Simultaneous determination of six antibiotics in human serum by high-performance liquid chromatography with UV detection

Antibiotics are widely used in intensive care patients to treat severe infections. To avoid bacterial resistance or toxic side effects, the determination of serum concentration of ABs is advisable. Therefore, in this study, we developed and validated a simple and fast high-performance liquid chromatography method with UV detection for the simultaneous determination of four β -lactam ABs (meropenem, imipenem, ceftazidime, and piperacillin) and two coadministered substances (cilastatin and tazobactam) in human serum. Sample preparation required a simple protein precipitation by methanol. The separation of the ABs occurred within a timeframe of 17 min. For this purpose, we used a Kinetex F5 column with a linear gradient of acetonitrile and phosphate buffer (pH 6.9). The UV detector recorded two separate chromatograms at 220 and 295 nm simultaneously. Validation has demonstrated that the method is linear, accurate, and precise within the clinically relevant range for each substance.

Therefore, to ensure its effectiveness, Pip is always coadministered with a β-lactamase inhibitor, mostly tazobactam (Taz, C 10 H 12 N 4 O 5 S).
Like Cila, Taz itself has no antibacterial effect. The structures of the different ABs are shown in Figure 1.
The aforesaid substances are broad-spectrum ABs, used as reserve ABs, and can only be administered intravenously. Therefore, these ABs are central to the treatment of sepsis and life-threatening infections in intensive care units. Because of their unusual pharmacokinetics, the risk of underdosing or overdosing these ABs has increased in critically ill patients, as they tend to have altered physiology. The results of insufficient AB coverage are AB resistance and an inadequate treatment of the patient, resulting, for example, in an increase in hospitalization time or even death. For these reasons, the rational use of AB, the correct dosing, and treatment duration are key factors in AB therapy. Therefore, therapeutic drug monitoring with a constant determination of the AB concentration in serum or plasma is necessary (Cazorla-Reyes, Romero-Gonzáles, Garrido Frenich, Rodríguez Maresca, & Martínez Vidal, 2014;McWinney et al., 2010).
Many studies on the determination of ABs in serum, plasma, and other human materials have been performed since the early 1980s.
The aim of this work was to develop and validate a single HPLC method for the determination of serum levels of the aforesaid six ABs, which are most frequently used in the intensive care unit of our hospital. The main focus of the method development was fast and easy sample preparation and the use of low-cost chemicals and consumables.  Roth (Karlsruhe, Germany). Sodium azide was supplied by Acros (Geel, Belgium). Potassium ferrocyanide trihydrate was obtained from SERVA (Heidelberg, Germany). All solvents had HPLC or comparable quality and all reagents were of analytical grade. For quality control (QC) sample preparation, we used the following pharmaceuticals: meropenem 500 mg (Hikma, Gräfelfing, Germany), Piperacillin/ Tazobactam 4 g/0.5 g, Imipenem/Cilastatin 500 mg/500 mg (both from Fresenius Kabi, Bad Homburg vor der Höhe, Germany), and Ceftazidime 500 mg (Dr. Eberth, Ursensollen, Germany). Ultrapure water was prepared using a Synergy UV purification system (Merck Millipore, Darmstadt, Germany).

| Reagents
We tested following substances for their use as internal standard

| Instrumentation and chromatographic conditions
For the separation and quantification of the ABs we used an Agilent 1260 HPLC system (Agilent, Waldbronn, Germany), equipped with a G7112B binary pump, a G7129A autosampler, a G7116A column oven, and a G7114A variable wavelength detector. In addition, we used an Agilent Flexible Cube (G4227A) for the on-line SPE. As extraction cartridge, we used a Phenomenex Strata C8 20 μm column (20 × 2.0 mm; Aschaffenburg, Germany).
The chromatographic separation was performed at 50 C using a Kinetex F5 (Phenomenex, Aschaffenburg, Germany) column (150 × 3.0 mm), packed with 2.6 μm diameter particles and protected by the corresponding pre-column. The mobile phase consisted of a gradient of solution A (50 mM phosphate buffer, pH 6.9, containing 0.004% sodium azide) and solution B (95% ACN). The gradient started at 0% B, increased to 25% B in 9 min, stayed constant for 1 min, and returned to starting conditions in 0.1 min up to a total runtime of 17 min for re-equilibration. A constant flow rate of 0.4 mL/min was used throughout the runtime. The injection volume was 10 μL. The variable wavelength detector recorded two separate chromatograms at 220 and 295 nm.
For the determination of protein concentration, we used an AU5800 clinical chemical routine analyzer (Beckman Coulter, Krefeld, Germany).  Table 2. Prior to analysis, 50 μL of each stock solution was diluted with 950 μL of drug-free serum.

