Investigation of early antibiotic use in pediatric patients with acute respiratory infections by high‐performance liquid chromatography

Abstract In this study, we developed and validated two reliable high‐performance liquid chromatography (HPLC) methods for the qualitative detection of six oral β‐lactams, which are commonly used in pediatric patients with acute respiratory infections (ARIs). Two distinct reverse‐phase chromatographic separations of six β‐lactams were obtained. Four β‐lactams (cefadroxil, cephalexin, cefaclor and cefixime) in urine were separated using a gradient program with a mobile phase consisting of K2HPO4 buffer (20 mm, pH 2.8) and acetonitrile on a LichroCART 250 × 4.6 mm, Purospher STAR C18 end‐capped (5 μm) column. Two remained β‐lactams (amoxicillin and cefuroxime) were analyzed using a gradient elution with the mobile phase containing K2HPO4 buffer (20 mm, pH 3.0) and acetonitrile on a LichroCart® Purospher Star C8 end‐capped column (5 μm, 125 × 4.6 mm). Good linearity within the range of 0.3–30 μg/ml for cefadroxil, cephalexin, cefaclor and cefixime, and 0.2–20 μg/ml for amoxicillin and cefuroxime, was attained. The precisions were <14%. The accuracies ranged from 85.87 to 102.8%. The two validated methods were then applied to determine these six antibiotics in 553 urine samples of pediatric patients with ARIs. As a result, 32.2% were positive with one or more of six tested β‐lactams. Cefixime was the most commonly detected agent, accounting for 9.8% of enrolled patients.

