Solvent‐Based and Matrix‐Matched Calibration Methods on Analysis of Ceftiofur Residues in Milk and Pharmaceutical Samples Using a Novel HPLC Method

ABSTRACT

Background

The matrix effect is accepted as one of the critical factors that affect trueness, and it is especially important in the analysis of critical compounds, such as drugs, toxins and pesticides, in complex matrices. Ceftiofur (CEF) is a leading drug used in dairy industry for pulmonary infections. Because milk has a complex matrix, which is rich in fats, proteins and many other compounds, the determination of CEF in milk products deals with issues of trueness. On the other hand, pharmaceuticals also have a complex composition as they contain not only the active pharmaceutical ingredient but also many excipients, such as preservatives and pH modifiers. Due to its frequent use, its residues can be found in meat products or exist as an excreted compound in urine, milk and so on.

Objective

The aim of this study was to investigate the effect of different calibration methods on quantitative estimation of CEF in milk and pharmaceutical samples.

Methods

CEF was analysed using a new HPLC method, in which separation was achieved on a C₁₈ bonded fused‐core silica particle column by using isocratic elution; the method was optimized by testing different chromatographic conditions and validated according to the ICH Q2(R1) and United States Pharmacopeia guidelines.

Results

Interestingly, it was found that CEF signals in milk samples were higher than the ones at the same level prepared in calibration solutions with a ratio of 11.28:1.

Conclusion

This finding reveals the necessity of matrix‐matched calibration for accurate quantitative determination of CEF in milk samples. Moreover, the milk in the market and the most used pharmaceutical formulations were analysed using the current method.

Prepared standard ceftiofur solution is added to 2 mL of milk sample.

The solution is stirred for 20 min.

For the protein, precipitation was added to 4 mL of acetonitrile to the solution.

The heterogeneous solution is stirred and sonicated for 20 min.

The solution is centrifuged at 5180 rpm for 15 min.

The filtrate is taken and filtered with a 0.22 m PVDF syringe filter.

The solutions prepared in this way are injected into the HPLC system, and a calibration curve is obtained.

Abbreviations

ADI

acceptable daily intake

BEN

benzocaine

CEF

ceftiofur

FAO/WHO

Food and Agriculture Organization and the World Health Organization

FDA

Food and Drug Administration

IS

internal standard

ME

matrix effect

MMC

matrix‐matched calibration

MRLs

maximum residue limits

MRPL

minimum required performance limit

1. Introduction

Ceftiofur (CEF), which is classified as a third‐generation, broad‐spectrum cephalosporin antibiotic, was introduced into the pharmaceutical market to use only on food‐producing animals in 1988 (Hornish and Katarski 2002). It bears an oxyimino–amino thiazolyl group as a 7‐b amino‐acyl substituent of the 7‐amino cephalosporin nucleus and contains a furoic acid thioester at the third position, which is an unexcelled substitution for the third‐generation cephalosporins (Hornish and Katarski 2002; Yancey et al. 1987). Following its administration, CEF mainly transforms into desfuroylceftiofur, which is also effective against bacteria (Papich 2016).

Today, veterinary drugs not only provide benefits for animal health but also mean economic profit for producers; on the other hand, because they possibly remain in various tissues, eggs and milk of animals, they are a source of risk for human health. Especially in long‐term exposure, they may cause a variety of critical problems, such as an increase in antibiotic resistance or allergy incidence. Therefore, at the international level, the Food and Agriculture Organization and the World Health Organization (FAO/WHO) jointly created the Codex Alimentarius to determine the acceptable daily intake (ADI) and maximum residue limits (MRLs) for each drug or pesticide. In Turkey, the legislation on veterinary drug residues in foodstuffs is the Turkish Food Codex Classification of Pharmacological Active Substances that may be in Food of Animal Origin and Regulation on MRLs, which is updated according to EU legislation and includes Directive 2377/90/EC (Gazete 2002). Many countries have restricted feeding the food‐producing animals with antimicrobial drugs. The responsible institutions in the European Union have issued some guidelines for the management of veterinary medicines in food of animal origin. The Council Directive 96/23/EC was established for measures on residues of pharmaceutical compounds in food products from animals (European Commission 1996). European Commission Regulation 470/2009/EC is another guideline, which is about the tolerance levels of pharmaceutical compounds (European Commission 2009). Regulation 37/2010 includes the use of pharmaceutical products and their MRL in food products from animal origin; besides, it lists compounds with no MRL that have been given that threaten the consumer but have not been not determined and observed (European Commission 2010). In addition, a minimum required performance limit (MRPL) has been presented to perform the analytical performance of methods for some non‐authorized substances. For antibiotics that do not have an MRL or MRPL value, the level of validation is determined according to the class of the active pharmaceutical compound (Kaufmann 2009). On the basis of the latest residue studies, the current guide offers the following MRLs for CEF: 1000 µg/kg (muscle), 6000 µg/kg (kidney), and 2000 µg/kg (fat) in cattle and pigs, and the Committee reaffirmed the MRLs as 2000 µg/kg for liver, cattle and pigs, and 100 µg/L for cow milk. Particularly, microbiological inhibitor tests used in conventional bulk milk control systems lack adequate sensitivity for the marker residue of CEF, desfuroylceftiofur.

