Multilaboratory Trial for Determination of Ceftiofur Residues in Bovine and Swine Kidney and Muscle, and Bovine Milk

30 HORNISH ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 86, NO. 1, 2003 DRUGS, COSMETICS, FORENSIC SCIENCES Multilaboratory Trial for Determination of Ceftiofur Residues in Bovine and Swine Kidney and Muscle, and Bovine Milk REX E. HORNISH, PHILIP J. HAMLOW, and SCOTT A. BROWN Pharmacia Corp., Animal Health Preclinical Development, 7000 Portage Rd, Kalamazoo, MI 49001 A multilaboratory trial for determining ceftiofur-related residues in bovine and swine kidney and muscle, and bovine milk was conducted following regulatory guidelines of the U.S. Food and Drug Administration, Center for Veterinary Medicine. The methods convert all desfuroyl -ceftiofur-related residues containing the intact -lactam ring to desfuroylceftiofur acetamide to establish ceftiofur residues in tissues. Four laboratories analyzed 5 sets of samples for each tissue. Each sample set consisted of a control/blank sample and 3 control samples fortified with ceftiofur at 0.5 R m,r m,and2r m, respectively, where R m is the U.S. tolerance assigned for ceftiofur residue in each tissue/matrix: 0.100 g/ml for milk, 8.0 g/g for kidney (both species), 1.0 g/g for bovine muscle, and 2.0 g/g for swine muscle. Each sample set also contained 2 samples of incurred-residue tissues (one > R m and one < R m ) from animals treated with ceftiofur hydrochloride. All laboratories completed the method trial after a familiarization phase and test of system suitability in which they demonstrated >80% recovery in pretrial fortified test samples. Results showed that the methods met all acceptable performance criteria for recovery, accuracy, and precision. Although sample preparation was easy, solid-phase extraction cartridge performance must be carefully evaluated before samples are processed. The liquid chromatography detection system was easily set up; however, the elution profile may require slight modifications. The procedures could clearly differentiate between violative (>R m ) and nonviolative (<R m ) ceftiofur residues. Participating laboratories found the procedures suitable for ceftiofur residue determination. Ceftiofur sodium (NAXCEL /EXCENEL Sterile Powder; Pharmacia Corp., Kalamazoo, MI) and ceftiofur hydrochloride (EXCENEL/EXCENEL RTU Sterile Suspension; Pharmacia Corp.) are indicated for the treatment of respiratory disease in cattle and swine. Treatment Received April 1, 2002. Accepted by JM August 15, 2002. Corresponding author s e-mail: Rex.E.Hornish@pharmacia.com. is administered by daily subcutaneous injection of 1 5 mg/kg/day for 3 5 days. Ceftiofur, whether derived from sodium or hydrochloride salts or from free acid, is rapidly metabolized by loss of the furoic acid moiety to produce the central active metabolite, desfuroylceftiofur (DFC; Figure 1; 1). This metabolite rapidly conjugates with cysteine and glutathione, may dimerize to a disulfide dimer, and may reversibly bind to proteins. Because parent ceftiofur has a short half-life of 5 10 min in plasma following its intramuscular (IM) administration in cattle and swine (2), DFC is the metabolite of interest for residue monitoring. DFC is readily liberated from the various conjugated forms by simply treating residue extracts with dithioerythritol (DTE). However, liberated DFC is not sufficiently stable in the absence of DTE to serve as a viable analyte. It is, however, readily converted with iodoacetamide to an acetamide derivative (desfuroylceftiofur acetamide; DCA) that is stable and highly suitable for liquid chromatographic (LC) analysis. An analytical procedure based on these chemical principles, i.e., the ceftiofur-dca residue procedure (LC DCA), was developed and validated for monitoring residues of ceftiofur in plasma, milk, and tissues obtained from a variety of species (2 4). Part of the approval process for any food animal drug entity is a validated method for determining drug residue in carcasses of animals that may have been exposed to the drug. The method is used to monitor the human food supply for such residues. It is the drug sponsor s responsibility to develop, validate, and demonstrate to the regulatory agencies that the method is