children are viral infections, oral antibiotics are frequently prescribed, especially when viral or bacterial etiologies are indistinguishable (Do et al., 2011;van Gageldonk-Lafeber et al., 2005). Hence, the misuse of antibiotics is a common practice in the community setting of ARI treatment in Vietnam and other countries (Hoa et al., 2009;Landers, Ferng, Mcloughlin, Barrett, & Larson, 2010;Llor & Cots, 2009).
Irrational and inappropriate use of antibiotics can be recorded from questionnaire-based surveillance of the patients performed by healthcare providers during physical examination at admission (GARP, 2010;Jansen et al., 2006;Otters, van Der Wouden, Schellevis, van Suijlekom-Smit, & Koes, 2004). However, in Vietnam, for most cases of ARIs requiring hospitalization, the reliability and objectiveness of the information about previous exposure to specific antibiotics provided by patients at admission are questionable and unverified. Identification of specific antibacterial agents used prior to hospitalization in these cases using an analytical method could be helpful to verify such information and help clinicians to choose a rational antibiotic regimen.
Among all oral antibiotics, the six β-lactams (cefadroxil, cephalexin, cefaclor, cefixime, amoxicillin and cefuroxime) are the most commonly used antibacterial agents for the treatment of mild to moderate respiratory bacterial infections in both hospital and community settings in Vietnam (GARP, 2010). All of six β-lactams are excreted at a high rate in urine in unchanged forms (AHFS, 2009), making urine a valuable and effective biological matrix for the qualitative or quantitative determination of antibiotics in patients (El-Gindy, El Walily, & Bedair, 2000;Samanidou, Hapeshi, & Papadoyannis, 2003;Samanidou, Ioannou, & Papadoyannis, 2004). Many studies have described highperformance liquid chromatography (HPLC) methods for the determination of oral cephalosporins in various matrices such as plasma (Mcateer, Hiltke, Silber, & Faulkner, 1987;Samanidou et al., 2003;Verdier et al., 2011;Wolff et al., 2013) and urine (Samanidou et al., 2003;Samanidou et al., 2004). In addition, Cazorla Reyes et al. also developed a liquid chromatography tandem mass spectrometry method for the analysis of 21 antibiotics including cephalosporins and various other antibiotics in plasma, urine, cerebrospinal fluid and bronchial aspiration matrix (Cazorla-Reyes, Romero-González, Frenich, Rodríguez Maresca, & Martínez Vidal, 2014). Although several HPLC methods have been developed, some are limited to quantification of one or two drugs (Bhinge, Malipatil, & Sonawane, 2014;Danafar, 2015;El-Gindy et al., 2000;Khan et al., 2011), while others have been developed for the detection of 5-12 cephalosporins (Denooz & Charlier, 2008;Legrand, Vodovar, Tournier, Khoudour, & Hulin, 2016;Verdier et al., 2011;Wolff et al., 2013). In addition, many methods using liquid chromatography coupled to tandem mass spectrometric detection (LC/MS-MS) were set up to measure simultaneously from five to 11 cephalosporins (Carlier, Stove, De Waele, & Verstraete, 2015;Colin, DE Bock, T'jollyn, Boussery, & Van Bocxlaer, 2013). Of note, several sample preparation procedures were applied to treat samples before injection into the HPLC systems: protein precipitation (Carlier et al., 2015;Khan et al., 2011;Legrand et al., 2016;McAteer et al., 1987;Samanidou et al., 2003;Verdier et al., 2011;Wolff et al., 2013), liquid-liquid extraction (Bhinge et al., 2014) and solid-phase extraction (SPE) (Colin et al., 2013;Nemutlu, Kir, Katlan, & Beksac, 2009). A pre-dilution step is commonly applied to treat urine samples because the high ionic strength of this matrix interferes with the analyte detection methods (Eshra, Hassan, & El-Walily, 1993;Kovach, Lantz, & Brier, 1991;Najib, Suleiman, El-Sayed, & Abdulhameed, 1987;Samanidou et al., 2003;Samanidou et al., 2004). Additionally, some new extraction procedures have recently been applied to treat complex matrices aiming at reducing solvent consumption and the use of toxic solvents, in accordance with the Green Analytical Chemistry concepts. These innovative procedures also allowed analytical performance to be improved using modest and routine instrument configurations such as UV-vis detectors, while avoiding the use of more complex and expensive systems such as tandem mass spectrometers (Kabir, Locatelli, & Ulusoy, 2017). In the present study, two reliable HPLC with diode array detection methods were developed and validated for the detection of the six antibiotics in urine samples. Subsequently, the two distinct validated methods were applied on the samples collected at hospital admission from pediatric patients with ARIs who attended to the outpatient clinic of Children's Hospital 1, Ho Chi Minh City (CH1) to investigate the early use of these antibiotics in the community setting. and cefuroxime (CFU) were purchased from Sigma-Aldrich (Singapore).
The blank urine samples were collected from healthy children.

| Equipment
The liquid chromatography system was a LaChrom Elite (Merck-Hitachi, Japan) composed of an autosampler L-2200, 2 pumps L-2130, a column Oven L-2350 and a diode array detector (DAD) L-2455

| Preparation of standard solutions and quality controls
Stock solutions of CFD, CPL (10 mg/ml) and CFO (5 mg/ml) were prepared by dissolving the standards in water. CFI, AMO and CFU (2 mg/ml) were prepared by dissolving the standards in MeOH.
The stock solutions were combined and further diluted with water to obtain fresh working solutions for calibration curves (CC) ranging from 3 to 300 μg/ml for CFD, CPL, CFO and CFI, and from 2 to 200 μg/ml for AMO and CFU. Quality control (QC) stocks and working solutions were independently prepared in the solvents described above. The concentrations of low, medium and high QC working solutions were 5, 100 and 250 μg/ml for CFD, CPL, CFO and CFI, and 4, 40 and 160 μg/ml for AMO and CFU, respectively.
Urine samples for CC determination were prepared by diluting

| Sample preparation
Urine samples were thawed at room temperature for 20 min then mixed for 20 s before being centrifuged at 10,000g for 5 min at room temperature. After 10 min rest on the bench, 50 μL urine samples were mixed with 200 μL water. The mixture was then added to 250 μL of 4% formic acid. The resulting mixture was mixed by vortexing for 15 s and then rested for 2 min prior to being centrifuged at 10,000 rpm for 4 min.