A review of the literature reveals that analysis of CEF was accomplished with using separation techniques, such as HPLC besides desfuroylceftiofur (Jacobson, Martinod, and Cunningham 2006), CEF crystalline‐free acid (Hibbard et al. 2002; Kilburn, Cox, and Backues 2016; Hope et al. 2012; Waldoch, Cox, and Armstrong 2017; Gardhouse et al. 2017; Waraich et al. 2017; Woodrow et al. 2016) and CEF‐free acid equivalents (Beconi‐Barker et al. 1995); capillary electrophoresis (Puig, Borrull et al. 2007; Puig, Tempels et al. 2007), and disk plate method (Owens et al. 1990), ELISA (Althaus et al. 2003), brilliant black reduction test (Molina et al. 2003), cathodic stripping voltammetry (Barbosa, Trindade, and Ferreira 2006; Barbosa et al. 2011), microbiologic assay (Mills et al. 2000), capillary‐based fluorescent immunoassay (Huth et al. 2002.), ultra‐violet spectrometry (Ribeiro and Schmidt 2017), LC–MS/MS (Becker, Zittlau, and Petz 2003; Keever, Voyksner, and Tyczkowska 1998) was also used. Among these, HPLC is the dominant technique because it allows quantitative determination in many different matrices. An overview of reported HPLC methods for the determination of CEF is presented in Table 1.

TABLE 1.

An overview of reported HPLC methods for the determination of ceftiofur (CEF).

Method	Column	Internal standard	Detection	Linearity range	RT (min)	LOD	LOQ	Application	References
HPLC (ion paired)	Phenomenex phenyl (250 mm × 4.6 mm ID, 3 µm)	—	289.6 nm	10–1000 ng/mL	11.0	50 ng/mL	NI	Milk serum	Tyczkowska et al. (1993)
HPLC (isocratic)	Microsorb C18 (250 mm, pd 5 µm, 100 µm)	—	254 nm	10–1000 µg/mL	17.0	10 ng/mL	NI	Plasma	Navarre et al. (1999)
HPLC (ion‐paired gradient)	Waters Nova‐Pak phenyl (150 mm × 3.9 mm ID, 4 µm)	—	270 nm	20–200 mg/kg	41.2	7 µg/kg	9 µg/kg	Raw bovine milk	Sørensen and Snor (2000)
HPLC (gradient)	Combination of Phenomenex C18 and Phenyl (50 mm × 4.6 mm i.d., 3 µm) and (50 mm × 4.6 mm ID, 3 µm)	—	266 nm	NI	NI	NI	0.1 µg/mL 0.1 µg/g 0.1 µg/g 0.1 µg/g	Plasma Lochial fluid Caruncles Endometrium	Okker et al. (2002)
HPLC (gradient)	PLRP‐S polymeric (150 mm × 2.1 mm ID, 5 µm, 100 Å)	—	266 nm	0–20 µg/mL	NI	0.36 µg/mL 0.27 µg/mL	0.5 µg/mL 0.5 µg/mL	Horse plasma Synovial fluid	De Baere et al. (2004)
HPLC (gradient)	Polymeric PLRP‐S (150 mm × 2.1 mm ID, 5 µm, 100 Å)	—	266 nm	NI	NI	0.35 µg/mL 0.25 µg/mL	0.5 µg/mL 0.5 µg/mL	Plasma Synovial	Pille et al. (2005)
HPLC (gradient)	LiChrospher C18 (250 mm × 4.6 mm ID, 5 µm)	—	292 nm	20.0–120.0 µg/mL	6.8	0.47 µg/mL	1.44 µg/mL	Drug substance and in sterile powder for injection	Souza et al. (2007)
HPLC (gradient)	Chromolith RP‐18e (100 mm × 4.6 mm ID, 5 µm)	Caffeine	265 nm	0.2–12 ng/µL	17.2 22.1 9.8	NI NI 0.4 ng/mL	NI NI 26.0 µg/kg	Milk	Karageorgou and Samanidou (2010)
HPLC (isocratic)	Hypersil C18 (250 mm × 4.6 mm ID, 3 µm)	—	290 nm	NI	15.2	10 µg/L	NI	Recycled water	Li et al. (2011)
HPLC (gradient)	Inertsil ODS‐3 (250 mm × 4 mm, 5 µm)	1,7‐Dimethyl‐xanthine	265 nm	0.5–20 ng/µL	36.1	11.8 µg/kg	35.7 µg/kg	Milk	Karageorgou, Samanidou, and Papadoyannis (2012)
HPLC (gradient)	Luna C18 (150 mm × 4.6 mm ID)	—	280 nm	0.1–10 mg/L	18.8	0.15 mg/L	0.5 mg/L	Plants	Cavenati et al. (2012)
HPLC (gradient)	Kinetex C18 (100 mm × 4.6 mm ID, 2.6 µm)	—	280 nm	NI	NI	3 µg/g	10 µg/g	Sediment Sludge	Carvalho et al. (2013)
HPLC (isocratic)	Hypersil BDS C18 (250 mm × 4.6 mm ID, 5 µ)	—	292 nm	50–150 µg/mL	7.6	NI	NI	Bulk form	Palur et al. (2013)
HPLC (gradient)	Waters C18 (150 mm × 4.6 mm ID, 5 µm)	—	254 nm	0.05–10 µg/mL	13.0	0.01 µg/mL	0.03 µg/mL	Milk	Chen and Ye (2016)
HPLC	C18	—	263 nm	NI	NI	0.01 µg/L 0.05 µg/L	NI NI	Milk or serum sample	Han et al. (2017)
HPLC (gradient)	C18 (250 mm × 4.6 mm; ID, 5 µm)	—	265 nm	0.1–20.0 µg/mL	24.0	0.03 µg/mL	0.10 µg/mL	Mueller–Hinton broth with foetal bovine serum	Nie et al. (2016)
HPLC	Nova‐Pak HR C18 (25 mm × 10 mm, 6 µm)	(3‐(Trimethylsilyl)‐1‐propanesulfonic acid	254 nm	0.4–40 µg/mL	16.5–17.5	NI	NI	Bovine plasma	Jacobson, Martinod, and Cunningham (2006)
HPLC (isocratic)	Bondapack C18 (250 mm × 4.6 mm ID, 5 µm	—	310 nm	0.1–10 µg/mL	6.8	0.03	0.11	Bubaline plasma	Adil et al. (2019)
HPLC (isocratic)	Eclipse XDB‐C18 (4.6 mm × 250 mm, 5 µm)	—	292 nm	NI	3.2	—	—	Goat plasma	Helbling et al. (2020)

