Stir bar-sorptive extraction , solid phase extraction and liquid-liquid extraction for levetiracetam determination in human plasma : comparing recovery rates

Levetiracetam (LEV), an antiepileptic drug (AED) with favorable pharmacokinetic profile, is increasingly being used in clinical practice, although information on its metabolism and disposition are still being generated. Therefore a simple, robust and fast liquid-liquid extraction (LLE) followed by highperformance liquid chromatography method is described that could be used for both pharmacokinetic and therapeutic drug monitoring (TDM) purposes. Moreover, recovery rates of LEV in plasma were compared among LLE, stir bar-sorptive extraction (SBSE), and solid-phase extraction (SPE). Solvent extraction with dichloromethane yielded a plasma residue free from usual interferences such as commonly co-prescribed AEDs, and recoveries around 90% (LLE), 60% (SPE) and 10% (SBSE). Separation was obtained using reverse phase Select B column with ultraviolet detection (235 nm). Mobile phase consisted of methanol:sodium acetate buffer 0.125 M pH 4.4 (20:80, v/v). The method was linear over a range of 2.8-220.0 μg mL-1. The intraand inter-assay precision and accuracy were studied at three concentrations; relative standard deviation was less than 10%. The limit of quantification was 2.8 μg mL-1. This robust method was successfully applied to analyze plasma samples from patients with epilepsy and therefore might be used for pharmacokinetic and TDM purposes.


INTRODUCTION
Levetiracetam (LEV; α-ethyl-2-oxo-1-pyrrolidine acetamide; Figure 1) is an antiepileptic drug (AED) approved by the Food and Drug Administration to be prescribed in polytherapy for management of partial onset seizures with or without secondary generalization in adults and children with refractory epilepsy, and as monotherapy for treatment of partial onset seizures in patients with newly diagnosed epilepsy (Mbizvo et al.,2012;Patsalos, 2004).
Its innovative mechanism of action based on the selective binding to a synaptic vesicle protein (SV2A) involved in neurotransmitter release (Macdonald;Rogawski, 2008) and the favorable pharmacokinetic profile makes LEV a promising AED in terms of efficacy and tolerability.Levetiracetam presents high oral bioavailability, rapid gastrointestinal absorption (T MAX = 1 h), linear pharmacokinetics over the therapeutic dosages range (1000 -3000 mg/day), minimal plasma protein binding (<10%), and elimination half-life estimated at 6 to 8 h in adults.According to Strolin Benedetti and colleagues (2003), LEV does not undergo chiral interconversion in humans.In contrast to most other AEDs, LEV elimination is not significantly dependent on the cytochrome P450 enzymatic system; about 66% of an oral dose are excreted unchanged in urine, and 24% undergo hydrolysis in blood to the inactive metabolite ucb L057 (Patsalos, 2004).As a relatively new AED, LEV pharmacokinetics is still under investigation, especially regarding its disposition and elimination in situations such as pregnancy and extremes of ages (Italiano, Perucca, 2013;Tomson, Landmark, Battino, 2013).
Despite the wide reference range for plasma concentration values (12-46 mg/L) and the positive pharmacokinetic profile, therapeutic drug monitoring (TDM) of LEV is advisable to improve efficacy and safety in special situations, namely pregnancy, renal impairment, in elderly and pediatric patients, to investigate intoxication, therapeutic compliance or overdose (Krasowski, 2010;Patsalos et al., 2008;Tomson et al., 2007).Several chromatographic methods have been published for quantification of LEV concentrations in biological fluids such as saliva and plasma (or serum) for TDM and in urine for pharmacokinetic aims.They include different subtypes of liquid chromatography with various detection systems (Contin et al., 2008;Grim et al., 2003;Guo et al., 2007;Jain et al., 2006;Juenke et al., 2006;Kuhn, Knabbe, 2013;Lancelin et al., 2007;Martens-Lobenhoffer, Bode-Boger, 2005;Matar, 2008;Pucci et al., 2004;Rao et al., 2004;Ratnaraj, Doheny, Patsalos, 1996;Shibata et al., 2012), gas chromatography (Isoherranen et al., 2000;Mecarelli et al., 2007) and microemulsion electrokinetic chromatography (Ivanova et al., 2003).An analytical nonchromatographic method, capillary electrophoresis, is also described (Shihabi, Oles, Hinsdale, 2003).Because of the advantages of high-performance liquid chromatography (HPLC) over other analytical methods, namely in terms of versatility and simplicity of sample preparation, as well as wide linearity range in detectors, relative low cost and widespread use in laboratories, it has become the gold standard method for TDM not only for LEV, but also for all other AEDs monitored (Queiroz et al., 2008).
Different sample pretreatment techniques have been described for LEV analysis, with recoveries ranging from 12% to 99%, according to the method and reagents.Solidphase extractions (SPE) of LEV performed with different polymeric sorbents such as C 18 and hydrophilic-lipophilic balanced copolymer are described in the literature (Isoherranen et al., 2000;Ivanova et al., 2003;Mecarelli et al., 2007;Pucci et al., 2004;Vermeij, Edelbroek, 1994).Protein precipitation is considered a simple, inexpensive and suitable pretreatment procedure especially for drugs minimally bound to plasma proteins.Preparation usually consists of adding organic solvents such as methanol or acetonitrile; formation of insoluble salts through addition of zinc sulphate or perchloric acid is also possible (Contin et al., 2008;Martens-Lobenhoffer, Bode-Boger, 2005;Pucci et al., 2004).Although dilutions can affect the sensitivity of the method, this procedure is considered a positive pretreatment method because LEV is not significantly bound to plasma proteins (Pucci et al., 2004).
More recently, stir bar-sorptive extraction (SBSE) has been evaluated for the enrichment of organic solutes from biological fluids.It is a non-exhaustive sample preparation technique based on the partitioning coefficient of the solutes between the non-polar silicone phase made of polydimethylsiloxane (PDMS) and the aqueous phase.Although this process is still under deeper investigation, it is becoming more popular for different matrices and analytes (David, Sandra, 2007), including therapeutic drugs (Queiroz et al., 2008).
Considering that analytical procedures allowing rapid and robust quantifications of LEV concentrations in biologic fluids are of great interest both for TDM and pharmacokinetic applications, this work aimed to present a study of a reliable analytical method based on comparisons of recovery rates from LLE, SBSE and SPE cleanup procedures followed by HPLC analysis for determination of LEV concentrations in plasma samples.Noteworthy, as the purpose is centered in the comparison of the above-mentioned extraction methods, validation aspects are shown as pertinent.

