Development and Validation of an LC-MS/MS Method and Comparison with a GC-MS Method to Measure Phenytoin in Human Brain Dialysate, Blood, and Saliva

Phenytoin (PHT) is one of the most often used critical dose drugs, where insufficient or excessive dosing can have severe consequences such as seizures or toxicity. Thus, the monitoring and precise measuring of PHT concentrations in patients is crucial. This study develops and validates an LC-MS/MS method for the measurement of phenytoin concentrations in different body compartments (i.e., human brain dialysate, blood, and saliva) and compares it with a formerly developed GC-MS method that measures PHT in the same biological matrices. The two methods are evaluated and compared based on their analytical performance, appropriateness to analyze human biological samples, including corresponding extraction and cleanup procedures, and their validation according to ISO 17025/FDA Guidance for Industry. The LC-MS/MS method showed a higher performance compared with the GC-MS method. The LC-MS/MS was more sensitive, needed a smaller sample volume (25 µL) and less chemicals, was less time consuming (cleaning up, sample preparation, and analysis), and resulted in a better LOD (<1 ng/mL)/LOQ (10 ng/mL). The calibration curve of the LC-MS/MS method (10–2000 ng/mL) showed linearity over a larger range with correlation coefficients r2 > 0.995 for all tested matrices (blood, saliva, and dialysate). For larger sample numbers as in pharmacokinetic/pharmacodynamic studies and for bedside as well as routine analyses, the LC-MS/MS method offers significant advantages over the GC-MS method.


Introduction
Sensitive and specific quantification methods are of critical importance when monitoring individualized drug therapy in patients or investigating drug concentration in forensic toxicology [1]. Critical dose drugs but also newly developed and designed complex drugs require analytical methods to check for effective drug delivery to target tissues and to minimize toxicity in sensitive organs or cells. When such drugs have to be used in patients with varying pharmacokinetics (PK) (e.g., ICU patients), an appropriate therapeutic drug monitoring (TDM), which allows, for example, to correlate the drug concentration in easy accessible plasma samples with those in the tissue of action, becomes even more relevant for a safe and efficient drug treatment [2].
Phenytoin (PHT) belongs to the most widely prescribed drugs to prevent and control most types of seizure disorders and to treat epilepsy [3]. It is one of the most often used critical dose drugs where insufficient or excessive dosing can have severe consequences such as seizures or toxicity. us, the monitoring and precise measuring of PHT concentrations in patients is crucial [4,5]. As an example, in forensic toxicology, epilepsy patients under PHT treatment who have been involved in an accident have to be analyzed in order to verify whether the PHT concentration was adequate or possibly the reason for the accident [6]. However, there are several characteristics of PHT including a relatively low therapeutic index, difficult pharmacokinetics (PK) and pharmacodynamics (PD), saturable oxidative biotransformation, and the nonlinear clearance, which complicate a therapeutic drug monitoring (TDM) aimed at preventing intoxication of patients or treatment failures [7]. us, researchers and practitioners are interested in specific, sensitive, robust, and cost-effective methods to identify PHT concentrations in patients. ereby, several compartments to measure the PHT concentration could be addressed such as blood, saliva, and CNS fluid (microdialysate). e correlation of PHT in different body compartments is not yet completely understood and has only recently been addressed by researchers who have compared the measurement of PHT in these different compartments with a GC-MS method [8]. While the GC-MS has long been the standard method in forensic testing, LC-MS/MS methods have become more common, as they generally offer some advantages over GC-MS [9]. Recently, researchers have developed an LC-MS/MS method to measure PHT in one specific body compartment (i.e., blood plasma or serum) [10]. Missing, however, is a thorough comparison of the performance of these two analytical methods in the detection and analysis of PHT in different body compartments (i.e., blood, saliva, and samples from brain tissue microdialysis).
e aim of the present study was to develop and validate an LC-MS/MS method for the measurement of PHT concentrations in different body compartments such as blood and saliva, as well as samples from brain tissue microdialysis often used in neurology and neurosurgery, where antiepileptic therapy is often mandatory [11,12], and to compare its efficiency with a formerly developed GC-MS method [8].
e fact that this established GC-MS method measured PHT in the same biological matrices (i.e., blood, saliva, and human brain dialysate) enables a reliable comparison with regard to the performance of GC-MS versus LC-MS/MS in measuring PHT in different body compartments. e two methods are evaluated and compared based on their analytical performance, appropriateness to analyze human biological samples, including corresponding extraction and cleanup procedures, and their validation according to ISO 17025/FDA Guidance for Industry [13]. Finally, the suitability of the two analytical methods for PK/PD studies, bedside measurement, and forensic use is discussed. In addition, the LC-MS/MS method developed in the current study is compared with an established LC-MS/MS method which measured PHT in blood plasma samples [10]. Calcium chloride, perchloric acid, citric acid monohydrate, potassium chloride, magnesium chloride hexahydrate, sodium chloride, sodium hydroxide, and the solvents (methanol, acetic acid 100%, and acetone) were of analytical grade and purchased from Merck (Darmstadt, Germany).