Abbreviations: ID, internal diameter; NI, not given information; pd, particle diameter.

Generally, setting up a calibration curve using the standard analyte solutions is the preferred way for quantitative analysis of complex matrices, such as milk, plasma or herbal extract; within this concept, it is assumed that the signal characteristics of the standard solutions are similar to those of the sample solutions in terms of qualitation and quantification. Contraversely, samples are subjected to multiple sample treatment steps and determined using the calibration curve; in addition, the accuracy of the method was generally analysed with recovery experiments. On the other hand, the Food and Drug Administration (FDA) guidelines specifically emphasize that calibration sets should be prepared by adding known analytes to the sample matrix for analysis in complex matrices. Besides, matrix‐matched calibration (MMC) may be a requisite for analysis of trace amounts in order to achieve accurate measurement of targeted compounds in natural matrices (Liu, Rochfort, and Cocks 2018; Seccia et al. 2008; Food and Drug Administration 2001). So, whether the matrix interference exists or not, use of MMC should be the plan A during method development for a warrantable assay.

The aim of the study presented herein was to examine the effects of different matrices (e.g. milk and pharmaceutical injectables) on quantitative determination of CEF; in addition, the development of a robust method which possesses better specificity when compared to the previously published methods for routine analysis of CEF in milk and pharmaceutical samples is accomplished. The compositional difference between milk and pharmaceutical injectables originates mainly from proteins, fats and lactose in milk and solvents, stabilizers and pH‐controlling agents in injectable solutions. The matrix effects were examined in this study, and quantitative determination of CEF was carried out in different matrices. This is the first article in which signal loss of CEF observed in milk samples due to matrix effects was overcome by using MMC; the method presented herein was validated according to the International Conference on Harmonization Q2(R1) (ICH Harmonised Tripartite Guideline 2005) and the United States Pharmacopeia (USP) (Pharmacopeia 2013).

2. Materials and Methods

2.1. Consumables and Instruments

2.1.1. Chemicals and Reagents

Reference materials of CEF sodium salt (CEF.Na) (93.6%, w/w) were purchased from Fluka, Buchs, Switzerland. Benzocaine (BEN) (≥99%, w/w, used as an internal standard), HPLC grade sodium dihydrogen phosphate (99.8%, w/w), phosphoric acid (86%, w/w), sodium hydroxide (pure), hydrochloric acid (37%, w/w), hydrogen peroxide (30%, w/w), acetonitrile and water were purchased from Sigma‐Aldrich (Missouri, ABD). Injectable suspensions of CEF were Cefcloren TeknoVet and Ceftivil TeknoVet (both contain hydrochloride salt of CEF at 50 mg/mL conc.), which were purchased from a local veterinarian.

2.1.2. Instrumental Conditions

All analyses were carried out using a Prominence series of HPLC system from Shimadzu Co. (Kyoto, Japan). The instrument is equipped with an LC‐20AT tandem dual‐plunger pump equipped with a low‐pressure gradient unit, SIL‐20AC HT auto‐sampler, CTO‐10ASVP column oven, DGU‐20A5R online degasser, CBM‐20A communications bus module and SPD‐M20A photodiode array detector (PDA). The dwell volume of the liquid chromatography system was determined to be about 500 µL. All standard and sample solutions were prepared using an RK 100 H model ultrasonic bath from Bandelin (Berlin, Germany), a model XSE 105 Dual Range model analytical balance and SevenMulti model pH meter from Mettler‐Toledo (Greifensee, Switzerland), a Rotina 380 R model centrifuge from Hettich (Tuttlingen, Germany) and a Heidolph Reax Top model vortex mixer (Schwabach, Germany).