MATERIAL AND METHODS
The plasma samples used in the present study were obtained from patients with epilepsy attending the Ribeirão Preto Clinical Hospital.Procedures were performed under written consent and approved by the local Ethics Committee (Approval nº 11133/2008).

Instrumentation and reagents
The analyses were performed on an HPLC system consisting of a Shimadzu LC 10AD pump (Japan), a Shimadzu SPD 10A ultraviolet (UV) detector, a Chromatopac C-R8A integrator (Shimadzu, Columbia, MD, USA) and a Rheodyne injector with a 20 μL loop.
Supleco ® cartridges (C 18 , 3 mL) were used for SPE.The procedure was carried out on a vacuum apparatus (Marconi, Brazil).For SBSE, the commercial stir bar Twister ® (10 mm long glass-encapsulated magnetic stir bar externally coated with 22 μg of PDMS) was obtained from Gerstel (Gerstel GmbH, Mulheim Ruhr, Germany).

Calibration curve
Stock solutions of LEV in water (0.22 mg mL -1 ) and of IS in methanol (260.0 μmol L -1 ) were prepared and stored at -20 °C.These solutions were stable for at least five months.Routine daily calibration curves were prepared for each analytical batch by adding LEV and IS working solutions to drug-free plasma in order to produce LEV concentrations of 7. 13, 11.88, 19.8, 33.0 and 55.0 µg mL -1 .