Materials and Method
Artificial cerebrospinal fluid (aCSF; dialysate solution) was prepared according to M Dialysis AB (Stockholm, Sweden) [14]. Blood CPDA-1 (anticoagulant citrate phosphate dextrose adenine solution; to simplify only named blood in the following) was obtained from the Blood Donor Center (Bern, Switzerland). Saliva was obtained from one of the investigators. 20-60 μL PHT-containing samples from patients collected from a 2 µL/min flow rate brain microdialysis and 2 mL of CPDA containing PHT patient blood samples were provided by the Department of Neurosurgery (Kantonsspital Aarau AG, Switzerland and Centre Hospitalier Universitaire Vaudois, Switzerland). All biological samples (blood and dialysates) were frozen and stored at −24°C. Before sample analysis, the samples were thawed at room temperature for 30 minutes and homogenized by shaking with a vortex for one minute.

Internal Standards, Calibrator Standard System Suitability Testing, and Sample Preparation.
e internal standard (IS) stock solution was prepared by adding 100 µL of the PHT-D 10 (100 µg/mL) to 9900 µL of MeOH. 5 mL of this solution was added to 95 mL of 1 M perchloric acid aqueous solution to get the final concentration of 50.0 ng/mL, which is used as IS working solution.
e PHT reference stock solution (1.00 mg/mL) was used to obtain eight calibration (Cal) solutions with concentrations of 2000, 1000, 500, 250, 100, 50, 20, and 10 ng/mL PHT. 20 µL of these Cal solutions were added to 980 µL of the biomatrices to get the Cal working solutions. For quality control (QC), solutions with 1600, 400, 30, and 10 ng/mL PHT were prepared out of PHT reference stock solution (1.00 mg/mL).
e IS working solution of 75 µL was added either to an aliquot of 25 µL Cal working solution, QC solutions, or 25 µL sample from patients containing PHT. e sample preparation for the LC-MS/MS consisted of pipetting 75 µL of IS working solution to 25 µL sample into a deep well plate (0.6 mL, Chemie Brunschwig AG, Basel, Switzerland) covered by a sealing mat (Silicone, Chemie Brunschwig AG, Basel, Switzerland). e well plates were rigorously shaken for 5 minutes and then centrifuged (4.500 U/min; Mikro 22R, Hettich Instruments, Andreas Hettich AG, Bäch, Switzerland) for 30 minutes at about 8°C ( Figure 1). e processed samples were ready for the LC-MS/MS analysis.
is system was operated by Analyst Software (Version 1.5.1, AB Sciex, Toronto, Canada).