transferable and useful to those laboratories charged with monitoring the safety of the human food supply. The purpose of this Sponsor-Monitored Method Trial was to test the applicability and viability of the method in laboratories that are not familiar with the method and to demonstrate that the method can be adapted and transferred to such laboratories for regulatory use. The method trial followed the recommended procedure/guidelines of the U.S. Food and Drug Administration, Center for Veterinary Medicine (FDA/CVM; 5). Four laboratories participated in this validation trial: a U.S. government laboratory, a Canadian government laboratory, a private commercial laboratory, and a Pharmacia Animal Health Laboratory. These laboratories were randomly labeled Laboratories A D to preserve anonymity. Each laboratory was required to analyze 5 sets of samples for each tissue and for milk. Each sample set consisted of a nonfortified control/blank sample and 3 control samples fortified with

HORNISH ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 86, NO. 1, 2003 31 Figure 1. Metabolism of ceftiofur in cattle, swine, sheep, and rats. ceftiofur at 0.5 R m,r m,and2r m, respectively, where R m is the U.S. tolerance assigned for ceftiofur residue in each tissue: 0.100 g/ml for milk, 8.0 g/g for kidney (both species), 1.0 g/g for bovine muscle, and 2.0 g/g for swine muscle. Each sample set also included 2 samples of incurred-residue tissues from animals treated with ceftiofur hydrochloride, where one sample was <R m and the other was >R m. Experimental Test Matrixes (a) Bovine and swine control/blank kidney and muscle samples. Each laboratory received ca 500 g ground/homogenized frozen bovine and swine kidney and similar samples of frozen bovine and swine muscle. The samples were obtained from a local abattoir and tested by the LC DCA method at maximum sensitivity in the sponsor s laboratory to ensure that they were devoid of ceftiofur residue. Samples were packaged in ca 45 g subsamples in 50 ml polypropylene screw-cap tubes, which were shipped in dry-ice. Upon receipt, the samples were transferred to a freezer maintained at < 15 C until analysis. (b) Bovine and swine incurred-residue kidney and muscle samples. Samples included 14 subsamples of 11 12 g each of ground/homogenized kidney, and 14 similar samples of muscle from cattle and swine treated by IM injection with ceftiofur hydrochloride. Cattle were treated IM with EXCENEL at 3 mg/kg/day for 3 days, and swine were treated IM at 7.5 mg/kg/day for 3 days. Animals were slaughtered at 2.5 and 24 h post-last-dose to harvest kidney and injection-site muscle to provide incurred-residue samples >R m and <R m,respectively. Samples were prepared in 20 ml polypropylene screw-cap tubes (7 samples each of kidney and muscle had ceftiofur residue <R m, and 7 of each tissue had ceftiofur residue >R m ) and supplied to each laboratory. Before use, each tissue type was assayed in triplicate by the sponsor s laboratory by using the DCA method to ensure that the residue concentration was appropriately above or below the R m target values. The samples were randomly numbered/codified to retain anonymity. The tubes were shipped in dry-ice. Upon receipt, the samples were transferred to a freezer maintained at < 15 C until analysis. Storage stability studies for ceftiofur residue in frozen kidney and muscle (bovine and swine) have indicated that ceftiofur residue is stable in these tissues for >1 year when stored at < 15 C (Pharmacia Animal Health, unpublished data). (c) Bovine control/blank milk samples. Approximately 500 ml whole bovine milk obtained from a local dairy farm was used to prepare control/blank milk samples. The samples were tested by the LC DCA analytical method to be sure they were devoid of ceftiofur residue and packaged in ca 45 ml subsamples in 50 ml polypropylene screw-cap tubes, packed in dry-ice, and shipped to each laboratory. Upon receipt, the samples were transferred to a freezer maintained at < 15 C until analysis. (d) Bovine incurred-residue milk samples. Fourteen incurred-residue subsamples of 11 12 ml each of whole frozen milk, obtained from a cow treated by intramammary infusion of ceftiofur hydrochloride, were prepared in 15 ml poly-