The chromatographic separation was achieved on an Ascentis Express C₁₈ reversed‐phase column (100 × 4.6 mm, 2.7 µm particle size, from Supelco, Bellefonte, PA, USA) with isocratic elution. In addition, two different reversed‐phase columns, that is Ascentis Express Phenyl Hexyl (100 × 4.6 mm², 2.7 µm ID) and Ascentis Express F₅ (100 × 4.6 mm², 2.7 µm ID), were tested for performance, system suitability and specificity comparison. The mobile phase was composed of acetonitrile:phosphate buffer (0.025 M, pH 2.6):water (28:10:62, v/v/v). The injection volume was 5 µL, and the column oven temperature was set at 40°C. The PDA detector was set at 289 nm wavelength to obtain maximum absorbance from CEF. The real‐time spectra were recorded at 40 Hz data sampling frequency and 0.640 s time constant.

2.2. Preparation of Solutions

2.2.1. Reference Solutions

For the preparation of CEF standard solutions, 12.17 mg CEF.Na was accurately weighed and transferred into a 10 mL volumetric flask. The substance was dissolved in water and made up to volume with water for the stock solutions of MMC and solvent‐based calibration. The concentration of CEF.Na was 1.27 mg/mL; this solution was diluted to obtain further solutions of CEF.Na.

For IS solutions, 1.29 mg BEN was accurately weighed and transferred into a 100 mL volumetric flask. The flask content was dissolved and diluted to volume with water. Concentration of the resulting solution of BEN was 12.9 µg/mL; this solution was diluted to the required level in further studies.

The cow milk samples which were required for MMC solutions were purchased from local producers, from non‐medicated dairy cows. The matrix‐matched standard solutions were prepared for the liquid–liquid extraction (LLE) procedure, according to Jank et al. (2015) and given in Figure 1A. Briefly, a 2.0 mL aliquot of homogenized whole milk was transferred into a 50 mL polypropylene tube with the required amounts of CEF.Na, which was prepared according to the method mentioned above, was added to 2 mL of cow milk samples separately. After that, spiked milk samples that contain different concentrations of CEF.Na were shaken vigorously for 20 min. The extraction assay consisted of adding 4.0 mL of ACN, divided into three aliquots (2.0 + 1.0 + 1.0 mL) to precipitate milk proteins and lipids, shaking with vortex the tubes between each addition for 3 min. Afterwards, the mixture was sonicated in ultrasonic bath for 20 min and centrifuged for 15 min at approximately 5180 rpm, at 5°C. The supernatant was filtered through 0.22 µm PVDF syringe filters, and 1000 µL of the solution was directly transferred into HPLC vials. Finally, 200 µL of IS solution was added to HPLC vials to make the final concentration of 1718.0 ng/mL.

FIGURE 1.

The detailed sample preparation procedure.

2.2.2. Injection Suspension Assay Preparations

The preparation of injection suspensions is given in Figure 1B. The suspension was shaken for 10 min and vortexed for 2 min. After that, ca. 232 mg of the suspension was weighed into a 100 mL volumetric flask, and 30 mL of tetrahydrofuran was added to dissolve the solid drug particles; after mixing, acetonitrile was added to make up the volume and vortexed. The solution was diluted 1:25 (v/v) with acetonitrile and filtered through a PTFE 0.22 µm syringe filter. The filtrate of 1000 µL was transferred into HPLC vials and spiked with IS solution.

2.2.3. Milk Sample Assay Preparations

The preparation protocol for the milk samples is as mentioned above; only 200 µL of IS solution was added to the samples after protein precipitation to make the final concentration of BEN 1718.0 ng/mL.

2.2.4. Mobile Phase Solutions

The buffer in the mobile phase was prepared as follows: Sodium dihydrogen phosphate of 34.0 mg and phosphoric acid of 421 µL were dissolved in 500 mL of water and sonicated for 10 min. The other mobile phase components (i.e. acetonitrile and water) and the buffer solution were filtered through non‐sterile membrane filters (47 mm I.D., 0.45 µm pore size, Sartorius, Germany) and degassed before use. Accurate mixing and preparation of the mobile phase components in the desired ratio was realized by the low‐pressure gradient unit in the pump.

2.3. Analytical Method Validation

2.3.1. System Suitability Testing

For system suitability testing, resolution (R _s), tailing factor (T _f), peak width, the number of theoretical plates (N), tailing factor (T) and the height equivalent to a theoretical plate (HETP) were calculated via Shimadzu LC LabSolutions 1.24 SP1 data integration software. In addition, peak capacity (PC) was evaluated to see the analysis performance of isocratic separation.

2.3.2. Specificity

For the method specificity, the signals of CEF and BEN were examined, and peak purities were investigated using a PDA detector. In addition, chromatograms recorded using different types of stationary phases were compared with each other, and possible interfering peaks were inspected in an orthogonal manner.

2.3.3. Linearity and Range

The linearity solutions were prepared to cover 20 different concentration levels for both types of working solutions, between 94.2 and 94,238.6 ng/mL for matrix‐matched and 94.2–18,839.2 ng/mL for solvent‐based. The range of the method corresponded to 5%–5000% and 5%–1000% estimated concentrations of the test solutions for both matrix‐matched and solvent‐based calibration sets, respectively. The calibration curves of both of the methods were plotted as CEF concentration versus corresponding signal ratio. Peak normalization ratio for each point was calculated as follows: ratio = (peak area [CEF]/retention time [CEF])/(peak area [BEN]/retention time [BEN]). Slope, intercept, certain intervals of the slope, the intercept at 95% confidence level (CI) and correlation coefficient were calculated.