Validation of the LC-MS/MS Method according to ISO 17025/FDA Guidance for Industry.
e validation was carried out according to ISO 17025/FDA Guidance for Industry including selectivity, sensitivity, accuracy, recovery of PHT, reproducibility and suitability of the calibration curves, stability of PHT, and matrix effects. e selectivity and sensitivity (absence of PHT) were verified by analyzing blank samples without PHT (extraction and matrix effects). For the accuracy, QCs and Cal samples were analyzed. e recovery of PHT was analyzed by measuring QCs at different levels.
e reproducibility and suitability of the calibration curves was measured by a complete series of Cal 1 to Cal 8 (LC-MS/MS) analyses. e limit of detection (LOD) and the limit of quantification (LOQ) were analyzed using Cal 1 (LC-MS/MS 10 ng/mL PHT). e LOD was checked as a signal-to-noise (S/N) ratio of more than 4 : 1. e LOQ was considered as 5 times the response to a blank sample. e stability tests consisted of the freeze-thaw stability of PHT, which was determined after 3 freeze-thaw cycles. e shortterm stability was analyzed by keeping the samples thawed at ambient temperature for at least 6 hours, frozen for at least 12 hours at −25°C ± 5°C, and again thawed, worked-up, and analyzed. Postpreparative stability was evaluated to determine whether an analytical run can be reinjected in the case of instrument failure and, furthermore, whether the preparation of a large number of samples could be done at once. erefore, one of the validation runs was analyzed a second time after 7 days. e described criteria for Cal curves, QC, accuracy, and precision had to be met.   Matrix effects were analyzed by comparing the calibration curves generated with the three matrices aCSF, blood, and saliva. PHT microdialysis and blood samples from patients were analyzed to demonstrate the suitability of the method for biological samples from patients.

Comparison of the LC-MS/MS and the GC-MS
Method.
e LC-MS/MS method was evaluated and compared with the GC-MS method published by Hösli et al. [8] with regard to its analytical performance, appropriateness to analyze human biological samples, including corresponding extraction and cleanup procedures, and its validation according to ISO 17025/FDA Guidance for Industry. e statistical data were calculated with Microsoft Excel and IBM SPSS Statistics 22. To compare the different matrices, a one-way ANOVA was calculated. e corresponding values were checked for significance by t-tests.

Validation of the LC-MS/MS Method.
e retention time (RT) for PHT and for PHT-D 10 (IS) was about 2.8 min (Figure 3). e selectivity and sensitivity were checked; all blank samples were negative. e recovery of PHT after precipitation with HClO 4 was 89.5% for QC1 (10 ng/mL) and 97.1% for QC3 (1600 ng/mL) compared to the amount found in unprepared samples (�100%). e LOD calculated as S/N ratio of 4 : 1 for this method in aCSF, saliva, and blood was set at <1 ng/mL. e LOQ calculated as 5 times the response/blank was 10 ng/mL PHT. For the accuracy, the Cal 1 to Cal 8 were assessed. e calibrator values showed min-max deviations of 1-8% for Cal 2 (20 ng/mL) to Cal 8 (2000 ng/mL) with 3% for Cal 1 (10 ng/mL). e calibration curves for all three matrices were linear. e regression coefficients (r 2 ) of the three different matrices were r 2 blood � 0.996 (n � 3), r 2 dialysate � 0.997 (n � 6), and r 2 saliva � 0.995 (n � 3). Reinjection after 7 days showed no difference in accuracy. e sample volume needed was 25 µL. e sample preparation time was about 2 min per sample (6 hours for 182 samples). e run time for one LC-MS/MS analyses was 7 min.

Comparison of the LC-MS/MS with the GC-MS
Method. After validation of the LC-MS/MS method, it was compared with the referred GC-MS method [8]. Table 2 shows the comparative results of the two methods for their analytical performance, appropriateness to analyze human biological samples, including corresponding extraction and cleanup procedures, and its validation according to ISO 17025/FDA Guidance for Industry (Table 2). e selectivity and the sensitivity were met by both methods, and the recovery showed no differences ( Table 2). But the accuracy differed between the two methods. e GC-MS method showed a higher variation at Cal 1 (20%) than the LC-MS/MS method (Cal 1 � 3%). As expected, the biggest difference in terms of analytical performance between the two methods was observed by the LOQ (GC-MS � 50 ng/mL; LC-MS/MS � 10 ng/mL) and LOD (GC-MS � 15 ng/mL; LC-MS/MS � <1 ng/mL) ( Table 2).
Both methods showed linear regression coefficients (r 2 ) higher than 0.995 in all three different matrices for the PHT calibration curve. e calibration range of the LC-MS/MS