| Sample preparation
Blood samples were collected without any anticoagulant, transported chilled to our laboratory, and centrifuged at 2750g and 20 C for 10 min. After centrifugation, sera were frozen at −80 C and analyzed within 3 days.
150 μL of ISTD (containing MeOH) was added to 100 μL of serum/ calibrator/QC sample to precipitate the proteins, and vortex mixed immediately. Subsequently, the samples were centrifuged for 10 min at 16,000g. 20 μL of the supernatant was diluted with 180 μL of purified water before injection (10 μL) for chromatographic analysis.

| VALIDATION
where k is the relative uncertainty result (k = 3); s X0 is the standard deviation of the method; t is the quantile of the t distribution; α is the level of significance (error type 1, α = 0.01); m is the number of measurements; n is the number of calibration levels; X is the content value; and Q X is the sum of squares.

| Accuracy and precision
The intra-and inter-assay precision and the bias from the target value were determined by measuring the two QC samples on 9 different days, each in six replicates. To check the method accuracy, 21 samples containing Mero were measured with an alternative in-house method (Roth et al., 2017), whereas 17 samples containing Cefta, 21 samples containing Pip/Taz, and 21 samples containing Imi/Cila were measured in an external laboratory using an in-house LC-tandem mass spectrometry method.

| Recovery
The recovery rates of the ABs were quantified by a sixfold measurement of eight aqueous solutions, using the same concentrations as used for the determination of the working range (including 0 mg/L), after addition of the ISTD and 1:10 dilution with purified water. The peak height ratios (to the ISTD) of the aqueous solutions were defined as 100% recovery. The recovery rate for two concentrations is described in detail in the "Results and Discussion" section.

| Stability
The stability of the stock solutions was determined by comparing injections of the stock solution stored at −80 C for 1, 2, or 3 months.
These solutions were measured without any sample preparation. For all stability tests, peak height of each AB was used for interpretation.
To analyze the stability of the AB in real samples, both QC samples were prepared and frozen at −80 C. After 4, 24, 48, and 72 h, aliquots were thawed and subjected to complete sample preparation procedure and, subsequently, measured by HPLC. All aliquots were compared with the native samples.
Vial stability was examined using both QC samples.  Figure 2.
As Taz could only be detected at pH values of 2.1 and 6.9, the gradient development for the separation of ABs was performed on the Kinetex F5 column with a pH of 6.9 in the aqueous phase. After optimization of the gradient, the final chromatogram of the calibrator (Pip: 100 mg/L, all other: 50 mg/L) for the separation of ABs is generated ( Figure 3).
As already shown by Lefeuvre et al. (2017) andVerdier et al. (2011), Imi can be eluted as a double peak due to the tautomeric forms of its amidine group. Accordingly, during our method development, Imi often appeared as a double or even as several peaks. In such cases, we always considered the retention time of the first Imi peak (e.g. in Figure 2). However, by increasing the column temperature from 25 to 50 C, Imi eluted as a slightly broader, but single peak from the column, which led us to use this column temperature for further measurements.
During method development, we also encountered problems with the HPLC pump due to the formation of polymers by ACN and due to growth of algae in the phosphate buffer at nearly physiological pH value. To prevent the clogging of the HPLC pump by ACN polymers, we decided to use 95% ACN (diluted with water) instead of pure ACN. To stop the growth of algae, we added 0.004% sodium azide to range from 0.9 to 400 mg/L and depends on the sample preparation technique as well. Even for protein precipitation, the ISTD is either in the precipitation reagent or additionally added (Augey et al., 1996;Casals et al., 2014;Dailly et al., 2011;Demetriades et al., 1986;Denooz & Charlier, 2008;Farin et al., 1999;Holt et al., 1990;Hu et al., 2013;Legrand et al., 2008;McWinney et al., 2010;Robatel et al., 2002;Veillette et al., 2016).
By using the Kinetex F5 column, with the so-called core-shell particles and a particle size of 2.6 μm, we are able to separate all six ABs with narrow peaks, showing a high separating capacity, with a feasible pressure ranging from approximately 160 to 180 bars during the gradient elution. Using the gradient elution, all six ABs could be separated within 17 min and had the following retention times: 3.8 min (Imi), 4.6 min (ISTD), 6.5 min (Taz), 7.2 min (Cefta), 8.0 min (Mero),

F I G U R E 3
Final chromatogram for the separation of antibiotics. ISTD, internal standard 8.4 min (Cila), and 12.5 min (Pip). All HPLC methods using gradient elution in the literature have total runtimes between 15 and 35 min, so our method is quite fast to separate our desired ABs (Denooz & Charlier, 2008;Legrand et al., 2008;Robatel et al., 2002;Veillette et al., 2016;Verdier et al., 2011).