32 HORNISH ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 86, NO. 1, 2003 Table 1. Overall summary for bovine kidney, composite of all laboratories Lab Data Recovery of fortified controls, % samples, g/g Concentration, g/g 4.00 8.00 16.00 Low High D Mean (n = 5) 88.1 84.6 89.2 1.06 11.0 Low 84.9 81.8 86.4 1.03 10.5 High 91.0 89.5 91.5 1.11 11.5 A Mean (n = 5) 83.3 79.7 a 83.0 1.00 10.1 Low 79.4 72.1 77.6 0.92 9.8 High 91.2 86.6 87.6 1.05 10.6 C Mean (n = 5) 102 98.3 a 106 1.43 12.3 a Low 94.0 94.0 99.0 1.23 11.2 High 120 102 117 1.74 13.5 B Mean (n = 5) 87.8 85.9 94.8 1.18 12.3 Low 81.8 80.1 90.0 1.07 10.4 High 94.8 94.2 104 1.36 14.0 D 2.7 4.0 2.4 2.8 3.8 A 5.5 7.9 4.8 5.7 3.3 C 10.6 3.6 8.3 14.1 7.7 B 5.5 7.1 5.7 9.4 10.6 Composite (all n) 4.0 8.0 16.0 Low High n 20 18 20 20 19 Mean 90.4 86.9 93.2 1.17 11.4 SD 9.5 8.1 10.1 0.203 1.22 RSD 10.5 9.3 10.8 17.3 10.7 a n = 4, either because a sample was lost during analysis or the result was discarded as an outlier based on the nearest neighbor test described by Dixon and Massey (6). propylene screw-cap tubes and supplied to each laboratory. Seven contained ceftiofur-related residue at a concentration <R m, and 7 contained ceftiofur-related residue at a concentration >R m. Before distribution to the 4 laboratories, the samples were randomly numbered to retain anonymity. The tubes were shipped in dry-ice. Upon receipt, the samples were transferred to a freezer maintained at < 15 C until analysis. Storage stability studies for ceftiofur residue in frozen milk have indicated that ceftiofur residue is stable in milk for at least 6 months when stored at < 15 C (Pharmacia Animal Health, unpublished data). (e) Blank/control fortification levels and incurred-residue samples. Each laboratory prepared its own fortified samples as follows: Blank/control tissue samples were fortified by preparing phosphate buffer solutions of ceftiofur at 100.0 and 10.0 g/ml. Aliquots were added to 10 g control tissues to prepare samples fortified to required concentrations. Blank/control milk samples were fortified by preparing a phosphate buffer solution of ceftiofur at 5.0 g /ml. Aliquots of 50, 100, and 200 L were added to 5.0 ml control milk to prepare samples fortified to required concentrations. Each complete set of kidney and muscle tissue samples consisted of a blank or control sample, 3 blank/control samples fortified with ceftiofur at 0.50 R m,r m,and2r m, respectively, and 2 samples of incurred-residue tissues (one at >R m and one at <R m ) from 2 animals of each species that had been treated with ceftiofur hydrochloride by IM administration. Each complete set of milk samples consisted of a blank/control sample, 3 blank/control samples fortified with ceftiofur (in phosphate buffer solution) at 0.50 R m,r m,and2r m,respectively, and 2 samples of incurred-residue milk from 2 different milkings of a cow that had been treated with ceftiofur hydrochloride by intramammary infusion. Analytical Methods The complete study protocols and laboratory procedures were sent to participating laboratories, which were instructed to follow instructions verbatim. A preliminary procedure for milk was previously described (3). In brief, a 5 ml milk sam-

HORNISH ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 86, NO. 1, 2003 33 Table 2. Overall summary for bovine muscle, composite of all laboratories Lab Data Recovery of fortified controls, % samples, g/g Concentration, g/g 0.50 1.00 2.00 Low High D Mean (n = 5) 84.8 84.2 83.4 0.426 2.42 Low 80.2 77.7 76.8 0.410 2.28 High 88.4 92.0 87.2 0.450 2.53 A Mean (n = 5) 78.9 81.9 81.2 0.417 2.13 Low 75.6 76.1 76.7 0.377 1.85 High 85.4 91.6 86.3 0.485 2.30 C Mean (n = 5) 98.2 98.0 94.8 0.618 2.68 Low 90.0 87.0 85.0 0.503 2.52 High 110 112 106 0.780 2.89 B Mean (n = 5) 93.2 92.1 93.0 0.518 2.71 Low 87.2 82.8 82.5 0.451 2.27 High 100 105 111 0.555 3.07 D 3.9 6.3 4.9 4.3 4.7 A 5.0 7.4 5.7 10.1 9.6 C 10.1 12.3 10.2 16.3 5.1 B 6.7 11.1 12.5 7.6 11.9 Composite (all n) 0.50 1.00 2.00 Low High n 20 20 20 20 20 Mean 88.8 89.1 88.1 0.495 2.48 SD 9.6 10.5 9.6 0.099 0.31 RSD 10.8 11.8 10.9 20.1 12.5 ple was incubated with DTE solution in ammonium acetate buffer (ph 8.9) at 50 C for 15 min to yield the common DFC intermediate. The free DFC