127–136 64–70 13–15 13–14 122–180 125–140 5.2–5.9 78–81 12–15

Sulphonamides Dapsone Sulphadiazine Sulphamethoxazole Sulphamethazine Sulphadimethoxine Sulphadoxine Sulphaquinoxaline Sulphachloropyridazine Sulphamerazine Sulphamethizole Sulphamethoxypyridazine Sulphamonomethoxine Sulphamoxole Sulphapyridine Sulphisoxazole Sulphathiazole

C12 H12 N2 O2 S C10 H10 N4 O2 S C10 H11 N3 O3 S C12 H14 N4 O2 S C12 H14 N4 O4 S C12 H14 N4 O4 S C14 H12 N4 O2 S C10 H9 ClN4 O2 S C11 H12 N4 O2 S C9 H10 N4 O2 S2 C11 H12 N4 O3 S C11 H12 N4 O3 S C11 H13 N3 O3 S C11 H11 N3 O2 S C11 H13 N3 O3 S C9 H9 N3 O2 S2

10 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100

4.3 3.5 4.6 4 5 4.6 5 4.4 3.8 4 4.3 4.1 3.9 3.7 4.7 3.6

0.06–0.10 0.02–0.07 0.03–0.07 0.05–0.09 0.05–0.05 0.03–0.08 0.04–0.08 0.05–0.12 0.02–0.08 0.05–0.11 0.03–0.05 0.02–0.07 0.07–0.11 0.02–0.07 0.05–0.06 0.01–0.07

1.9–3.3 1.4–3.2 2.8–5.1 1.7–3.7 2.7–3.7 2.5–3.0 2.5–3.1 2.1–3.8 1.4–1.9 2.0–2.5 2.5–3.2 4.4–8.3 1.6–2.1 1.3–1.5 1.5–3.2 1.2–1.6

0.019–0.042 0.008–0.135 0.004–0.046 0.302–0.315 0.005–0.029 0.004–0.004 0.004–0.005 0.013–0.025 0.005–0.007 0.006–0.074 0.003–0.009 0.023–0.054 0.005–0.031 0.003–0.004 0.005–0.208 0.005–0.005

0.996–0.997 0.978–0.996 0.982–0.997 0.998–0.999 0.993–0.999 0.997–0.999 0.994–0.999 0.976–0.999 0.995–0.997 0.976–0.996 0.989–0.999 0.998–0.999 0.989–0.999 0.995–0.998 0.996–0.998 0.992–0.998

90–102 89–108 95–98 94–115 97–101 95–102 96–103 93–97 93–102 92–118 93–99 83–109 87–107 99–115 83–92 88–95

9–20 7–15 11–20 7–11 6–15 8–15 4–18 8–16 6–13 10–14 8–15 9–19 11–15 6–11 10–13 8–11

18–26 9–18 18–23 7–15 10–18 15–18 10–23 13–22 9–17 15–27 12–17 10–30 16–19 6–14 13–22 12–14

15–18 130–159 158–175 123–149 133–160 147–159 132–176 143–173 128–155 147–188 137–155 132–198 151–163 119–147 142–171 139–147

Pyrimidine Trimethoprim

C14 H18 N4 O3

50

3.6

0.07–0.09

2.1–7.3

0.005–0.123

0.997–0.998

95–109

6–8

6–8

60–62

Compound

SigmaFite

rb , f

Accuracyg (%)

rs h (%)

Rs i (%)

CC␤j (␮g/kg)

R.J.B. Peters et al. / J. Chromatogr. A 1216 (2009) 8206–8216

8209

Table 1 (Continued ) RTb (min)

RTc (min)

Massd (ppm)

200 200 200 200

3.9 4.5 3.7 4.4

0.03–0.05 0.05–0.06 0.05–0.09 0.02–0.05

1.9–3.5 2.4–3.2 1.7–3.0 1.3–3.4

0.005–0.033 0.005–0.019 0.037–0.049 0.012–0.117

0.984–0.996 0.985–0.998 0.983–0.996 0.974–0.997

84–109 63–92 81–88 75–120

9–18 6–23 11–21 15–33

14–19 8–52 17–24 20–42

289–322 251–538 314–358 332–478

C7 H11 N3 O2 C7 H11 N3 O3

10 10

4.8 4.1

0.03–0.08 0.06–0.09

2.4–3.3 1.6–4.1

0.020–0.048 0.053–0.093

0.975–0.996 0.996–0.999

84–141 84–128

14–30 7–14

22–49 10–15

17–26 13–14

Tranquillizers Azaperol Azaperone Propionylpromazine Acetopromazine Xylazine Haloperidol Chloropromazine Carazolol

C19 H24 FN3 O C19 H22 FN3 O C20 H24 N2 OS C19 H22 N2 OS C12 H16 N2 S C21 H23 ClFNO2 C17 H19 ClN2 S C18 H22 N2 O2

1.6 2 10 2 2 2 4 10

3.7 3.9 5.5 5.1 4.1 5.1 5.7 4.5

0.06–0.08 0.11–0.15 0.05–0.10 0.09–0.09 0.08–0.14 0.12–0.12 0.05–0.06 0.03–0.05

2.2–5.6 5.7–8.3 1.4–5.0 3.6–8.2 4.4–6.4 2.2–6.4 2.0–6.2 1.8–2.4

0.093–0.122 0.046–0.072 0.004–0.146 0.038–0.222 0.022–0.031 0.026–0.071 0.033–0.068 0.010–0.028

0.993–0.995 0.992–0.995 0.940–0.995 0.992–0.996 0.994–0.998 0.979–0.998 0.992–0.993 0.996–0.998

92–107 88–94 71–230 83–162 89–104 94–101 74–341 90–98

6–12 13–16 12–23 11–24 8–12 11–30 0–23 9–13

7–13 16–27 20–45 13–29 9–15 19–46 0–36 9–17

1.9–2.3 3.0–3.8 15–101 2.9–3.9 2.6–3.0 3.3–5.0 4.0–8.8 13–15

NSAIDs Piroxycam Propyphenazone Indoprofen Tolmetin Ketoprofen Naproxen Fenbufen Carprofen Diclofenac Niflumic acid Phenylbutazone Flufenamic acid Mefenamic acid Meclofenamic acid Isopyrin Isoxicam Tenoxicam Sulindac Indomethacin Flunixin Meloxicam Fenoprofen Tolfenamic acid