| Sample preparation
We first tried to perform an on-line SPE using an Agilent Flexible Unfortunately, after precipitation with perchloric acid, the chromatograms could not be clearly interpreted. Even neutralization and precipitation of the perchlorate with 2.5 M potassium carbonate solution did not improve the chromatograms. Besides, protein precipitation with trichloroacetic acid formed a peak that interfered with Imi, and that with MeOH led to the formation of double peaks. Protein precipitation with zinc sulfate had no impact on the chromatograms. However, the supernatant of the protein precipitation with zinc sulfate led to the formation of further precipitates after a short period. Addition of 0.1 M NaOH (precipitation of zinc hydroxide) or potassium ferrocyanide (precipitation of zinc ferrocyanide) could further prevent the formation of precipitates. As the formation of double peaks after protein precipitation with MeOH could be avoided by a subsequent dilution with water, we further removed the proteins (in 100 μL serum) by the addition of 150 μL of MeOH. After centrifugation, the supernatant was diluted 1:10 with water and directly used for HPLC measurement.
A further advantage of MeOH precipitation is that the ISTD can directly be dissolved in MeOH.
Compared with other HPLC methods mentioned in the literature, the sample volume required for our method is quite low (100 μL vs volumes ranging from 12.5 to 1000 μL serum/plasma; median: 350 μL, as reviewed in the "Introduction" section). In the literature, the Except Taz, for all other ABs the limits of detection described in this study were below the median value of the limits mentioned in literature.

| Linearity
The results for the determination of linearity and LOD/LOQ are summarized in Table 1. The highest calibration standard S7 represents the upper limit of the working range. We found the complete working range to be sufficient to cover the therapeutic range of all ABs.
Because of the good linearity, only one calibration point was used for further measurements.

| Accuracy and precision
The data for the determination of precision and accuracy are specified in Table 2. All values correspond to the specification of the GTFCh (≤20% for QC low; ≤15% for QC high). The comparison with an inhouse method and/or an external laboratory revealed a good linear correlation between the two methods for all ABs. Therefore, we plotted Passing-Bablok regression for the comparison of Mero with the in-house method and the new method ( Figure 4).

| Recovery
The recovery rate of a low and high concentration standard of each analyte is shown in Table 3.

| Stability
Stock solutions for all ABs were stable for at least 3 months at −80 C as no significant difference in peak height could be detected between freshly prepared solutions and solutions stored for up to 3 months.
The sample stability test for each AB revealed that the prepared and refrigerated QC samples are stable for at least 3 days at −80 C.
The differences between each measuring point corresponded to the usual variation of inter-day measurements. The values for QC low are presented in Table 4.
Because samples in our laboratory are stored for a maximum of 3 days, the stability at −80 C is sufficient for our purposes, and therefore, no stabilizer addition was required. A comparison with the literature also confirmed this. Naicker et al. (2018)  the less stable of the tested substances. Legrand et al. (2008) noted that Imi (besides Mero) remains stable for 3 days at −80 C, mentioning that Imi is the less stable one. However, by addition of a stabilizer, the authors found that the stability of both substances increased up to 50 days at −80 C. For Cila no data on stability in serum/plasma could be found in the literature.
However, Imi showed a limited stability in the prepared sample.
After 24 h in the autosampler, only approximately 70% of the initial value could be detected. Within 6 h more than 10% of the initial Imi concentration was degraded. All other ABs showed good vial stability and no significant degradation, as more than 90% of the initial concentration could still be measured after 24-h storage in the autosampler. Among these ABs, Mero was the less stable one with the greatest decrease in concentration after 24 h.

| Routine analysis
The method we describe in this study has been used for more than 9 months in our laboratory to determine serum levels of ABs in criti- This is promoted by the measurement of Taz at the lower wavelength of 220 nm, the low response factor of Taz, and the low retention time.
However, Taz is always co-administered with Pip.

| CONCLUSION
We have developed and validated a method for the determination of four β-lactam ABs (meropenem, imipenem, ceftazidime, and piperacillin) and two coadministered substances (Cila and tazobactam) in human serum in the clinically relevant range. The analysis was performed by a single run using HPLC and UV detection. By using a less common pentafluorophenyl column as a stationary phase, we successfully separated the ABs. Our method is based on a simple, but sufficient, protein precipitation and a subsequent dilution for sample preparation. The stability analysis confirmed the sufficient stability of samples and stock solutions at −80 C, even without the addition of a stabilization buffer, as well as vial stability within the working range of our laboratory. Among the six analytes, the carbapenems Imi and Mero were less stable than the other substances. The method described in this study showed good validation performance in terms of sensitivity, accuracy, and precision and has proven its feasibility in the routine analysis performed in our laboratory.

CONFLICT OF INTEREST
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.