was then extracted onto a C 18 solid-phase extraction (SPE) column, where it was reacted in situ at room temperature with iodoacetamide (in phosphate buffer) to form the stable DCA derivative. (Note: Precautionary measures printed on the label for handling the toxicant iodoacetamide should be followed.) Further purification was performed on a strong cation exchange (SCX) SPE column. The DCA was then eluted from the SCX column with methanol ammonium acetate buffer (15 + 85, v/v), then analyzed via gradient LC by using a C 18 column with UV absorbance detection at 266 nm. Because a reference standard was not available for the DCA derivative, a standard curve was prepared by fortifying whole blank milk with parent ceftiofur at various concentrations ranging from 0.050 to 0.50 g/ml. These milk solutions were then processed by incubation, derivatization, and SPE procedures. Previous work showed that the validated limit of quantitation (LOQ) for the milk method was 0.050 g/ml and that the estimated limit of detection (LOD), based on 3 standard deviations above the background mean noise in blank determinations, was 0.015 g/ml (Pharmacia Animal Health, unpublished data). The basic procedure for bovine tissues was adapted from the plasma method previously described (3), but it required modification due to interferences in tissues (4). The procedure for swine tissue was adapted from this bovine method and required essentially no modification (4). In brief, 10 g tissue sample was homogenized with 140 ml extractant solution (0.4% DTE in 0.05M, ph 9, borate buffer). A 15 ml aliquot (10% of the total, representing 1 g tissue) was incubated at 50 C for 30 min, and the resultant free DFC in solution was then reacted with iodoacetamide at room temperature for 30 min to yield DCA. The DCA derivative was first extracted onto a C 18 SPE column, followed by successive purifications on a strong anion exchange (SAX) SPE, and then on an SCX SPE column. (Note: The procedure may be interrupted after either the C 18 SPE or SAX SPE step if the entire procedure cannot be completed in 1 day.) The derivatized residue was analyzed via gradient LC as

34 HORNISH ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 86, NO. 1, 2003 Table 3. Overall summary for swine kidney, composite of all laboratories Lab Data Recovery of fortified controls, % samples, g/g Concentration, g/g 4.00 8.00 16.00 Low High D Mean (n = 5) 85.6 82.8 90.0 4.13 14.0 Low 76.1 79.8 84.2 4.01 13.2 High 93.2 86.0 96.2 4.29 14.9 A Mean (n = 5) 93.0 91.3 91.4 4.46 14.6 Low 91.9 89.0 86.6 4.19 13.2 High 95.5 94.9 94.2 4.76 16.0 C Mean (n = 5) 101 90.0 a 105 a 4.61 a 15.6 a Low 96.0 78.0 98.0 4.18 14.3 High 112 97.0 111 4.94 16.6 B Mean (n = 5) 94.6 91.0 100 4.83 16.0 Low 84.5 83.5 92.5 3.96 14.8 High 103 96.1 110 5.78 17.4 D 7.2 3.0 5.8 2.7 4.6 A 1.6 2.4 4.7 6.0 6.9 C 6.4 9.3 5.1 7.0 6.3 B 8.2 5.3 8.4 13.4 7.2 Composite (all n) 4.0 8.0 16.0 Low High n 20 19 19 19 19 Mean 93.5 88.7 96.2 4.50 15.0 SD 7.8 5.7 8.4 0.448 1.22 RSD 8.3 6.4 8.7 9.9 8.1 a n = 4, either because a sample was lost during analysis or the result was discarded as an outlier based on the nearest neighbor test described by Dixon and Massey (6). for milk. The standard curve was prepared by fortifying a phosphate buffer solution with parent ceftiofur at various concentrations, and then processing these solutions through the method. Phosphate buffer solutions, rather than tissue matrix extracts, were used for the standard curve to reduce the variability of matrix components, which may adversely impact the chromatographic response characteristics of the standards. Concentrations ranged from 0.075 to 12.5 g/ml. Previous work showed that the validated LOQ for tissues was 0.1 g/g and that an estimated LOD, based on 3 standard deviations above the background mean noise in blank determinations, was 0.015 g/g. The method has been tested and evaluated for recovery at fortification levels up to 16.0 g/g, although a 1:3 dilution (as described below) was required to bring the concentration of the sample solution down to within the limits of the standard curve (Pharmacia Animal Health, unpublished data). The laboratories began with a method familiarization phase of 4 steps: (1) During the setup and initial