2.3.4. Accuracy

Recovery experiments were used to test the accuracy. The recovery tests were carried out at three different concentration levels, corresponding to low (80%), medium (100%) and high (120%) fortification. For each level, three parallel sets have been prepared. Spiked samples were re‐analysed, and average recovery with standard deviation, RSD% and CIs was estimated at 95% CI.

2.3.5. Precision

Precision tests consisted of intraday and interday (intermediate) experiments, and it was determined by analysing reference solutions and sample solutions of injectable suspension. Eight repetitive analyses were performed using the described method, on the same day and 3 consecutive days; in addition, both MMC and solvent‐based calibration methods were also tested. The findings were statistically evaluated like mean, standard error of mean, standard deviation, RSD% and confidence interval at 95% CI. Moreover, variations among inter‐day groups were examined with a one‐way ANOVA test performed using Prism 6.0 software (GraphPad, CA, USA).

2.3.6. Limit of Detection (LOD) and Limit of Quantification (LOQ)

LOD and LOQ values of the method were calculated according to the signal/noise ratio according to the ICH recommendations; signal/noise ratio was equalized to 3 for LOD and to 10 for LOQ. Then, CEF solutions at the LOD and LOQ were injected six times for each, and the responses in the chromatograms were evaluated.

2.3.7. Robustness

The robustness was examined by applying deliberate changes in the methodological parameters by ±10% and analysing calibration solutions at 100% level for both calibration curves. The effects of each parameter on the method were examined separately and compared with the optimum conditions. The results were calculated to recover system suitability tests, such as relative retention time, resolution, the number of theoretical plates and tailing factor.

3. Results

3.1. Method Optimization

In the presented study, various mobile phase characteristics, such as type and ratio of organic solvent, buffer pH and concentration and component ratio, have been examined to provide a good chromatographic separation with proper system suitability. Initially, methanol and acetonitrile were tested for the organic components of the mobile phase; when methanol was used, the peaks were spread out, and the baseline noise was increased. On the other hand, better separation efficiency and sharper peaks were observed with the use of acetonitrile; in addition, baseline noise decreased significantly compared to methanol. The additional advantages of acetonitrile over methanol were lower absorption in the UV‐region, decreased viscosity and resulting better mass transfer. After a series of attempts, acetonitrile:phosphate buffer (0.025 M):water (28:10:62, v/v/v) mixture was found to be a suitable mobile phase. pK _a values of CEF and BEN were 3.7 and 2.5, respectively. The buffer type and pH were selected from this perspective, and pH was adjusted to ca. 2.6 (van den Berg 2013; Pichon, Schmidtmayer, and Ulbricht 1981). Under the mentioned condition, although CEF molecules were non‐ionized, BEN molecules were 50% ionized because the pH of the mobile phase was close to the pK _a of BEN. At given pH, CEF molecules had a strong hydrophobic interaction with the column's functional group (C₁₈), whereas IS was less ionized and possessed weaker hydrophobic interaction. Because the CEF molecule has many hydrophilic atoms with n orbitals, its hydrophobic interaction with the C₁₈ stationary phase was limited and thus rapidly eluted. Although BEN molecule is smaller than CEF, it is more apolar and has slower elution. The logP values of the CEF and BEN are ca. 0.2 and 1.9, respectively. Under the mentioned pH condition, the analyte could be eluted first and IS eluted afterwards. The pH of the mobile phase was measured to be about 3.2. Moreover, to obtain a balanced chromatogram in terms of baseline noise and peak sharpness, data collection frequency, best wavelength for detection and time constant values were also studied (details are not given here).

In the method development steps, it was decided to use an IS to increase method robustness. Some critical points to take into consideration were interaction with the stationary phase, absorptivity and hydrophobicity of candidate compounds; another important goal to be achieved was to keep the elution order of the substance as the last in the analysis to prove that the analytes have been completely eluted from the column and that the chromatographic separation has ended safely. Caffeine, ibuprofen, loratadine, rosuvastatin, mianserin, naproxen sodium, diflunisal, diclofenac, agomelatine, irbesartan and BEN were tested to be an IS. Among the mentioned compounds, BEN was found to be a suitable compound (pK _a 2.5, logP 1.86), which was eluted just after the CEF peak at the end of the run. BEN was selected due to its close but higher retention than CEF, similar absorbance characteristics, and acceptable system‐suitability results at optimized conditions.

Analysis of CEF using Ascentis Express C₁₈, Ascentis Express Phenyl Hexyl and Ascentis Express F₅ columns provided representative chromatograms shown in Figure 2, respectively. High speed and efficiency of sub‐2 µm particles can be gained by using fused‐core silica particle columns with lower system back‐pressure. Within this study, method transfer among these different types of silica particle columns was successfully accomplished. On the other hand, minor differentiations on relative retentions, tailing factor and peak height were observed: Particularly, the tailing factors of the peaks observed when F₅ and phenyl hexyl bonded silica columns were higher than that of C₁₈‐bonded silica column. The system back‐pressure observed was 149 bars when using Ascentis Express C₁₈ approximately. Among the columns tested, C₁₈‐bonded fused‐core silica column generally gave better results in terms of tailing factor and PC (Figure 2). The system‐suitability results for CEF were calculated from the data recorded under optimized conditions and presented in Table 2. Moreover, chromatograms of CEF and BEN under optimized conditions for matrix‐matched and solvent‐based calibrations are given Figure 3A,B, respectively. A good agreement was found when the results were compared with the recommended values.