C15 H13 N3 O4 S C14 H18 N2 O C17 H15 NO3 C15 H15 NO3 C16 H14 O3 C14 H14 O3 C16 H14 O3 C15 H12 ClNO2 C14 H11 Cl2 NO2 C13 H9 F3 N2 O2 C19 H20 N2 O2 C14 H10 F3 NO2 C15 H15 NO2 C14 H11 Cl2 NO2 C14 H19 N3 O C14 H13 N3 O5 S C13 H11 N3 O4 S2 C20 H17 FO3 S C19 H16 ClNO4 C14 H11 F3 N2 O2 C14 H13 N3 O4 S2 C15 H14 O3 C14 H12 ClNO2

100 40 40 100 100 200 100 500 100 100 100 100 100 100 100 100 100 100 100 100 30 100 200

5.7 5.8 5.8 6.2 6.4 6.4 6.6 7.2 7.5 7.4 7.8 8 8.1 8.1 3.5 6.7 4.7 5.8 7.5 6.6 6.5 7.1 8.3

0.04–0.07 0.05–0.06 0.04–0.07 0.03–0.09 0.03–0.08 0.05–0.10 0.04–0.05 0.02–0.07 0.02–0.07 0.08–0.14 0.02–0.06 0.09–0.12 0.08–0.14 0.09–0.15 0.04–0.05 0.04–0.12 0.03–0.09 0.06–0.13 0.05–0.1 0.05–0.12 0.02–0.09 0.06–0.1 0.09–0.17

2.2–2.8 1.9–2.7 1.4–2.2 1.7–2.4 1.3–2.9 1.9–5.2 1.9–3.4 2.0–2.5 1.6–2.5 1.3–1.6 5.3–7.7 1.8–2.5 1.7–2.6 2.4–4.1 2.2–6.7 1.0–2.0 1.6–2.4 2.2–3.3 1.4–2.2 2.6–2.7 2.4–3.3 1.7–6.3 1.8–2.4

0.005–0.006 0.004–0.037 0.006–0.013 0.014–0.027 0.073–0.313 0.010–0.012 0.042–0.250 0.020–0.048 0.019–0.047 0.004–0.005 0.028–0.152 0.060–0.104 0.011–0.049 0.046–0.130 0.015–0.109 0.004–0.019 0.008–0.071 0.004–0.012 0.016–0.054 0.003–0.005 0.005–0.006 0.028–0.053 0.019–0.063

0.998–0.999 0.995–0.999 0.987–0.999 0.993–0.999 0.995–0.998 0.994–0.995 0.992–0.997 0.970–0.995 0.976–0.992 0.995–0.998 0.924–0.979 0.995–0.997 0.992–0.996 0.924–0.999 0.979–0.997 0.994–0.998 0.964–0.999 0.994–0.996 0.990–0.996 0.995–0.999 0.992–0.998 0.995–0.996 0.949–0.997

97–1274 95–119 96–121 105–114 98–103 95–170 93–115 62–111 87–104 99–103 32–76 73–110 66–104 67–100 77–98 94–111 47–97 94–100 69–105 98–111 94–108 82–99 93–106

8–55 3–7 8–17 8–19 10–11 12–21 12–33 23–46 23–42 8–13 27–81 15–63 19–52 20–50 12–18 9–12 10–102 12–20 14–63 3–8 11–22 12–27 19–59

10–110 3–9 11–22 8–22 12–17 16–68 14–52 46–99 33–81 10–25 43–95 25–86 33–72 32–50 15–29 9–21 14–92 14–20 24–85 6–13 20–22 18–41 32–62

131–459 44–52 53–68 125–172 138–155 307–644 145–269 1252–2119 207–365 131–182 241–413 181–381 209–336 206–264 150–194 129–168 145–401 145–164 179–380 119–142 49–51 159–234 410–606

Coccidiostat Robenidine

C15 H13 Cl2 N5

100

6

0.06–0.09

1.7–5.1

0.018–0.114

0.931–0.989

76–138

19–53

29–96

196–415

Ionophores Monensin Na Salinomycine Na Narasin Na Semduramycin Na Lasalocid Na

C36 H61 O11 -Na C42 H69 O11 -Na C43 H72 O11 -Na C45 H75 O16 -Na C34 H53 O8 -Na

50 50 50 50 80

11.5 11.3 12.2 10.6 12.2

0.05–0.13 0.03–0.08 0.05–0.08 0.03–0.07 0.03–0.08

1.6–2.1 1.5–2.4 2.2–2.9 1.9–2.2 2.8–6.7

0.029–0.079 0.040–0.109 0.054–0.086 0.013–0.039 0.06–0.172

0.984–0.993 0.957–0.993 0.806–0.986 0.95–0.986 0.37–0.37

113–563 51–133 16–115 53–190 101–143

14–63 37–73 27–44 21–66 158–205

41–126 65–119 37–37 36–173 149–170

116–256 156–245 50–111 109–333 470–526

Amphenicols Florphenicol Thiamphenicol

C12 H14 Cl2 FNO4 S C12 H15 Cl2 NO5 S

50 100

4.6 3.9

0.07–0.11 0.06–0.11

2.5–3.4 2.7–3.5

0.053–0.068 0.031–0.07

0.996–0.997 0.994–0.996

94–115 89–121

11–18 4–17

14–21 9–21

73–83 129–169

Elemental composition

VLa (␮g/kg)

Tetracyclines Tetracycline Doxycycline Oxytetracycline Chlorotetracycline

C22 H24 N2 O8 C22 H24 N2 O8 C22 H24 N2 O9 C22 H23 ClN2 O8

Nitroimidazolen Ipronidazol Hydroxy-ipronidazol

Compound

a b c d e f g h i j

SigmaFite

rb , f

Accuracyg (%)

rs h (%)