Comparison of the LC-MS/MS Method with a Formerly Established LC-MS/MS Method.
Recently, a LC-MS/MS method has been developed which measures PHT in blood plasma or serum [10]. For the measurement of PHT in blood, the newly validated LC-MS/MS method can hence also be compared with this recently published study. e two methods show some similarities such as an identical IS (100 µg/mL PHT-d10), similar sample volumes needed (25 µL versus 20 µL [10]), and a comparable retention time (2.8 min versus approximately 2.1 min [10]). Both methods showed linear regression coefficients (r 2 ) higher than 0.99 in the blood matrix.
e accuracy was similar as both studies showed deviations of <10%. With regard to the calibration range and the calibration solution, the two LC-MS/MS methods differ. While the LC-MS/MS method developed in this study showed a calibration range from 10 ng/mL to 2000 ng/mL, the calibration curve of the published LC-MS/MS method [10] ranged from 100 ng/mL to 4000 ng/mL. e calibration solution in the current study was the respective biological matrix (e.g., blood). In the published study [10], phosphate-buffered saline was used as the calibration solution.

Discussion
In this study, a LC-MS/MS method to measure PHT in different biological samples was successfully validated and compared with a similarly validated GC-MS method [8]. Overall, the LC-MS/MS method showed to be a more specific analytical method with a higher general performance (Table 2). e LC-MS/MS method needed less sample volume, less chemicals, and less analytical time and therefore resulted in less costs for the sample preparation.
Concerning the LOD, there was a huge difference between the two methods. e LOD of the LC-MS/MS method was 15 times better than the one of the GC-MS methods: the LOD of the LC-MS/MS method was <1 ng/mL compared to 15 ng/mL for the GC-MS method (increments by a factor of ten). Similarly, the difference in LOQ was 5 times lower in LC-MS/MS (10 ng/ml) compared to GC-MS (50 ng/mL). e LOQ for the LC-MS/MS could be set even lower than 10 ng/ml PHT (Cal 1). e FDA guidelines which claim a minimal reproducibility at the LOQ level of 20% were well below (deviation to target PHT amount: <8% in aCSF (n � 6), <4% in blood (n � 3), <9% in saliva (n � 3); accuracy: aCSF 103%, blood 101%, and saliva 106%). e LOQ of the GC-MS method and hence the lowest concentration level (Cal 1 at 50 ng/ml) of the calibration curve showed a deviation value of 19%. e LC-MS/MS method, in contrast, showed a value of only 3% deviation at the lowest Cal (10 ng/mL).
is difference is of high importance, as samples with even lower concentrations could be reliably analyzed. e calibration range (from 10 ng/mL to 2000 ng/mL) of the LC-MS/MS method was twice as large as of the GC-MS method (50 ng/mL to 1200 ng/mL). is indicates that the LC-MS/MS method is more powerful and effective over a larger range of concentration, since the linearity is given over a larger area (10 ng/mL-2000 ng/mL) compared to the GC-MS method (50 ng/mL-1200 ng/mL).
As IS, two different substances were used. MPPH as a structurally related compound was used for the GC-MS method. As IS for LC-MS/MS, deuterated PHT (PHT-D 10 ) was used, which is the same molecule as PHT and differs only by the molecular mass (+1). All the physicochemical processes upon cleanup and analysis are identical or highly  similar for PHT and PHT-D 10 . MMPH, however, could be chemically affected in a different way than PHT, which could lead to a systematic bias in a given situation [15].
Regarding the sample preparation procedure, the LC-MS/MS (Figure 1) showed an important advantage compared to the GC-MS method as it only needs 3 steps of sample preparation compared to 11 steps necessary for the GC-MS method including a solid-phase extraction (SPE) and derivatization with a more critical chemical trimethylsulfonium (TMSH) (Figure 4). is resulted in significant shortening of the overall analysis: Preparation of the samples before injection for GC-MS is about ten times more time consuming than for the LC-MS/MS. For the GC-MS method, researchers needed 5 hours to prepare 25 samples (5 samples/h), whereas for the LC-MS/MS method 182 samples were prepared in 6 hours (30.3 samples/h), which corresponds to 6 times the amount of prepared samples per hour compared to the GC-MS method.