FIGURE 2.

Representative chromatograms of solvent‐based calibration solutions recorded using different columns (C_CEF = 1883.9 ng/mL, C_BEN = 1718.0 ng/mL). BEN, benzocaine.

TABLE 2.

The system‐suitability data for ceftiofur (CEF) (C = 1883.9 ng/mL).

	Observed value
	Ascentis express C18		Ascentis express F5	Ascentis express phenyl hexyl
Parameter/Calibration method	Matrix‐matched	Solvent‐based	Solvent‐based	Solvent‐based	Acceptance criteria
Retention time (min)	2.397	2.398	3.194	6.548	—
Relative retention time (min)	0.579	0.580	0.745	0.7492	—
Relative standard deviation (%) of retention time	0.097	0.032	0.087	0.068	RSD ≤ 1%
Precision for relative area	0.012	0.003	0.024	0.026	RSD ≤ 1%
Injection precision for retention time (min)	0.000	0.013	0.008	0.010	RSD ≤ 1%
Capacity factor (k′)	1.349	1.067	1.465	1.527	k′ ≥ 2
The number of theoretical plates (N)	19,329	33.598	50.299	46.523	N > 2000
Resolution (R s)	7.776	11.406	13.891	15.225	>2
Tailing factor (T)	1.348	1.327	1.245	1.296	≤2
Peak width a	0.178	0.136	0.147	0.314	≤1

Abbreviation: RSD, relative standard deviation.

^a Calculated according to USP.

FIGURE 3.

Chromatogram of CEF (1883.9 ng/mL) and BEN (1718.0 ng/mL) under optimized conditions for matrix‐matched (A) and solvent based (B) calibrations. BEN, benzocaine; CEF, ceftiofur.

3.2. Method Validation

Specificity was determined via the analysis of CEF solutions that were kept under different stress conditions. CEF solutions at 1 mg/mL concentration were used for stress tests by mixing them with 0.1 N hydrochloric acid, 0.1 N sodium hydroxide and 3% (w/w) hydrogen peroxide at both room temperature and 60°C. Samples were collected at different times (at 60 and 120 min), analysed and new results were compared with the initial ones. Related results are given in Table 3. No degradation peak, which could interfere with CEF peak, was observed, presenting adequate specificity of the method. In addition, peak purities for selectivity were calculated in the PDA detector for degradation medium, milk sample matrix and pharmaceutical suspension sample matrix. In those matrices, the lowest peak purity was found to be 0.999632 and 0.999851 for CEF and BEN, respectively. The PDA evaluated the analyte's peak purity and found no evidence of co‐elution from samples.

TABLE 3.

Recovery of ceftiofur (CEF) under different stress conditions.

	Recovery (%) (at room temperature)		Recovery (%) (60°C)
Conditions	60 min	60 min	60 min	60 min
0.1 N HCl	79.02	76.14	72.68	67.70
0.1 N NaOH	—	—	—	—
3% (w/w) H2O2	31.68	9.63	—	—
UV254	92.21	91.03	85.15	78.20

In liquid chromatographic analyses, use of an internal standard with normalization contributes to the accuracy and precision of the method, especially where matrix effects are influental on the analyte loss; by this way, some inadvertent or unpredictable effects of external or internal factors, such as changes of ambient temperature, mobile phase composition and pH on the quality of analysis, can be decreased (Snyder, Kirkland, and Dolan 2011). In the current study, calculated ratios were found to be directly proportional to calibration concentrations. Table 4 displays statistical data of the linearity studies.

TABLE 4.

Statistical data for the linearity of ceftiofur (CEF).

	Parameter	MMC	Solvent‐based
Intraday (n = 18)	Linearity range (ng/mL)	94.2–9423.9	94.2–18,839.2
	Slope	0.0115	0.001019
	Intercept	0.13	−0.013271
	Regression coefficient	0.99981	0.9999
	SD of slope	0.00004	0.0000015
	SD of intercept	0.09118	0.00948
	ANOVA	F (1.16) = 0.1734 p > 0.05	F (1.16) = 0.1805 p > 0.05
Inter‐day (n = 18, k = 3)	Slope	0.0115	0.00102
	Intercept	0.2337	−0.0574
	Regression coefficient	0.99974	0.99799
	SD of Slope	0.03661	0.0721
	SD of intercept	0.00002	0.000002
	ANOVA	F (2.51) = 0.0.9999 p > 0.05	F (2.51) = 0.0.9998 p > 0.05
	LOD (ng/mL)	15.03	10.42
	LOQ (ng/mL)	37.59	31.66

Abbreviation: MMC, matrix‐matched calibration.

The chromatograms of CEF at the LOD and LOQ levels for under optimized conditions for matrix‐matched and solvent‐based medium were given in Figure 4A–D, respectively. When compared with previously reported techniques, the proposed method herein has the advantages of having a wider working range and lower LOD and LOQ in nanogram level (Han et al. 2017; Chen and Ye 2016).

FIGURE 4.