Rs i (%)

CC␤j (␮g/kg)

VL: validation level is the concentration spiked to samples in ␮g/kg, validation was carried out 0.5, 1.0 and 1.5 times VL. RT: retention time in minutes as determined in chromatograms of analyte standard solutions. RT: deviation from the expected RT of the analyte. Mass: mass measurement error relative to the calculated exact mass of the analyte. SigmaFit: numerical expression for the difference in the measured and calculated isotope pattern of the analyte. r2 : squared regression coefficient. Accuracy: the recovery of the spiked analyte concentration expressed as a percentage. Repeatability expressed as the relative standard deviation in seven analyses performed on 1 day. Within-lab reproducibility expressed as the relative standard deviation in the three series of seven analyses on 3 days performed by different analysts. CC␤: detection capability calculated from the within-lab reproducibility at the validation level 1.0 times VL.

B with a final hold for 2 min with a flow rate of 0.4 ml/min. The injection volume was 20 ␮l. The UPLC system was interfaced splitless to a Bruker Daltonics micrOTOF mass spectrometer equipped with an orthogonal electrospray ionisation (ESI) source. The instrument was operated in the positive ion mode using a mass range of 100–1000 Da. The trigger time was 33 ␮s and 10,000 spectra were

summed up equalling 0.33 s time resolution. The capillary voltage of the ion source was set at 3500 V and the capillary exit at 100 V. The nebulizer gas pressure was 1.5 l/min and drying gas flow 8 l/min. The drying temperature was set at 200 ◦ C. Instrument calibration was performed externally prior to each sequence with a sodium formate/acetate solution,

8210

R.J.B. Peters et al. / J. Chromatogr. A 1216 (2009) 8206–8216

consisting of 3.3 mM sodium hydroxide in a mixture of water/isopropanol/formic acid/acetic acid (1:1:1:3, v/v). The theoretical exact masses of calibration ions with formulae Na(HCO2 Na)2–8 and Na(CH3 CO2 Na)2–8 (HCO2 Na)2–8 in the range of 100–1000 Da were used for calibration. Automated post-run internal mass scale calibration of individual samples was performed by injecting the calibrant at the beginning and at the end of each run via a six-port divert valve equipped with a 20 ␮l loop. Calibration was performed based on calibrant injection at the beginning of the run. The calibrant at the end of the run was for manual verification of calibration stability. The calibrator ions in the post-run internal mass scale calibration were the same as in the instrument calibration. 2.4. Sample preparation Laboratory samples received within the framework of an annual monitoring program were homogenised using a blender and stored in a freezer. Typically, an analytical sample of 2 g of the homogenised laboratory sample was collected, mixed with 10 ml of an acetonitrile/Milli-Q water mixture (6:4, v/v) and shaken manually to effect protein precipitation. After an intensive shaking period of 30 min, the samples were centrifuged for 15 min (3600 g at T = 10 ◦ C). From the supernatant 3 ml was collected and diluted to 60 ml with Milli-Q water. The diluted sample extract was applied to a StrataX SPE column (60 mg) that was conditioned with 3 ml of methanol followed by 3 ml of water. Following passage of the sample extract the column was washed with 3 ml water. In case of egg samples the column was eluted with 3 ml of a methanol/ethyl acetate mixture (1:1, v/v), for fish and meat samples elution was achieved with 3 ml of a methanol/acetonitrile mixture (1:1, v/v). The eluate was evaporated under a stream of nitrogen at 40 ◦ C till near dryness and the residue was re-dissolved in 25 ␮l of acetonitrile and vortexed for 30 s. Finally, 225 ␮l of 0.1% formic acid in Milli-Q water was added and the extract was analysed by HRLC–ToF-MS in the full scan mode. 2.5. Quantification After acquisition the specific [M+H]+ ions (see Table 1) were extracted from the spectra using the Target Analysis software supplied by Bruker. For all compounds a standard extraction window width of 20 ppm was used. Identification was based on retention time and accurate mass relative to external standards. Calibration curves were calculated from the response of matrix matched standards (MMS), e.g. six blank samples fortified with five concentrations of each specific drug. The MMS were analysed on each of the 3 days and in the same series as the actual samples and sample concentrations were calculated using the linear regression method.

an expected limited sensitivity of the ToF detector for that specific drug. On the other hand, for other drugs the VL was set below the MRL or MRPL level of interest to prevent overloading of the analytical column. Blank samples were fortified at 0.5, 1.0 and 1.5 times the VL level with all drugs of interest and seven replicates of each sample were analysed on 1 day. The procedure was repeated on 2 additional days. From the data the repeatability, intra-laboratory reproducibility (both expressed as the relative standard deviation, RSD) and accuracy were calculated. Guidelines for acceptable repeatability and the intra-laboratory reproducibility values at different analyte concentrations were adopted from the 2002/657/EC document. The accuracy is expressed as the average recovery in the samples at the VL level and a range of 70–120% was considered acceptable for a multi-compound quantitative screening as in this study. The linearity was determined for a concentration range of 0, 0.5, 1, 1.5, 2 and 2.5 times the VL level. On each validation day the calibration curves were constructed and the squared regression coefficients (r2 ) calculated for each compound. Squared regression coefficients >0.99 were considered acceptable. The detection capability (CC␤) at the VL was determined from decision limit CC␣ which in turn was calculated from the standard deviation at the VL level using the following equations: CC␤ = CC␣ + 1.64 × SDVL

and

CC␣ = VL + 1.64 × SDVL

The robustness of the method was tested by analysing four samples of each matrix in duplicate, choosing four slightly different sample pre-treatment/extraction procedure for each duplicate. The first duplicate was analysed by using the developed procedure while for the second duplicate the extraction time was reduced from 30 to 15 min. For the third duplicate the SPE column was run dry for 10 min after application of the aqueous extract and before the wash step. For the fourth duplicate the final residue after evaporation of the solvent was left dry for 30 min before re-dissolving. The method is considered robust when the RSD of a specific compound within these eight analyses is smaller than, or equal to the RSD of the corresponding intra-laboratory reproducibility. The specificity of the method was checked by the analysis of 20 blank samples for each matrix. The chromatograms were monitored for peaks that can potentially interfere with the analytes of interest. Stability experiments were not performed in this study since these data were already available from other in-house studies. All compounds of interest were stable for at least 1 month with the exception of the penicillin’s. 3. Results 3.1. Sample pre-treatment