test, the laboratory assembled and tested the LC and data systems by using solutions of DCA at 6 concentrations that were provided by Pharmacia. (2) System suitability was determined by generating several standard curves with solutions of ceftiofur prepared to standard concentrations. (3) SPE cartridges were tested by acceptance/equivalency tests that processed a series of fortified samples. If the recovery was 80%, the laboratory proceeded with step 4, or readiness to perform. (4) This final familiarization step tested the overall preparation of the laboratory by analyzing a set of control/blank and fortified samples appropriate for each tissue. This initial phase was not monitored for Good Laboratory Practice (GLP) compliance (21 CFR Part 58). The method trial phase consisted of analyzing 5 sets of samples for each tissue. The control/blank samples were fortified with ceftiofur solutions. The incurred-residue samples were randomly coded to preserve anonymity, but were paired so that each day s run contained one sample <R m and one sample >R m. Because the 2 R m fortified kidney samples produced a final LC sample concentration beyond the range of the

HORNISH ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 86, NO. 1, 2003 35 Table 4. Overall summary for swine muscle, composite of all laboratories Lab Data Recovery of fortified, % controls samples, g/g Concentration, g/g 1.00 2.00 4.00 Low High D Mean 82.0 88.2 87.5 0.440 3.05 Low 79.4 86.7 85.4 0.420 2.99 High 86.4 89.9 89.8 0.460 3.15 A Mean 84.3 86.1 88.2 0.462 3.06 Low 81.6 83.1 84.8 0.436 2.64 High 87.9 92.2 92.5 0.541 3.32 C Mean 97.8 99.0 96.8 0.560 3.05 a Low 91.0 94.0 91.0 0.524 2.91 High 106 108 101 0.592 3.26 B Mean 97.4 98.9 98.4 0.532 3.30 Low 83.9 86.5 84.8 0.506 2.61 High 121 122 111 0.553 3.80 D 3.3 1.4 1.9 4.9 2.1 A 2.9 4.3 3.3 9.6 8.5 C 6.2 5.5 4.6 5.0 5.5 B 15.4 14.8 12.3 3.9 14.8 Composite (all n) 1.00 2.00 4.00 Low High n 20 20 20 20 19 Mean 90.4 93.0 92.7 0.499 3.06 SD 10.7 9.6 7.9 0.056 0.39 RSD 11.8 10.3 8.5 11.3 12.7 a n = 4, because a sample result was discarded as an outlier based on the nearest neighbor test described by Dixon and Massey (6). standard curve, kidney samples were diluted (1:3) by taking a 5.0, rather than a 15 ml, aliquot of the initial homogenate/extract for incubation and subsequent processing. In addition, the incurred-residue kidney samples were run both undiluted and diluted by processing a 15 and a 5 ml aliquot, respectively. The method trial phase was conducted according to GLP regulations (21 CFR Part 58). Data Collection and Statistical Analysis All records were maintained according to GLP standards (21 CFR Part 58). Deviations from the study protocol, including slight deviations from the method procedure, were to be noted. All adverse events were described in detail. For any samples that required re-analysis, the reasons for re-assay were clearly identified. All raw data were delivered to Pharmacia for a complete data audit and compilation into a final report for submission to the FDA/CVM for the approval of method as the regulatory determinative method to specifically monitor milk and tissues for ceftiofur residue. Results and Discussion No laboratory found it necessary to adjust the sample preparation steps, although Laboratory C used a different centrifuge speed than specified and performed shaking manually rather than with a mechanical shaker. Laboratory D performed final chromatography differently because of an LC pump that produced a shorter retention time than that referenced in the study procedure. The evaluation and use of the standard curve and calculation of results were also performed according to protocol. Deviations were generally caused by not excluding those standards that back-calculated to 10% but these deviations had very little (negligible) impact on the study results. (Note: The criteria for standards followed the SOP developed by Pharmacia for defining acceptable standard response factors in which the observed value must be within 10% of the predicted back-calculated value to be acceptable and retained in defining the standard curve. Otherwise, the standard is discarded and a new standard curve computed, with a second iteration of back-calculated values.)