The chromatograms of CEF at the LOD and LOQ levels of matrix‐matched (A and B) and solvent‐based calibration (C and D) studies, respectively. BEN, benzocaine; CEF, ceftiofur.

Precision consists of two components: repeatability is the variations observed on a single instrument used by a single analyst, and intermediate precision refers to variations observed when applying the same method in a laboratory as a result of different application days, instruments and analysts. In order to perform precision assays, freshly prepared solutions at 1883.9 ng/mL concentration of CEF (regarded as 100%) were analysed 3 different days by different analysts with six replicates. The results were calculated as the mean recovery percentage and RSD% (Table S1). For the samples which were checked within the day and between days, there was no noticeable difference. All solutions were freshly prepared to ensure the stability of the analyte in solutions.

The described method protocols were tested for accuracy in the working range, as mentioned above. The standard addition method was used for recovery experiments by making six independent determinations over three different concentrations covering the whole range. Results are summarized as in Table S2.

Robustness was investigated via analysing CEF solutions prepared for matrix‐matched and solvent‐based calibration methods by introducing small changes in the organic component ratio in the mobile phase, mobile phase pH, flow rate, column temperature and detector wavelength. The difference in the parameters (e.g. retention time, number of theoretical plates and tailing factor) were evaluated after three replicates at final instrumental conditions. The results demonstrate the robustness of the procedure (Table S3).

The stability of CEF solutions was investigated by the analysis of CEF solutions after storage under different conditions. The 95% CI results are given in Table S4.

3.3. Matrix‐Matched Calibration

Several preliminary trials were conducted to analyse milk samples during the method development stage. In these analyses, despite the use of IS, interferences were observed due to the matrix effect, which cannot be avoided with LLE, and the analysis error cannot be drawn within acceptable limits. MMC was proposed to compensate for the interferences. As the milk matrix is rich of many different types of compounds, it is of high importance to determine the effects of matrix components on CEF extraction. The signals from the CEF molecules in the milk matrix were compared to the signals in the solvent. The comparison was denoted as signal suppression/amplification (SSE), that is slope ratios for MMC and EC, was calculated. This ratio was found to be 11.28 (greater than 1). It indicated that the CEF signal in the matrix medium was more than the solvent medium; besides, enhanced CEF analysis effects in the milk sample experimentally. Moreover, signal suppression/amplification caused by matrix effects was observed strongly in all milk sample types in a similar proportion.

The reason may be that the apolar compounds with the LLE are removed from the medium together with proteins and fats, whereas polar compounds remain in the environment. These polar compounds might increase the signal intensity by attaching to CEF molecules by molecular interactions such as adsorption or electrostatic attraction. Besides, the aim was taken to make the LLE procedure quite simple, easy and focused only on protein–fat precipitation with minimal analyte loss. Although many different LLE procedures were tested, similar signal enhancement was observed, and matrix‐matched standards had to be used for precise and accurate quantification.

3.4. Analysis of CEF‐Injectable Suspensions and Milk Samples

At the final stage of the experimental studies, methods were applied to real samples. Efficient extraction and rapid sample preparation procedures were searched for by doing intensive literature examination, and LLE was found to be suitable (Jank et al. 2015). In addition to conventional approaches in the LLE, many types of filters on the market were also examined, and PVDF‐type membrane filters, which enabled the least loss of analyte and minimum interaction with other compounds in the matrix, were chosen. There were no interferences to CEF peaks observed in the pharmaceutical preparations (Ceftivil and Cefcloren injectable suspension [50 mg/mL]) and milk samples in the assay. Typical chromatograms of injectable solution Ceftivil and Cefcloren and milk analyses are shown in Figure 5A–C, respectively. Each sample solution was injected into the system in triplicate, and average values were calculated for statistical evaluation (the statistical results are given in Table 5). In conclusion, the real samples were successfully analysed by the improved method that displays excellent performance from an analytical perspective.

FIGURE 5.

Assay chromatograms of injectable suspensions (Ceftivil (A) and Cefcloren (B)), and milk sample (C).

TABLE 5.

Assay results of Ceftivil and Cefcloren injectable suspension (n = 3 for each) and milk samples (n = 3).

Parameters	Ceftivil	Cefcloren
Mean ± SD (mg)	49.10 ± 0.79	48.58 ± 0.22
Median (mg)	49.72	48.57
RSD (%)	1.6166	0.4570
Standard error of mean (mg)	0.3000	0.084
Bias (%)	−1.79	−2.84
Confidence interval a	48.27–49.94	48.35–48.81

Sample	Source	Concentration of residual CEF b (μg/L)
IY	Cow milk (UHT)	ND
SL	Cow milk (UHT)	73.80
MY	Cow milk (UHT)	ND
PL	Cow milk (UHT)	80.19
DY	Cow milk (UHT)	ND
MSY	Cow milk (UHT)	80.55
BY	Cow milk (UHT)	ND
YL	Cow milk (UHT)	83.15
SL	Cow milk (UHT)	75.39
YY	Cow milk (UHT)	ND
MGY	Cow milk (UHT)	81.27
TY	Cow milk (UHT)	ND
YY	Cow milk (UHT)	72.35
KS	Cow milk (Raw)	ND
KOS	Cow milk (Raw)	77.39

Abbreviations: CEF, ceftiofur; ND, not detected; RSD, relative standard deviation; SD, standard deviation.