2.6. Validation The developed method was validated based on the procedure described in Ref. [11] which is in accordance with EU Commission Decision 2002/657/EC [15] for a quantitative screening method. The validation study of the veterinary drugs in egg, fish and meat was carried out at three concentration levels chosen around a validation level (VL). This validation level was equal to the MRL level, the minimum required performance limit (MRPL) level, or to a “specific level of interest” if no MRL or MRPL level was available, as is the case for many of these drugs in these matrices. This specific level of interest was based on the drug characteristics, e.g. class of compounds, or based on the MRL or MRPL level of the specific drug in other matrices. All concentration levels used in this validation study are described in Table 1. For some compounds the VL was set somewhat higher than the MRL or MRPL level of interest due to

A first necessity for this multi-compound analysis is the development of a generic sample preparation method suited for the extraction of veterinary drugs from different food matrices of animal origin. Preparation/extraction methods for the multicompound determination of veterinary drugs in milk, meat, fish and feed have been described in the literature and are in use in our laboratory [11,13,14,17,20]. In the first four references, an extraction using acetonitrile is used followed by an SPE clean-up procedure. Mol et al. [14] tested a number of solvents and combinations thereof, and while the most suitable solvent depended on the matrix, extraction with water/acetonitrile was chosen as the default method. In a recent publication Noguchi determined veterinary drugs in livestock food and seafood, again using an extraction with acetonitrile followed by an SPE clean-up procedure [21]. That many authors prefer acetonitrile over methanol or

R.J.B. Peters et al. / J. Chromatogr. A 1216 (2009) 8206–8216

8211

Fig. 1. Typical HRLC–ToF-MS total ion chromatogram of a blank sample of meat (top), a standard mixture in solvent (second) and a selection of extracted ion chromatograms from an extract of a meat sample spiked with 100 veterinary drugs at the validation level (1 × VL). Shown are the extracted mass spectra with a mass window of 0.01 Da.

ethyl acetate is not surprising since the latter extracts too many matrix compounds, complicating further clean-up procedures and making direct extract analysis using HRLC–ToF-MS difficult. Furthermore, acetonitrile affords protein precipitation which is a first clean-up step in case of matrices of animal origin. The only different choice was made by Xiaofei who tested methanol, acetonitrile and McIlvaine-EDTA buffer for the simultaneous extraction of tetracyclines and quinolones from eggs and as expected found that McIlvaine-EDTA buffer showed the highest recoveries for these analytes [22]. In this method acetonitrile/water combinations in different ratios were tested for all three matrices. A 6:4 acetonitrile/water ratio was selected, which, taking into account the water and protein contents, results in conditions comparable to the extraction of milk in the previous study [11]. Ultra-filtration and polymer based C18 -SPE columns (Oasis and StrataX) were tested for clean-up of the extracts. While ultra-filtration resulted in losses of analytes the primary extracts had to be diluted in water to avoid breakthrough during the SPE step. It is obvious that the use of a C18 -SPE clean-up will lead to losses of some (very polar) analytes. On the other hand, the SPE step leads to a concentration of analytes enabling lower detection limits for prohibited compounds. While the analytes in fish and meat extracts could be eluted from the SPE column using 3 ml of a 1:1 methanol/acetonitrile mixture, for egg this resulted in low recoveries for a number of analytes. The recoveries for egg improved when the elution was carried out with 3 ml of a 1:1 methanol/ethyl acetate which, however, also resulted in a less efficient clean-up. That a stronger eluent is needed in the case of eggs is probably a result of protein binding of analytes to egg proteins on the SPE column [23]. Egg proteins (albumin) are smaller and more water-soluble than proteins in fish and meat (myosin) and probably less efficiently precipitated by acetonitrile resulting in a higher protein loading on the SPE column and subsequent interference with the adsorbed analytes. Attempts to remove the interfering proteins by salting out of proteins by the addition of ammonium sulphate

during extraction, and after extraction but before centrifugation, did not improve the results [24]. 3.2. HRLC–ToF-MS screening method To develop the screening method, a solvent-based standard with all veterinary drugs of interest was analysed. Based on the chromatographic retention times, the specific accurate masses and the isotope pattern ratios calculated by the software from the elemental composition of the veterinary drugs, the method is constructed. The isotope pattern matching algorithm (SigmaFit) is a feature of the Bruker Daltonics micrOTOF that can be used as an additional identification tool to accurate mass measurement. In the method the different combinations of retention time, accurate mass and SigmaFit value are defined with their acceptable tolerances. Since the method was intended as a multi-matrix screening method no drug specific parameter windows were defined for the accurate mass and SigmaFit but a standard extraction window of 20 ppm and 0.5 Sigma were used, respectively. For reasons unknown the Sigma window had to be set at a relatively high value because otherwise even compounds with a smaller SigmaFit value were identified correctly. Following analysis of a real sample the full scan chromatogram is processed and a list of detected compounds (included in the screening method) is generated by the Target Analysis software. The final result list was produced with an in-house developed MS Excel based script [25]. Table 1 presents the retention times of the target analytes in the chromatogram. A chromatogram of a blank sample and that of a standard mixture of the compounds in solvent are the top chromatograms in Fig. 1, showing that most analytes elute in the region of 3–8 min while most matrix compounds elute after 8 min. Note that the matrix loading is relatively high in this method, since an absolute amount of 0.5 g sample is concentrated to a final volume of 250 ␮l from which 20 ␮l is injected, corresponding to approximately 80 mg of matrix equivalent. Taking into account the relative