36 HORNISH ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 86, NO. 1, 2003 Table 5. Overall summary for bovine milk, composite of all laboratories Lab Data Recovery of fortified controls, % samples, g/ml Concentration, g/ml 0.05 0.10 0.20 Low High A Mean (n = 5) 97.2 104 102 0.068 0.331 Low 84.0 98.0 97.5 0.060 0.285 High 104 107 108 0.076 0.382 B Mean (n = 5) 103 106 99.8 0.063 0.317 Low 96.0 98.0 92.0 0.057 0.290 High 110 114 112 0.068 0.341 C Mean (n = 5) 98.6 98.4 99.0 0.062 0.293 Low 94.0 95.0 97.0 0.052 0.241 High 104 101 104 0.066 0.338 Mean (n = 5) 102 105 101 0.067 0.315 D Low 98.0 100 94.5 0.065 0.282 High 108 109 110 0.074 0.341 A 8.2 3.7 3.7 10.4 11.7 B 5.1 5.4 8.2 7.7 5.8 C 4.4 2.7 2.9 9.0 16.4 D 3.8 3.3 6.6 5.6 7.1 Composite (n = 20) 0.05 0.10 0.20 Low High Mean 100 104 101 0.065 0.314 SD 5.7 4.9 5.5 0.006 0.034 RSD 5.7 4.7 5.4 8.8 10.9 (a) Bovine kidney. Table 1 summarizes the combined results for bovine kidney for the 4 participating laboratories. The overall study recovery, computed for individual results at the 3 concentrations tested by the 4 laboratories was 90.4% at 4.0 g/g, 86.9% at 8.0 g/g, and 93.2% at 16.0 g/g. The ranges of recovery at the various concentrations are shown in Table 1. The relative standard deviations (RSDs) of mean recoveries for fortified samples from the individual laboratories were all <10%, except for the 4.0 g/g samples analyzed by Laboratory C, where the RSD was 10.6%. When the 18 20 individual determinations at each concentration were averaged as an overall assessment of method performance across the 4 laboratories at each concentration, overall recovery RSDs (9.3 10.8%) indicated that reasonable method performance criteria for precision were met for bovine kidney (5). The mean concentrations reported for incurred-residue samples by the 4 laboratories ranged from 1.0 (Laboratory A) to 1.43 g/g (Laboratory C) for low-concentration samples and from 10.1 (Laboratory A) to 12.3 g/g (Laboratories B and C) for high-concentration samples. The means of the RSDs for the composite/combined results were 17.3 and 10.7% for the low- and high-concentration samples, respectively. Laboratory C had the highest variability with the low-concentration incurred-residue samples (RSD = 14.1%), but all other laboratories had a lower variability at both concentrations, with RSDs <11%. (b) Bovine muscle. Table 2 summarizes results for bovine muscle. Two standard observations in 5 runs at one laboratory (Laboratory C) were discarded for failure to meet the criteria of variance within 10% of the predicted value (see Note above). The overall study recovery, computed for individual results at the 3 concentrations tested by the 4 laboratories, was essentially identical for the 3 levels: 88.8% at 0.50 g/g, 89.1% at 1.0 g/g, and 88.1% at 2.0 g/g. The ranges of recovery at the various concentrations are shown in Table 2. The RSDs of mean recoveries for fortified samples from individual laboratories ranged from 3.9 to 12.5%. Across the 4 laboratories at each concentration, overall recovery

HORNISH ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 86, NO. 1, 2003 37 RSDs (10.8 11.8%) indicated that reasonable method performance criteria for precision were met for bovine muscle (5). The mean concentrations reported for incurred-residue samples by the 4 laboratories were reasonably consistent, ranging from 0.417 (Laboratory A) to 0.618 g/g (Laboratory C) for low-concentration samples, and from 2.13 (Laboratory A) to 2.71 g/g (Laboratory B) for high-concentration samples. The means of the RSDs were 20.1 and 12.5% for the low- and high-concentration samples, respectively. Laboratory B had the highest variability with the low-concentration incurred-residue samples, with RSD = 16.3. (c) Swine kidney. Table 3 summarizes results for swine kidney for the 4 participating laboratories. Results for one fortified sample in each of 2 runs for Laboratory C appeared to be anomalies and were excluded as outliers (at the 10% level) in the calculation of mean recoveries by using the nearest neighbor test for outliers (6). The overall study recovery, computed for individual results at the 3 concentrations tested by the 4 laboratories was 93.5% at 4.0 g/g, 88.7% at 8.0 g/g, and 