From the exposure side, the volumes are much larger and the exposure to the chemicals are more prolonged with the GC-MS method compared to the sample cleanup for the LC-MS/MS method. Especially, the derivatization agent TMSH is critical to handle because of toxicity. e risk of serious and even irreversible effects through inhalation, skin contact, or eye exposure is well known. TMSH is also considered to be teratogenic. erefore, the potential health risk for the laboratory staff handling the samples can be reduced by the LC-MS/MS method and the elimination of a safety critical agent. e amount of biological samples needed for the GC-MS method (50 µL) was twice as much as for the LC-MS/MS (25 µL). e sample volume is a critical point for PK/PD studies, where, for example, by continuing dialysis from brain in neurosurgical patients only small volumes of samples per time point/period are available. For 50 µL dialysate about 25 minutes collecting time is necessary at the usual flow rate of ∼2 µL per minute [12,16]. erefore, not a requested specific time point, but a rather large time segment is represented which can influence the requested results. e reduced sample volume needed (25 µL) for the LC-MS/MS analyses reduces the dialyses time needed per sample to about 15 minutes. e smaller the dialysis time, the more precise correlations of the respective tissue concentration with plasma/blood samples can be made.
Furthermore, LC-MS/MS also has the shorter run time. e time needed for 100 GC-MS analyses would be approximately 50 hours. e LC-MS/MS method, in contrast, needs only 11 hours and 40 minutes for 100 analyses. is is a time saving of more than 38 hours. While this may not be highly relevant for forensic purposes, for bedside and routine analyses (real-time) and PK/PD studies with larger numbers of samples, this factor is relevant. Also, when the time between taking a sample and the result needed is short, as it is in TDM to adjust subsequent dosing for PHT treatment, this time saving is crucial. e costs for one way materials per sample was about 50% lower for the LC-MS/MS compared to the GC-MS method. Especially because no SPE device was needed. Also, the reduced work load for the laboratory technician must be considered as an imported cost factor.
Finally, the appropriateness of the method also depends on the biological matrix. Both methods can generally be used to measure PHT in blood and saliva, as the sample volume is less limiting. As mentioned before, however, for dialysates, the most difficult aspect is to get enough sample volume. erefore, the LC-MS/MS method needing only half of the sample volume compared to the GC-MS method is more suited for microdialysate measurements. With respect to the LOD/LOQ, the LC-MS/MS method is also better suited for PK/PD studies, as it allows to include patients with low PHT dosages.
In addition, the newly established LC-MS/MS method was compared with a recently published LC-MS/MS method [10]. While this study measured PHT only in one body compartment (i.e., blood plasma or serum), the current LC-MS/MS method was developed and validated for the measurement of PHT in different body compartments (i.e., blood, saliva, and samples from brain tissue microdialysis often used in neurology and neurosurgery). e calibration range of the published LC-MS/MS method [10] (from 100 ng/mL to 4000 ng/mL) is appropriate for the measurement of PHT in blood plasma. As the PHT concentrations in brain tissue dialysates are much smaller than in blood plasma, the LC-MS/MS method of the current study was more appropriate for such samples, showing a lower calibration range from 10 ng/mL to 2000 ng/mL. Finally, as the aim of this study was to measure PHT in different biological matrices, a general substitute solution for blood plasma such as phosphatebuffered saline [10] could not be used. Instead, the fluid of the respective body compartment was used as calibration solution (e.g., artificial cerebrospinal fluid (aCSF) for the measurement of PHT in the brain tissue dialysates). is also eliminates a potential analytical bias due to matrix effects.

Conclusion
In this study, a LC-MS/MS method to measure PHT in different biological samples (i.e., human brain dialysate, blood, and saliva) was developed and validated under circumstances that ensured a high comparability with an established GC-MS method [8]. Overall, the study concludes that LC-MS/MS is not only better performing in human PHT concentration measuring or comparable drug PK/PD studies but is the only one to be used for bedside analysis.
e time-consuming sample preparation and the long run time of the GC-MS method delay the result, which is critical in TDM. e higher sensitivity, the smaller needed sample volume, the better LOD/LOQ, the less time-consuming cleaningup and sample preparation procedure, and the shorter run time make the LC-MS/MS method the preferred analytical procedure.

Conflicts of Interest
e authors declare that they have no conflicts of interest.