^a Confidence interval at 95% confidence level (mg).

^b It is 100 μg/L for Regulation of Turkish Food Codex and FAO/WHO.

4. Discussion

All parameters considered to affect the chromatographic analysis of the CEF were examined in detail during method optimization. This study was conducted on a fused‐core particle LC column; however, it was also possible to examine variations due to different interaction characteristics of CEF with the columns that have different phase chemistries and system suitability data. For instance, the organic modifier type and ratio in the mobile phase were decided based on these interactions. The pK _a values of CEF and IS were the most important factors in the mobile phase's pH selection. Because CEF was a highly polar compound, its rapid elution made the selection of IS a challenge.

In the developed method, peak purities were also calculated in each sample matrix and forced degradation medium for selectivity studies. In ultraviolet absorption studies, peak purity is an important parameter in demonstrating selectivity with a diode array detector (Marzouk et al. 2024). In a single‐wavelength ultraviolet‐absorption visible detector, peak purity is tried to be determined by whether there is any shoulder, tail or valley in the analyte peak. However, the absence of such a bulge in the peak is not a definite guarantee. Because the interference may not be visible due to low resolution. On the other hand, with the photodiode array and diode array detector, the entire wavelength spectrum emitted by the sample light source is scanned and separated into wavelengths with a diffraction grating (Papadoyannis and Gika 2004). It does not require mechanical scanning and spectral acquisition is provided in a very short time with HPLC sensitivity. Successful peak purity determination is made using more than one spectrum. There are successful studies on this subject, especially in methods recommended for drug analysis in the presence of degradation products. In fact, Ahmed et al. (2024) and Hussein et al. (2023), despite working with an ultraviolet visible region detector in their study, preferring to use a diode array detector to calculate peak purity in selectivity studies is a notable example for stability indicating or bioanalytical method validation studies (Wadie et al. 2024).

In the validation studies of the method, the most significant handicap was solved by using MMC. Thanks to the detailed examination of all optimization parameters and the use of IS, the successful performance was achieved within limits recommended in the ICH guideline. All necessary validation data were presented to avoid any uncertainty in the applicability of the method.

Because the CEF analysis of the developed method provides quantitative determination in both pharmaceutical preparations and milk samples, there are two different sample applications. It can be evaluated as a unique HPLC method that was easy to prepare samples for analysis of CEF residue in dairy milk products and independent of matrix effects thanks to the use of MMC and IS. The method also had a straightforward sample preparation procedure and gave high accuracy results. Another advantage of the method was its simple and readily preparable mobile phase composition, which was environmentally friendly, low cost and allowed rapid analysis time such as 5 min.

In order to demonstrate the applicability of the method, the most common milk samples in the market and samples from local producers, which were not applied by pasteurization, were collected and analysed with sufficient repetition. CEF was found to be below the MRL limits in all samples; it should be underlined that CEF is one of the most preferred antibiotics in treating dairy animals. Besides, local producers did not declare use of CEF because there were no sick animals in the sample collected group of animals.

5. Conclusion

The current method is the first HPLC method in the literature, which has fully optimized analytical properties; it was validated according to ICH guidelines, and matrix effects in the analysis of milk samples were examined. The study reports two types of calibration methods: matrix‐matched and solvent‐based calibration. The matrix effects on CEF analysis were found in milk samples. These effects increase the CEF signal in the milk sample and cannot be minimized entirely with LLE. So an MMC technique was developed to minimize this effect. When compared with previously reported techniques, the developed matrix‐matched and solvent‐based methods herein have many advantages of rapidity, simplicity, reliability, precision, wide working range, LOD and LOQ in nanogram levels (Han et al. 2017; Chen and Ye 2016). Because the fused‐core column packing material was first used in CEF analyses, fast separation, low detection and quantification limits could be achieved. Three types of fused‐core column chemistries, which have C₁₈, phenyl hexyl, and pentafluorophenyl functional groups, were examined for the separation of CEF in milk sample and pharmaceutical preparations. The fused‐core column with C₁₈‐bonded phase gave excellent resolution between CEF and IS. Compared with other reported techniques, the current work requires cheaper reagents and has a more straightforward sample preparation procedure (Tyczkowska et al. 1993; Karageorgou, Samanidou, and Papadoyannis 2012; Chen and Ye 2016).

Moreover, this is the first study in which MMC for CEF analysis was applied for milk samples; fifteen milk samples were collected from the local markets and analysed. The developed method offers a short analysis time for CEF, which is a prerequisite in the routine analysis of pharmaceutical preparations and biological samples. Thus, the method described herein is proposed as a suitable candidate for the screening of related drugs and milk samples.

Author Contributions

Saniye Özcan contributed to conception and design, supervision, analysis and acquisition of data and revising the manuscript critically for important intellectual content. Serkan Levent handled conception and design, acquisition of data, analysis and interpretation of data and revised draft. Nafiz Öncü Can was involved in conception and design, supervision, funding and revising the manuscript.

Ethics Statement

No living creature received any invasive or non‐invasive treatments in this study. Milk samples were purchased from markets and local producers. Therefore, it is not subject to ethical permission.

Consent

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.

Human and Animal Rights

Not applicable.

Supporting information

Supporting Information

Data Availability Statement

The data supporting the findings of this study are available within the article.