Analytical conditions
Chromatographic separation was achieved isocratically, at room temperature, using an analytical column (250x4 mm i.d.) with precolumn (4x4 mm) packed with LiChrospher 60 RP-select B, 5 μm particle size (Merck ® ).Mobile phase consisted of a mixture of methanol and sodium acetate buffer 0.125 M, pH 4.4 (20:80, v/v) daily prepared and degassed.The mobile phase was delivered into the HPLC system at a flow-rate of 1.0 mL min -1 .The UV detector was set at 235 nm.

Sample preparation for SPE
The SPE cartridge was conditioned with 2 mL of methanol followed by 2 mL of water.A mixture of 100 mL of plasma with 25 mL of IS and 3 mL of borate buffer (0.1 M, pH 9.0) was mixed in a vortex-type shaker for 1 min, and loaded into the cartridges which were previously dried under low vacuum at room temperature.The cartridges were then washed with 5 mL of water and dried under vacuum for 5 min.Elution was performed with 2 mL of dichloromethane under low vacuum at room temperature.The eluate was transferred into conic tubes to be evaporated under constant airflow.The dry residue was reconstituted in 50 mL of the mobile phase and, after vortexing for 10 s, 20 mL were injected into the HPLC system.

Sample preparation for SBSE
Prior to first use, the stir bar was placed into a vial containing acetonitrile and conditioned for 24 h.A mixture of 100 mL of plasma sample with 3 mL of borate buffer (0.1 M, pH 9.0) and 25 mL of IS was vortexed for 1 min at room temperature.The stir bar was immersed into the sample and the extraction was performed under a magnetic stirring rate of 1200 rpm for 30 min at 50 ºC.For desorption, the stir bar was firstly removed with tweezers, carefully dried in clean tissue and put in a glass tube containing 2 mL of dichloromethane, ensuring total immersion.Desorption was performed under a magnetic stirring rate of 1200 rpm for 30 min at 50 ºC.After the desorption step, the stir bar was removed by means of a magnetic rod and the solvent was evaporated to dryness at room temperature under airflow.The dry residue was reconstituted in 50 mL of mobile phase and 20 µL were injected into the HPLC system after shaking for 10 s in a vortex-type shaker.
After each the successive extraction the used stir bar was cleaned in methanol for 30 min at 50 °C (under a magnetic stirring rate of 1200 rpm), followed by a drying step using a lint-free tissue.

Sample preparation for LLE
Liquid-liquid extraction was based on that previously proposed by Ratnaraj, Doheny and Patsalos (1996) with minor modifications herein described.Twenty five microliters of IS were added to 25 µL of NaOH 5 M, 2 mL of dichloromethane and 100 µL of plasma.The mixture was shaken in a vortex-type shaker for 1 min and then centrifuged at 1800 rpm for 5 min.The organic phase was transferred to conic tubes and evaporated under constant airflow at room temperature.The dry residue was reconstituted in 50 mL of mobile phase and, after shaking for 10 s in a vortex-type shaker, 20 mL were injected into the HPLC system.

HPLC method validation
Recovery rates were determined at the five different concentrations from the calibration curve.Levetiracetam was extracted in triplicate from human blank plasma according to each method proposed above and the UVpeak heights were compared to the direct injection of standard solutions of equal concentrations.
In order to obtain intra-and inter-assay precisions, aliquots (n=10 and n=5, respectively) of blank plasma containing LEV and IS at three different concentrations were analyzed.The inter-assay precision was determined based on analyses of blank plasma containing LEV and IS solutions at the same three concentrations for five consecutive days.Accuracy was checked using three concentrations within the linear range of the procedure.The accuracy values accepted showed variations of less than 15%.
Linearity was determined by analyzing drug-free plasma samples (n=3) containing LEV and IS solutions at concentrations from 2.85 to 220.0 μg mL -1 , a range wide enough to cover the routine calibration curve (from 7.13 to 55.0 µg mL -1 ).The concentration range was estimated based on the regression curve (y=ax + b) and correlation coefficient (R 2 ).
The limit of quantification (LOQ) was determined by analyzing drug-free plasma samples spiked with decreasing concentrations of LEV standard solution.The LOQ was considered the lowest concentration quantified with a relative standard deviation (RSD) of less than 10%, obtained from five determinations.