| Method validation
A method validation of six β-lactams in urine was performed for selectivity, carryover, sensitivity, linearity, recovery, precision and trueness in accordance with USA Food and Drug Administration (FDA) bioanalytical method validation guidelines (FDA, 2013).

| Selectivity, carryover and sensitivity
Selectivity was evaluated by analyzing urine samples from six different healthy children and the chromatograms were evaluated for any sign of potential interferences with the drug identification and measurement. No interfering components were considered to be present when the signal was <20% of the limit of quantification (LLOQ) for analytes.
Carryover was evaluated by injecting two blank samples in the system after a duplicate of the highest calibration standard (upper limit of quantification, ULOQ) was run. The signal in the blank sample run after injection of the highest calibration standard should not be greater than 20% of the LLOQ. Sensitivity was defined by the limit of detection (LOD) and the limit of quantification (LLOQ). The LOD was expressed as the concentration producing a signal three times higher than the noise from blank samples. The LLOQ was defined as the lowest concentration of analyte which could be determined with acceptable trueness and precision. The analyte signal of the sample with the LLOQ should be from 5 to 10 times higher than the baseline noise of six blank samples.

| Calibration curve and linearity
Linearity was assessed using the calibration curve consisting of eight calibration standards (including blank samples). For each antibiotic, calibration curves were obtained using the peak area of analytes (y) vs concentrations (x). The best regression models with or without data transformations and weighting factors were evaluated and selected using model options in HPLC System Manager Software-EZChrom Elite version 3.18 (Hitachi-Japan). This was intended to control the heteroskedasticity of our data and have the best predicted backcalculated concentrations from unknown samples.

| Extraction recoveries
Antibiotic extraction recoveries were measured at three different levels (corresponding to the QC samples with low, medium and high concentrations) and determined by comparing the peak areas of the analytes spiked in urine after the extraction process with the peak area of an unprocessed standard solution prepared with identical concentrations.

| Precision and trueness
The intra-day precision and trueness were evaluated at three different Acceptance criteria were as follows: trueness error had to be within 85-115% of the nominal value, and intra-and inter-precisions had to be <15%.

| Method application on clinical samples
The urine samples were collected from pediatric patients who had a  Table 2 and their recoveries are shown in Table 3.
In the development of chromatographic conditions for six βlactams, we intended to simultaneously analyze these antibiotics in one run. However, SCX, the best SPE sorbent for four β-lactams

| Validation
The validation methods of six β-lactams in human urine was performed to assess selectivity, sensitivity, recovery, linearity, precision and trueness in compliance with the FDA's bioanalytical method validation guidelines (FDA, 2013).

| Selectivity, carryover and sensitivity
No interference peaks were observed in the drug-free human urine samples following sample pretreatment procedures for six β-lactams.
The analytes were well defined and separated from matrix contaminants, with symmetrical peak shapes at the respective retention time for the six β-lactams. In all validation batches, no interferences from urine were found co-eluating with analytes, thus ensuring the selectivity of the methods. This was also confirmed throughout the entire clinical sample analysis.
Injection of blank samples directly after injection of the ULOQ showed a signal <20% of the LLOQ for each antibiotic, thus satisfying the acceptance criteria for the carryover.

| Calibration and linearity
The concentration range of the methods consisted of eight points (including blank urine). Methods were found to be linear over concentrations of 0, 0.3, 0.6, 2, 6, 12, 20 and 30 μg/ml (CFD, CPL, CFO and CFI) and 0, 0.2, 0.5, 1, 2, 5, 10 and 20 μg/ml (AMO and CFU). The calibration curve was determined by plotting the peak areas vs. the concentrations of analytes using regression analysis. The best models were quadratic, log-log transformed and 1/x weighting regression models (for CFD and CPD) and 1/x 2 regression models (for CFO and CFI). For AMO and CFU, linear regression with log-log transformation and nonweighting were the optimal models. The high coefficients of regression were achieved (r 2 > 0.99) for all tested β-lactams.