96.2% at 16.0 g/g. The ranges of recovery at the various concentrations are shown in Table 3. The RSDs of mean recoveries for fortified samples from the individual laboratories were all <10%. Across the 4 laboratories at each concentration, overall recovery RSDs (6.4 8.7%) indicated that reasonable method performance criteria for precision were met for swine kidney (5). The mean concentrations reported for the incurred-residue samples by the 4 laboratories were reasonably constant, ranging from 4.13 (Laboratory D) to 4.83 g/g (Laboratory B) for low-concentration samples and from 14.0 (Laboratory D) to 16.0 g/g (Laboratory B) for high-concentration samples. The RSDs at each concentration for all 4 laboratories were all <10%, except for low-concentration samples by Laboratory B at 13.4%. (d) Swine muscle. Table 4 summarizes results for swine muscle for the 4 participating laboratories. In 2 of 5 runs for Laboratory C, one standard observation in each run was discarded for failure to meet the criteria of a variance within 10% of the predicted value. The overall study recovery, computed for individual results at the 3 concentrations tested by the 4 laboratories was 90.4% at 1.0 g/g, 93.0% at 2.0 g/g, and 92.7% at 4.0 g/g. The ranges of recovery at the various concentrations are shown in Table 4. The RSDs of mean recoveries for fortified samples from individual laboratories were all <10% except for all samples assayed by Laboratory B, where RSDs were all >10%. Across the 4 laboratories at each concentration, overall recovery RSDs (8.5 11.8%) indicated that reasonable method performance criteria for precision were met for swine muscle (5). The mean concentrations reported for incurred-residue samples by the 4 laboratories were reasonably consistent, ranging from 0.440 (Laboratory D) to 0.560 g/g (Laboratory C) for low-concentration samples and from 3.05 (Laboratories D and C) to 3.30 g/g (Laboratory B) for the high-concentration samples. The means of the RSDs were 11.3 and 12.7% for lowand high-concentration samples, respectively. Laboratory B had the highest variability with the incurred-residue samples (RSDs >10% for the high-concentration samples), but all other laboratories had low variability, with RSDs <10%. (e) Bovine milk. Table 5 summaries the combined results for the 4 participating laboratories. One standard observation in one run for Laboratory C was discarded for failure to meet the criteria of a variance within 10% of the predicted value. The overall study recovery, computed for individual results at the 3 concentrations tested by the 4 laboratories was 100.4% at 0.050 g/ml, 104% at 0.10 g/ml, and 101% at 0.20 g/ml. The ranges of recovery at the various concentrations are shown in Table 5. One sample at 0.050 g/ml for Laboratory A appeared to be an outlier, but was not excluded because its value (84%) was >80%. The averages of the RSDs of mean recoveries for fortified samples were all <10%. Across the 4 laboratories at each concentration, overall recovery and RSDs (4.7 5.7%) indicated that reasonable method performance criteria for precision were met for bovine milk (5). The mean concentrations reported for incurred-residue samples by the 4 laboratories were reasonably consistent, ranging from 0.062 g/ml (Laboratory C) to 0.068 g/ml (Laboratory A) for the low-concentration samples, and from 0.293 (Laboratory C) to 0.331 g/ml (Laboratory A) for the high-concentration samples. The means of the RSDs were 8.8 and 10.9% for low- and high-concentration samples, respectively, and were acceptable across the laboratories. (f). Figure 2 shows the composite results for all samples and participating laboratories. The combined recovery of ceftiofur residue was 90.3% from bovine kidney, 88.6% from bovine muscle, 102% from bovine milk, 92.8% from swine kidney, and 92.0% from swine muscle. The RSDs ranged from 5.3 (bovine milk) to 11.0% (bovine muscle). In general, Laboratories A and D achieved consistently fewer variable results; the RSDs for both fortified control samples and incurred-residue samples were <12% in all cases. Laboratory C experienced a higher variability with bovine kidney and muscle (RSDs from 3.6 to Figure 2. (percent recovery) for all tissues in the method trial.