RESULTS
Table I displays the concentration range to which linearity was determined, the regression curve and correlation coefficient referred to the range, and the comparison of LEV recovery rates from the different extraction procedures evaluated.
Validation parameters for LLE are shown in Table II.Figure 2 displays representative chromatograms from the validated method applied to clinical samples.

DISCUSSION
Three different sample preparation procedures were tested and compared for LEV extraction from human plasma.Recovery rates were assumed as the most important figure of merit in comparing the cleanup procedures, i.e., methods with low recovery rates (<80%)

LEV conc. (μg.mL -1 ) Precision Intra-assay RSD (%) n=10
Precision Inter-assay RSD (%) n=5 Accuracy error (%) 13.0 5.4 5.0 4.9 15.0 6.0 5.0 9.6 30.0 7.0 6.6 3.8 RSD: relative standard deviation; LEV: levetiracetam.were considered unsatisfactory for our aims and eventually were not validated.Although recovery rate was the most important parameter in determining the validity of the method, other aspects such as complexity of execution, time-consumption and amounts of organic solvents required were also balanced in order to discuss the most suitable method for TDM and pharmacokinetic aims.
The stationary phase (C 18 ) here selected for SPE analyses is considered a reasonable choice according to Isoherranen and colleagues (2000), Ivanova and colleagues (2003), Vermeij and Edelbroek ( 1994) and Chollet ( 2002), although Pucci and co-workers (2004) argue against the reproducibility of the results obtained from this sorbent among different batches.Considering the high hydrophilic characteristic of LEV, our present elution performed with dichloromethane can possibly explain the low recovery rates found (Table I) and to what is reported elsewhere (Isoherranen et al., 2000;Pucci et al., 2004;Vermeij,Edelbroek, 1994;Zufia et al., 2010).Among the organic solvents commonly used in SPE, dichloromethane has a polarity index of 3.4, while it is 6.6 for methanol.This is suggestive that mixtures with higher polarity indexes such as acetonitrile and methanol (Zufia et al., 2010) or methanol and water (Vermeij, Edelbroek, 1994) would extract LEV from stationary phase efficiently.Despite the main advantages of SPE, namely the purification and enrichment of analytes from complex matrices and the use of low amounts of organic solvents, it is an expensive and time-consuming procedure, not reliable especially for analyses of large amounts of samples as in pharmacokinetic studies and routine TDM analyses.Furthermore, in our specific case, the low recovery rates found would not be suitable for neither pharmacokinetics nor for TDM purposes.
To our knowledge, no publication in the literature has tried to extract LEV from any matrix using SBSE.This extraction method has proven to be an interesting and environmentally friendly alternative to LLE considering its insignificant organic solvent consumption and higher limits of detection (David, Sandra, 2007).However, the even lower recovery rates obtained from SBSE (Table I) make this procedure a non-suitable cleanup method.As bars with different coatings other than PDMS are released, this method could be optimized for extracting LEV from human plasma.
Liquid-liquid extraction modified from Ratnaraj and co-workers (Ratnaraj;Doheny;Patsalos, 1996) proved to be the most satisfactory of the evaluated methods.Due to the high recovery rates (Table I), validation parameters were established only for LLE, as illustrated in Table II.The LOQ found was 2.85 μg mL -1 .Calibration curves were linear in the range of 2.85-220.0μg mL -1 , with R 2 = 0.991.Linear regression equation was defined as y=0.0245x+0.1842.Accuracy and intra-and interassay precision were studied at three concentrations, as summarized in Table II.The RSD were less than 10%.
The small amount of biological matrix needed (100 μL) makes this a suitable method even for children.The pretreatment steps presented here are free of toxic solvents such as petroleum ether and chloroform (Isoherranen et al., 2003) and provided results without any interference from drugs normally used with no prescription.Accordingly, there were no interfering peaks related to endogenous nor exogenous compounds, as can be observed on the chromatograms (Figure 2), where drugfree human plasma, spiked human plasma and clinical samples from patients with epilepsy show the ability of the method to quantify LEV unequivocally.
Several mobile phases were tested and the proposed (methanol:sodium acetate buffer 0.125 M, pH 4.4, 20:80 v/v) was the simplest that yielded satisfactory separation in a acceptable run time with retention times of 6.5 min for LEV and 8.0 min for IS (Figure 2).The complete elution was obtained in less than 20 min through isocratic mode, in contrast to Vermeij and Edelbroek ( 1994) and Martens-Lobenhoffer and Bode-Boger ( 2005) who performed gradient runs for their analyses.Remarkably, the isocratic mode of elution is still the preferred choice when compared to gradient elution, possibly because of improvements in column efficiencies as well as issues related to equilibration time, throughput, ease of use and availability of suitable pumps (Chollet, 2002).
Solvent extraction with NaOH and dichloromethane, the mobile phase described, and detection at 235 nm showed high sensitivity and resolution for detection of LEV in human plasma.Different wavelengths between 205 and 240 nm were tested.Considering that real clinical samples can be difficult to yield chromatographic peak separations because of comedications and endogenous interferences, a high wavelength set at 235 nm showed to be the best for selectivity instead of the UV detection at 205 nm (very unspecific) described elsewhere (Contin et al., 2008;Martens-Lobenhoffer, Bode-Boger, 2005;Pucci et al., 2004;Zufia et al., 2010).
The polytherapies frequently prescribed to approximately one third of all patients with epilepsy who do not achieve satisfactory seizure control with currently available AEDs in monotherapy (Sander, 2003) often include drugs such as carbamazepine, phenytoin, phenobarbital and benzodiazepines.Figures 2C and 2D illustrate the selectivity of the method on two plasma sample chromatograms, as they show the absence of interfering peaks of lamotrigine (2C) and carbamazepine (2D), two commonly used AEDs, in LEV and IS analyses.Chromatograms shown in Figure 2C and 2D were obtained from adult patients with epilepsy chronically treated with lamotrigine 200 mg/day and carbamazepine 600 mg/day, respectively, after receiving a single oral dose of LEV 1000 mg.Although no reference range has been definitely established for LEV, clinical studies published using adult patients with epilepsy show LEV plasma concentrations mostly ranging from 12 to 46 µg mL -1 (Patsalos et al., 2008), values fully in accordance with our proposed quantification range.