8212

R.J.B. Peters et al. / J. Chromatogr. A 1216 (2009) 8206–8216

Fig. 2. Median values of deviations in retention time (a), mass measurement error (b), and SigmaFit value (c) in relation to matrix and validation level.

amounts of the organic constituents in the matrices (protein, carbohydrates and fat) the amount of matrix is about 4 times higher than in the case of milk in the previous study. This especially is the case for eggs were the more water-soluble proteins probably result in a higher matrix loading which is not only visible in the chromatograms but also in the retention time deviations. These deviations in the chromatograms of egg, fish and meat samples spiked at concentrations of 0.5, 1.0 and 1.5 VL ranged from −0.26 to 0.20 min with the median of the absolute deviations ranging from 0 to 0.09 min. As shown in Fig. 2 the retention time deviations depend on concentration and especially on the matrix, with the highest deviations found for the lowest concentration in eggs. Probably the

relatively high matrix loadings interfere with the chromatography on the analytical column resulting in the relatively high observed retention deviations for the lowest standard. Table 1 also presents the accurate mass measurement data obtained in the sample extracts. In general for a ToF-MS having a mass resolution of ∼10,000 FWHM and an external calibration, a deviation of the measured accurate mass versus the calculated mass of 10 ppm is acceptable, especially considering the sometimes low concentration levels [26]. In our previous study, where a Waters-Micromass LCT Premier ToF-MS was used, we found that for more than 80% of the studied compounds the mass accuracy was within the 10 ppm acceptability limit and that most of the compounds that did not comply eluted in the region where most of the matrix compounds eluted [11]. In this study we found an average mass measurement error of 3 ppm (median 2.5 ppm) with little difference between the three matrices. While for >98% of the studied compounds the mass accuracy was below the 10 ppm limit, in individual analyte measurements excursions up to 20 ppm were encountered. These results are comparable with those of Ojanpera and Kolmonen, who both applied the same type of ToF-MS instrument for the analyses of drugs in urine and found a mean mass measurement error of 2.51 ppm and <5 ppm for all compounds, respectively [27,28]. An interesting observation is that the average mass measurement error slightly decreases with increasing concentration for each of the three matrices, as illustrated in Fig. 2a. That the deviations in mass accuracy are related to concentration is also shown in a recent study by Bristow who observed an increase of the deterioration in mass measurement accuracy at both, very low and very high ion abundances [29]. Different from our previous study the highest mass measurement errors are found not only in the region after 8 min where most of the matrix compounds elute from the column, but also in the retention time region of 3–8 min, where most of the analytes elute from the analytical column. A similar relation as for the mass accuracy is expected for the SigmaFit value, an exact numerical comparison of the theoretical and measured isotopic patterns of a compound. The results, illustrated in Fig. 2c, show that for each of the three matrices the SigmaFit value decreases with increasing concentration. This is according to expectations since the signal-to-noise ratio increases with concentration resulting in a better defined isotope pattern. In general, the average SigmaFit value was around 0.04 (median around 0.01) with a slight increase from meat to fish to egg. This seems to reflect the complexity of the matrices since the results of the validation study in terms of repeatability and CC␤ are also better for meat than for egg. A comparison of the SigmaFit values with the elemental composition seems to suggest that higher SigmaFit values are found for compounds containing chlorine with an average SigmaFit value of 0.050 ± 0.012 versus 0.036 ± 0.016 for non-chlorine compounds. However, no firm conclusions can be drawn from this since many more non-chlorine – than chlorine – containing compounds are involved in this study resulting in an unbalanced comparison. In addition, no correlation was found between elemental composition and mass measurement error, and none of both, SigmaFit and mass measurement error, showed a relation with the molecular mass or the retention time in the chromatogram. Not surprisingly, the highest deviation in SigmaFit are found in the retention time region of 3–8 min, where most of the analytes elute from the analytical column, and after 10 min where most of the matrix compounds elute from the column. 3.3. Repeatability, reproducibility and accuracy The results in Table 1 show that the repeatability of the full method for compounds in meat at the validation level is <10% for 50% of all compounds and >20% for less than 10% of the compounds. Apart from some individual compounds like enroflaxin and

R.J.B. Peters et al. / J. Chromatogr. A 1216 (2009) 8206–8216

8213

compounds in the chromatogram are not the only source of the general decrease of the repeatability. Another probable reason causing this decrease is the fact that at least five different types of fish, ranging from low-fat to high-fat containing fish, were used in the validation. Especially, this difference in fat content may influence the recovery of less polar compounds and the retention behaviour of the SPE column, both leading to a decrease of the repeatability. Fig. 3a shows the median repeatability in relation to analyte concentration and matrix. While there is a decrease of the repeatability with decreasing analyte concentration, the general impression is that the matrix is a more important factor. Fig. 3b shows the results for median intra-laboratory reproducibility, three series of seven samples processed and analysed on different days and by different analysts. For meat the reproducibility is about 1.5 times higher than the repeatability which is more or less according to expectation. However, for fish and egg the reproducibility is only slightly higher than the repeatability which indicates that the matrix is a more dominating factor than the moment in time or the analyst performing the analysis. The reproducibility was <20% for 75% of the analytes spiked to meat at any of the three concentrations. For 95% of the analytes the reproducibility was <40% and poor reproducibility’s (>40%) were only found for the ionophores that elute in the last part of the chromatogram in between the matrix peaks. For the other matrices and concentrations the situation is similar with the exception of the 0.5 VL level in fish. For reasons unknown the results of one series deviated for all compounds suggesting a deviation in calibration, final extract volume or injection volume. Therefore, internal standards will be included in future method development. For analytes in meat the accuracy at 1.0 VL ranges from 47% to 145% with 92% of the compounds in the range of 70% to 120%. A low accuracy was found for phenylbutazone (47%) and semduramycin (53%), and for robenidine (138%) and monensin (145%). For egg similar results were found with 80% of the compounds with an accuracy between 70% and 120%. While piroxicam showed a very low accuracy with values around 1000% for all three validation levels, the accuracy value of monensin increased with decreasing concentration. Both results show that the matrix interferes with the quantification of the compound. For fish the accuracy ranges from 50% to 118% with the median values at all three concentrations about 20% lower than those for meat and egg. For 80% of the compounds the accuracy is in the range of 70% to 120%. For fish a low accuracy is found for phenylbutazone (55%) and fenoprofen (58%). The median values for the accuracy at the three validation levels and the three matrices are given in Fig. 3c. Taking into account the minimal performance criteria we have set, we have to conclude about 20% of the compounds cannot be quantified accurately since these compounds have a repeatability >20%, a reproducibility >40% and/or an accuracy <70% or >120%. In that case accurate quantification is only possible if a standard addition procedure or compound specific optimisation has been applied. Fig. 3. Median values of repeatability (a), reproducibility (b), and accuracy (c) in relation to matrix and validation level.