38 HORNISH ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 86, NO. 1, 2003 16.3%) than with swine kidney and muscle (RSDs <10%). Laboratory B had higher variability with bovine muscle (RSDs from 6.7 to 12.5%) and swine muscle (RSDs from 3.9 to 15.4%) than with kidney samples. No major concerns were expressed by any of the participating laboratories. All of the participating laboratories completed the method trial as per the study protocol once they assembled, set up, and established the LC/UV detection system. They were directed to complete a series of pretrial exercises and readiness-to-perform tests. During the familiarization run, Laboratory D found that the retention time (Rt) obtained for the DCA peak (7.8 min) was less than that required by the protocol (9 11 min). The cause of the reduced Rt was identified as a smaller dead volume in the LC equipment used by the laboratory compared with the equipment used by the sponsor. Although Rt could be lengthened to 10.4 min by modifying the gradient program, this laboratory used the protocol conditions with Rt of 7.8 min because the degree of separation of the analyte peak from co-extracted materials was acceptable. Laboratory C used a centrifuge speed of 29 000 rpm instead of the recommended 48 000 rpm (for centrifugation of the extract for either milk or tissues just before the C 18 SPE step) because of equipment limitations. The same laboratory used manual agitation for extraction as opposed to a shaking water bath. However, these changes had no apparent effect on the results, which were consistent with those of the other laboratories. Laboratory D noted that one swine muscle sample consistently stopped flowing after the NaOH wash of the C 18 SPE. The sample had a thick layer of fat after centrifugation, but skimming fat from the sample had no effect on its rate of flow through the C 18 SPE columns. Positive pressure was applied to complete the analysis. No explanation for this single apparent aberration was provided. Once initial exercises were completed and the methods were performing well, the trial itself ran smoothly. The important result from this trial was that the performances of both tissue and milk methods were acceptable by all defined criteria for method/procedure suitability to qualify them as suitable regulatory methods (5). None of the laboratories expressed serious concerns about the methods, except perhaps for lengthy tissue sample preparation; a minimum of8hwas needed to prepare a set of kidney or muscle samples. Nevertheless, the methods are reliable and appropriate for the specific detection and quantitation of ceftiofur-related residue in bovine and swine kidney and muscle and bovine milk. No changes or modifications to the laboratory procedures were recommended by any of the participants. However, Laboratory A noted that stopping points during the sample preparation should be indicated when the full 8 h preparation time cannot be completed in a single working day, such as after an SPE step. Laboratory A also correctly pointed out that the use of the term recovery for the milk results, as used in the interpretation of the results, was incorrect but instead should be termed accuracy, because the matrix (and method process) was used to prepare the standards for the standard curve. Thus, true percent recovery cannot be determined when a matrix-prepared calibration curve is used, as in this milk procedure, because recovery of drug from milk is already factored into the calibration curve data. This likely explains why the calculated mean recovery at all concentrations and across the range of concentrations in milk was 100%. Conclusions The methods met acceptable performance criteria in terms of recovery, accuracy, and precision. Sample preparations were easily implemented and conducted, but took about 8 h to complete. Care must be exercised to properly evaluate the performance of the SPE cartridges before implementing the method with unknown samples. The LC detection system was easily set up and maintained. The procedures could clearly differentiate between a violative (>R m ) and nonviolative (<R m ) ceftiofur-residue sample in bovine and swine kidney and muscle and bovine milk. The participating analysts found that the overall procedures were workable, and thought they would be suitable for routine analyses of ceftiofur residue in these tissues and could, therefore, serve as the regulatory methods for ceftiofur-related residue in bovine and swine kidney and muscle, and bovine milk. Acknowledgments This study was funded by Pharmacia Corp. References (1) Beconi-Barker, M.G., Roof, R.D., Vidmar, T.J., Hornish, R.E., Smith, E.B., Gatchell, C.L., & Gilbertson, T.J. (1996) in Veterinary Drug Residues Food Safety, ACS Symposium Series 636, American Chemical Society, Washington, DC, Ch. 9, pp 70 84 (2) Jaglan, P.S., Cox, B.L., Arnold, T.S., Kubicek, M.F., Stuart, D.J., & Gilbertson, T.J. (1990) J. Assoc. Off. Anal. Chem. 73, 26 30 (3) Jaglan, P.S., Yein, F.S., Hornish, R.E., Cox, B.L., Arnold, T.S., Roof, R.D., & Gilbertson, T.J. (1992) J. Dairy Sci. 75, 1870 1876 (4) Beconi-Barker, M.G., Roof, R.D., Millerioux, L., Kausche, F.M., Vidmar, T.J., Smith, E.B., Callahan, J.K., Hubbard, V.L., Smith, G.A., & Gilbertson, T.J. (1995) J. Chromatogr. B Biomed. Appl. 673, 231 244 (5) U.S. Food and Drug Administration (July 1994) Guideline for Approval of a Method of Analysis for Residues, Section V, Center for Veterinary Medicine, Rockville, MD (6) Dixon, W.J., & Massey, F.J. (1969) Introduction to Statistical Analysis, 3rd Ed., McGraw-Hill, New York, NY