CONCLUSIONS
This paper presents a simple, robust and fast method for determination of LEV in human plasma using reverse-phase HPLC combined with UV detection.The method is compatible with pharmacokinetic studies and TDM.Based on figures of merit results, the LLE procedure proved to be adequate to LEV analyses from sub-therapeutic to toxic levels.The assay was successfully applied to the analysis of real samples once it could be considered selective, especially for patients with epilepsy in chronic treatment with AEDs other than LEV.Taken into account that the core of the present study is the methods comparison essentially based on their recovery rates, LLE was considered a shorter step, precise and an inexpensive sample preparation method to be routinely used instead of the other methods evaluated.

FIGURE 2 -
FIGURE 2 -Chromatograms from the validated method.(A): drug-free plasma sample (blank); (B): plasma spiked with LEV and IS to the final LEV concentration of 55 μg mL -1 ; (C): clinical sample from an adult patient with epilepsy on lamotrigine chronic treatment (200 mg/day) obtained 9 hours after the ingestion of a LEV single dose (1000 mg); (D): clinical sample from an adult patient with epilepsy on carbamazepine chronic treatment (600 mg/day) obtained 1 hour after the ingestion of a LEV single dose (1000 mg).IS: internal standard; LEV: levetiracetam.

TABLE I -
Comparison of LEV recovery rates from the different extraction procedures LEV: levetiracetam; LLE: liquid-liquid extraction; SBSE: stir bar-sorptive extraction; SPE: solid phase extraction.

TABLE II -
Intra-and inter-assay precision and accuracy data for LEV according to the validated method