3.4. Linearity and CCˇ

erythromycin, the compounds showing poor repeatability are found in that second part of the chromatogram where the matrix elutes. Egg was considered to be the most difficult matrix of the three but still we find that 40% of the compounds at the VL level have a repeatability <10% while 30% show a repeatability >20%. As in meat, a few individual compounds show poor repeatability while most of the compounds with poor repeatability’s are found in the region of the chromatogram where the matrix elutes. For fish the situation is a bit different. Only 25% of the compounds show repeatability’s <10% while, on the other hand, 20% show repeatability’s >20%. This result suggests that overlapping peaks of matrix

The squared regression coefficient (r2 ) is used to evaluate the linearity and the average r2 for each matrix is given in Table 1. The matrix matched calibration curves show an r2 > 0.99 for 92% of the compounds in meat with exceptions for lincomycin, phenylbutazone and a few tetracyclines. For fish 85% of the compounds have an r2 > 0.99 with exceptions for tetracyclines and the ionophores. Finally, for eggs only 75% have an r2 > 0.99, exceptions here are the tetracyclines, ionophores and many of the sulphonamides. In general, these results are comparable or better than the linearity determined for the same compounds in a milk matrix using a Waters-Micromass LCT Premier ToF-MS and using drug specific extraction windows [11]. The Bruker micrOTOF used in

8214

R.J.B. Peters et al. / J. Chromatogr. A 1216 (2009) 8206–8216

this study utilizes an analog-to-digital converter (ADC) which has an increased ion abundance dynamic range for accurate mass measurement and thus, should allow more reliable quantitative measurements than the Waters time-to-digital converter (TDC) [29]. A limit of detection is not determined but may be estimated as 3 times the repeatability of the compound in a particular matrix. CC␤ values are calculated from the reproducibility at the level of interest, e.g. 1.0 VL, and addition to the 1.0 VL level. CC␤ is therefore always larger than the validation level at which it is determined. The CC␤ values for all compounds in each matrix are given in Table 1. In general, for 90% of the compounds the CC␤ values are within 2 times the VL level at which they were determined. Exceptions are the NSAIDs carprofen, phenylbutazone, mefenamic and meclofenamic acid, and the ionophores. For these compounds in the matrix meat the determined CC␤ values are within 2.5 times the VL level at which they were determined. In fish the CC␤ values are comparable to meat, however, for the NSAID’s carprofen and phenylbutazone they are about 4 times the validation level. For eggs the CC␤ values of compounds eluting in the chromatogram before 7 min are comparable with those in the matrices meat and fish, all within 2 times the VL level. For compounds eluting after 8 min, mainly NSAID’s and ionophores, the CC␤ values run up to 3–4 times the VL level, again showing the effects of the matrix on the performance of the method, especially the instrumental analysis, even if a modern HRLC–ToF-MS is applied. This means that HRLC–ToF-MS alone will not be able to distinguish an unlimited number of compounds in any matrix. Determination of a large percentage of veterinary drugs in urine, milk and even meat appears to be feasible. However, if matrices become more complicated as with fish and egg, more adequate sample preparation techniques appear to be required. 3.5. Robustness, specificity/selectivity The robustness of the method was tested by analysing four samples of each matrix in duplicate, choosing four somewhat different sample pre-treatment/extraction procedures for each duplicate. Since the relative standard deviation within these eight analyses was smaller than the intra-laboratory reproducibility, it is concluded that the method is robust. The specificity of the method was checked by the analysis of 20 known-blank samples for each matrix. The chromatograms were monitored for the analytes of interest or for peaks that can potentially interfere with the analytes of interest. Since no peaks (>LOD or >3 times the repeatability) were detected in these samples, it is concluded that the method is also specific/selective.

have to be determined for screening methods. For the determination of CC␤ the 2002/657/EC document recommends two options: The first is to use the variability of the analytical response, e.g. analyse at least 20 spiked samples on at least three different days and calculate CC␤ from the reproducibility. The second approach is the use of ISO 11843 that determines CC␤ (and CC␣) based on a linear regression model analysing sample material on different days and spiked at different concentration levels [31]. ISO 11843 states that the test samples should be analysed and that the results (not corrected for recovery) should be linear regressed versus the concentration, e.g. y = ax = b. CC␣ and CC␤ can now be determined as the critical values of the response variable [32,33]:

 ˆ CC␣ = CMRL + t,˛ a

1+

(xMRL − x¯ )2 1 + IJ (xij − x¯ )2

where a is the slope of the regression line which equals the recovery of the analyte, CMRL is the MRL value of the analyte, t,˛ the associated t-value,  is an estimation of the residual standard deviation of the regression function, I the number of replicates per concentration, J the number of concentrations of the spiked samples, xMRL is the referenced MRL value of the analyte and x is the mean of the xij values. In the same way the detection capability CC␤ can also be determined with ISO 11843 by using the equation:

 ˆ CC␤ = CMRL + ı,˛,ˇ a

1+

(xMRL − x¯ )2 1 + IJ (xij − x¯ )2

where ı,˛,ˇ is a statistical function that can be fairly approximated by 2t,˛ . If the analyte is a forbidden substance and there is no MRL value, the values for CMRL and xMRL are simply set to 0. Using a balanced approach (equal number of samples on each day), a minimum of 20 samples, 3 days and 3 spike levels, this would require a minimum of 27 samples to determine CC␤. The repeatability and within-lab reproducibility are determined by preparing a number of samples with concentrations of 0.5, 1 and 1.5 times VL (or MRL). The 2002/657/EC document prescribes a minimum of 20 samples resulting in seven replicate analyses at each concentration level on 1 day, carried out on at least 3 different days. As an alternative, this type of calculations can be performed using the single factor ANOVA function in an EXCEL spreadsheet [34]. In that case the repeatability variance (Sr2 ) and within-lab repro2 ducibility variance (SR2 ) can be estimated from the estimators Srep 2 with: and Sday

3.6. An alternative validation protocol

2 Sr2 = Srep

Newly developed methods have to be validated, but validation according to EU Commission Decision 2002/657/EC is costly and takes a lot of effort and time. Discussions about new protocols for the validation of qualitative screening methods go in the direction of evenly costly and perhaps even more complicated protocols [30]. This validation effort can be reduced if a smaller number of samples can be used. However, this will lead to an increase of the uncertainty associated with the performance characteristics and as a consequence, these characteristics will be no longer comparable with those of methods validated according to 2002/657/EC. Therefore, alternative validation protocols should determine the same characteristics and with the same uncertainty as determined by the 2002/657/EC procedure. The number of degrees of freedom of the determination of each characteristic can be used to compare the uncertainty. According to the 2002/657/EC document, the detection capability, repeatability and reproducibility, selectivity and specificity, and robustness and stability, are the performance characteristics that

2 and S 2 themselves can be determined from the mean squares: Srep day

and

2 Srep = MSrep

2 SR2 = Sr2 + Sday

and

2 Sday = (MSday − MSrep )/nday

Following the 2002/657/EC document the number of degrees of freedom associated with the repeatability standard deviation is 6. Therefore, the degrees of freedom for the replicate in the ANOVA calculation, nrep − nday , should be 6 to ensure comparability, and since the 2002/657/EC document requires that the within-lab reproducibility is carried out on at least 3 days, nday = 3, resulting in nrep = 9 and thus, a total of 27 samples. This means that by analysing nine samples fortified at three concentration levels on 3 different days, a total of 27 samples, CC␣, CC␤, the recovery, repeatability and within-lab reproducibility can be determined. Since the numbers of degrees of freedom are at least equal to that in the 2002/657/EC strategy, the uncertainty in the performance parameters is comparable, e.g. they are of comparable quality.

R.J.B. Peters et al. / J. Chromatogr. A 1216 (2009) 8206–8216

Fig. 4. Comparison of CC␤ values of the veterinary drugs in meat (triangles), fish (diamonds) and egg (circles) determined according to the 2002/657/EC approach and an alternative approach requiring a lower total number of analysis. The results for the matrix egg deviate most from the imaginary straight line because the repeatability is less than for the two other matrices.

This alternative approach was used to calculate the repeatability, within-lab reproducibility and CC␤ using only three of the seven daily replicates at each validation level and on the three validation days. In total 10 different data sets were used to determine the uncertainty in each performance characteristic. The results show that the performance characteristics determined by the alternative approach (total of 27 samples) are comparable with those previously determined using a total of 63 samples. In general, the CC␤ values determined by the alternative approach appear to be slightly (10%) higher, but the differences are usually within 20% as is illustrated by Fig. 4 showing the relation between the CC␤ values determined using both methods. Larger differences were only encountered for a number of compounds in the matrix egg, probably as a consequence of the lower repeatability of analysis in this matrix. Nevertheless, the results show that the high number of samples required in the 2002/657/EC validation strategy can be reduced by about 50% using a different strategy. For the validation of screening methods this may be a cost-effective alternative for the 2002/657/EC strategy for confirmatory analysis. 4. Conclusions Several examples have been published showing and advocating that HRLC–ToF-MS is probably the most powerful tool for multicompound methods as is also demonstrated in this study. However, the present study not only shows a multi-compound capability, but in addition a multi-matrix capability. The presented method was shown to be suitable for the analysis of about 100 veterinary drugs in the matrices meat, fish and egg. While the results are satisfactory for 70–90% of the veterinary drugs involved in this study, it is also shown that application of HRLC–ToF-MS cannot replace an adequate sample preparation. The median mass measurement error of the ToF-MS instrument used in this study was 2.5 ppm with little difference between the three matrices for the veterinary drugs spiked at concentrations ranging from 4 to 400 ␮g/kg. The mass measurement error slightly decreased with increasing concentration. We conclude that for >98% of the studied compounds the mass measurement error