| Recoveries
The mean recoveries of six β-lactams in all QC levels are shown in Table 3. The lowest recovery was 65.87% for AMO and the highest was 102.5% for CFO.

| Precision and trueness
The precision and trueness of six β-lactams, overall, were all satisfactory. The intra-and inter-assay relative standard deviations were always <14%. The accuracies ranged from 85.87 to 102.78% at all QC levels for six β-lactams in samples. The detailed results of the assay are shown in Table 3.
Even though the stability test was not performed as per the guidance instruction in our current HPLC methods, the stability of these antibiotics in different matrices was well established and confirmed in previous publications (Denooz & Charlier, 2008;El-Gindy et al., 2000;Legrand et al., 2016;Nemutlu et al., 2009;Samanidou et al., 2003;Verdier et al., 2011). In addition, internal standards were not applied in this assay to correct any internal bias in the whole procedures. This could explain the significant variance seen at some levels.
However, the precision and trueness were in fair agreement and satisfied the guidance.

| Application to clinical samples
The urine samples were collected from 563 pediatric patients under 16 years of age (50% of patients were less than 2 years old, and 95% of patients under 5 years old). The validated methods were successfully applied to determine the six β-lactams in urine samples (10 patients with severe ARIs were anuria). Figure 2 presents the results of a qualitative measurement of the β-lactams in clinical samples.
The results demonstrated that 32.2% of patients had positive results with at least one out of the six investigated β-lactam antibiotics. Our findings were similar to a study conducted in India, which also indicated that 31% of 64 children with febrile illness had antibiotics in the urine (Mathew et al., 2010). Of note, amoxicillin was found as the most frequently used antibiotic in the Indian study, at 17.2%, compared with cefixime (9.8%) in our study. However, both amoxicillin and cefixime were detected at the highest proportion and the same rate (8.8-25/286) in patients <2 years of age in our study. The difference might be due to the considerable disparity in the sample size as well as the kinds of antibiotics tested in the two studies. In the Indian study, the urine samples were assayed for five antibiotics including trimethoprim/sulfamethoxazole, amoxicillin, ciprofloxacin, cephalexin and cefuroxime while we tested for amoxicillin, cephalexin and cefuroxime and three other β-lactams (cefadroxil, cefaclor and cefixime). Since other commonly used antibiotics (macrolides, fluoroquinolones and sulfamethoxazole/trimethoprim) were not tested by our current HPLC methods, we might assume that more than one-third of patients with mild ARIs in our study had been given antibiotics prior to hospitalization.
As regulated by the Drug Administration of Vietnam, antibacterial agents must be purchased with a medical prescription (VNMOH, 2003). However, it has been shown that most antibiotics were easily obtained without prescription in the community drugstores of Vietnam (Hoa et al., 2009;Landers et al., 2010;Larsson et al., 2000;Llor & Cots, 2009). Our finding was consistent with this real situation and provided evidence that patients easily took the antibiotics without prescription for their fever illness. In our study, cefixime, a broad-spectrum third-generation cephalosporin, was the most commonly detected agent, accounting for 27% of tested antibiotics and about 9.8% (54/553) of enrolled patients. A major reason for dispensing broad-spectrum antibacterial agents in the community setting might be due to inadequate antibiotic coverage by health workers in the pharmacies and/or private clinics (Hulscher, Grol, & Van Der Meer, 2010). In addition, 3.2% (18/553) of urine samples were positive with more than one β-lactam antibiotic. Our findings demonstrated that patients might receive more than one antibiotic in a visit or they might visit multiple private clinics and/or pharmacies to obtain antibiotics without attempting to identify the causative organisms during their illness period. The overuse of broadspectrum cephalosporins and inappropriate combination of antibiotics raised a red alert on the misuse of antibiotics in Vietnam