8215

criterion of <10 ppm is met; however, single excursions for individual measurements up to 20 ppm were observed. The SigmaFit value, a new feature of this ToF-MS for isotope pattern matching, also decreases with increasing concentration. In addition, the results show an increase of the SigmaFit value with increasing matrix complexity, e.g. going from meat, to fish, to egg. While the average SigmaFit value is 0.04, the median is 0.01 indicating some high individual deviations. As with the mass measurement error, the highest deviations are found in those regions of the chromatogram where most compounds elute from the column and we therefore conclude that the SigmaFit value is influenced by co-eluting analytes or matrix compounds. The median repeatability of the method ranges from 8% to 15% with the better values for the higher validation levels and the simpler matrix (meat). The median reproducibility ranges from 15% to 20% with little difference between matrices and concentrations. The median accuracy is in between 70% and 100% with a few compounds showing higher values due to matrix interference. While the linearity is less in difficult matrices like egg, the squared regression coefficient is >0.99 for >90% of the compounds. The detection capability, CC␤, is within 2 times the validation level for >90% of the compounds studied. CC␤ values much higher than the validation level are found for some compounds with low sensitivity in the ToF-MS. The method is robust as well as specific. From the results above we conclude that the method meets the previously set performance characteristics criteria for >90% of the compounds in meat, for >80% of the compounds in fish and for >70% of the compounds in egg, clearly showing the influence of the matrix on method performance. Finally, an alternative validation strategy is proposed and tested for screening methods. While the results calculated for repeatability, within-lab reproducibility and CC␤ show a good comparison for the matrices meat and fish, and a reasonable comparison for the matrix egg, only 27 analyses were required to obtain these results versus 63 analyses in the traditional, 2002/657/EC, approach. We conclude that for the validation of screening methods this alternative is a cost-effective validation procedure. Acknowledgement This study was financial supported by the Dutch Ministry of Agriculture, Nature and Food Quality (project # 87163901). References [1] Official Journal of the European Union, L224 of 18 August 1990, Council Regulation 2377/90/EC of 26 June 1990 laying down a Community procedure for the establishment of maximum residue limits of veterinary medicinal products in foodstuffs of animal origin, Brussels, Belgium, 1990. [2] Official Journal of the European Communities, L224 of 18 August 1990, Council Regulation 2377/90/EC, consolidated version of the Annexes I to IV updated up to 22.12.2004 obtained from http://www.emea.eu.int. [3] J.F. García-Reyes, M.D. Hernando, A. Molina-Díaz, A.R. Fernández-Alba, Trends Anal. Chem. 26 (2007) 828. [4] M. Danaher, H. De Ruyck, S.R.H. Crooks, G. Dowling, M.J. O’Keeffe, J. Chromatogr. B 845 (2007) 1. [5] A.A.M. Stolker, U.A.Th. Brinkman, J. Chromatogr. A 1067 (2005) 15. [6] K. Granelli, C. Branzell, Anal. Chim. Acta 586 (2007) 289. [7] R. Yamada, M. Kozono, T. Ohmori, F. Morimatsu, M. Kitayama, Biosci. Biotechnol. Biochem. 70 (2006) 54. [8] A.A.M. Stolker, T. Zuidema, M.W.F. Nielen, Trends Anal. Chem. 26 (2007) 967. ˜ [9] P. Munoz, J. Blanca, M. Ramos, M. Bartolomé, E. García, N. Méndez, J. Gomez, M.M. de Pozuelo, Anal. Chim. Acta 529 (2005) 137. [10] C. Soler, Y. Picó, Trends Anal. Chem. 26 (2007) 103. [11] A.A.M. Stolker, P. Rutgers, E. Oosterink, J.J.P. Lasaroms, R.J.B. Peters, J.A. van Rhijn, M.W.F. Nielen, Anal. Bioanal. Chem. 391 (2007) 2309. [12] A. Kaufmann, P. Butcher, K. Maden, M. Widmer, Anal. Chim. Acta 586 (2007) 13. [13] A. Kaufmann, P. Butcher, K. Maden, M. Widmer, J. Chromatogr. A 1194 (2008) 66. [14] H.G.J. Mol, P. Plaza-Bolanos, P. Zomer, Th.C. de Rijk, A.A.M. Stolker, P.P.J. Mulder, Anal. Chem. 80 (2008) 9450.

8216

R.J.B. Peters et al. / J. Chromatogr. A 1216 (2009) 8206–8216

[15] Official Journal of the European Communities, L221 of 17 August 2002, Commission Decision 2002/657/EC of 12 August 2002 concerning the performance of analytical methods and the interpretation of results, Brussels, Belgium, 2002. [16] J. van Loco, A. Janosi, S. Impens, S. Fraselle, V. Cornet, J.M. Degroodt, Anal. Chim. Acta 586 (2007) 8. [17] M. Hernando, M. Mezucua, J. Suarez-Barcena, A. Fernandez-Alba, Anal. Chim. Acta 562 (2006) 176. [18] R. Galarini, A. Piersanti, S. Salamida, L. Fioroni, Anal. Chim. Acta 586 (2007) 130. [19] J.P. Antignac, B. Le Bizec, F. Monteau, F. Andre, Anal. Chim. Acta 483 (2003) 325. [20] R. Romero-Gonzalez, J.C. López-Martínez, E. Gómez-milán, A. Garrido-Frenich, J.L. Martínez-Vidal, J. Chromatogr. B 857 (2007) 142. [21] S. Noguchi, H. Terada, Y. Tamura, Shokuhin Eiseigaku Zasshi 49 (2008) 177. [22] J. Xiaofei, J. AOAC Int. 91 (2008) 461. [23] C.A. Kan, Residues of veterinary drugs in eggs and possible explanations for their distribution between egg white and yolk, Dissertation Bergischer Universität Wuppertal, Germany, 2003. [24] P.R. Foster, P. Dunnill, M.D. Lilly, Biotechnol. Bioeng. 13 (2004) 713.

[25] A. Gersen, RIKILT – Institute of Food Safety, Wageningen UR, The Netherlands, 2008. [26] A.M.F. Laures, J.C. Wolff, C. Eckers, P.J. Borman, M.J. Chatfield, Rapid Commun. Mass Spectrom. 21 (2007) 529. [27] S. Ojanpera, A. Pelander, M. Pelzing, I. Krebs, E. Vuori, I. Ojanpera, Rapid Commun. Mass Spectrom. 20 (2006) 1161. [28] M. Kolmonen, A. Leinonen, A. Pelander, I. Ojanpera, Anal. Chim. Acta 585 (2007) 94. [29] T. Bristow, J. Constantine, M. Harrison, F. Cavoit, Rapid Commun. Mass Spectrom. 22 (2008) 1213. [30] V. Gaudin, P. Sanders, P. Göwik, S. Uhlig, Guide for analytical validation of screening methods, draft document version 4, September 2007, presented at National Reference Laboratories Meeting at Fougeres, France, July 2008. [31] ISO 11843: 1997 Capability of detection – Part 1: Terms and definitions, Part 2: Methodology in the linear calibration case, 1997. [32] E. Verdon, D. Hurtaud-Pessel, P. Sanders, Accred. Qual. Assur. 11 (2006) 58. [33] P.Th. Wilrich, Appl. Stochastic Models Business Ind. 18 (2002) 339. [34] E. Mullins, Statistics for the Quality Control Laboratory, The Royal Society of Chemistry, Cambridge, United Kingdom, 2003.