Isotope dilution-liquid chromatography-tandem mass spectrometry-based candidate reference measurement procedures for the quantification of total and free phenytoin in human serum and plasma

Chemicals and reagents

LC-MS-grade methanol (CAS No. 67-56-1) was purchased from Biosolve (Valkenswaard, The Netherlands). Dimethyl sulfoxide (DMSO) (CAS No. 67-68-5^®, American Chemical Society (ACS) reagent, ≥99.99 %) and acetic acid (CAS No. 64-19-7) were purchased from Sigma Aldrich (Taufkirchen, Germany). Phenytoin (CAS No. 57-41-0, Cat. No. PHR1139, Lot No. LRAB7779) was purchased from Supelco (Bellefonte, PA, USA). The deuterated ISTD [²H₁₀]-phenytoin (CAS No. 65854-97-9, Cat. No. C2090, Lot No. LSG-ALS-13-030-B1) was purchased from Alsachim (Illkirch Graffenstaden, France). HPLC-grade isopropanol (CAS No. 67-63-0) was purchased from Riedel de Haën (Seelze, Germany). Analyte-free human serum (Cat. No. 637816) was obtained from Merck (Darmstadt, Germany); TDM-free serum used as surrogate matrix (ID. No. 12095432001) was obtained from Roche Diagnostics GmbH (Mannheim, Germany); native plasma matrix (lithium heparin [Li-heparin]), dipotassium (K₂) ethylene diamine tetraacetic acid (EDTA), and tripotassium (K₃) EDTA were obtained from anonymized leftover patient samples (n≥5), and water was purified in-house using a Millipore Milli-Q^® 3 UV system from Merck (Darmstadt, Germany).

NMR solvent, pyridine-d₅, and the qNMR ISTD methyl-3,5-dinitrobenzoate (Cat. No. 94681, Lot No. BCBW3152) were obtained from Sigma Aldrich (Taufkirchen, Germany). Five-millimeter, 10-inch NMR tubes were purchased from Sigma Aldrich and/or Euroisotop GmbH (Hadfield, United Kingdom).

General requirements for laboratory equipment

Certified and calibrated equipment was used. The minimum sample weight for the microbalance used (XPR2, Mettler Toledo, Columbus, OH, USA) was determined according to United States Pharmacopeial Convention (USP) guidelines (USP Chapters 41 and 1251). Positive displacement pipettes were used to measure organic solvents and serum. Volumetric glassware (Class A volumetric flasks) that fulfilled the criteria of ISO 1042 and USP were used to prepare stock and spike solutions.

qNMR for determination of the purity of the standard materials

qNMR experiments were performed on a Jeol 600 MHz NMR (Jeol Ltd, Tokyo, Japan) equipped with a helium-cryoprobe head. A single-pulse ¹H{¹³C}NMR was utilized for quantitation (aromatic ortho protons δ=7.85 ppm; 4H; methyl-3,5-dinitrobenzoate as SI traceable qNMR ISTD; pyridine-d5 as solvent) with an inter-scan delay of 70 s. Pyridine-d₅ was utilized as solvent owing to its aromatic solvent-induced separation effects, which lead to optimal dispersion of the aromatic proton resonance. Any other solvent would prove incapable of this kind of resolution of aromatic protons. We also used two-dimensional total correlation spectroscopy to allocate the undoubtedly assignment of the quantitative resonance. Further details about NMR acquisition and processing parameters are available in the Supplementary Material 3 (Table S1, Figures S1–S3).

Preparation of calibrators and quality control samples

Two calibrator stock solutions were prepared by weighing phenytoin in tin boats on a microbalance. To obtain two 10 mg/mL stock solutions for total phenytoin, 50.0 mg of phenytoin was weighed twice and dissolved in 5 mL DMSO. For free phenytoin, 10 mg of phenytoin was weighed and dissolved in 10 mL DMSO in volumetric flasks to produce a 1.00 mg/mL stock solution. The final concentration of the stock solutions was calculated based on the purity of the material (99.7±0.4 %, determined by qNMR [k=2]) and the amount weighed. Each stock solution was then used to prepare four working solutions. Afterwards, the stock and working solutions were further diluted with DMSO in 5 mL volumetric flasks to produce eight levels of spike solutions for each total and free phenytoin. Finally, calibrator solutions (‘calibrators’) for total phenytoin were prepared by a 1 + 99 dilution (v + v) of the spike solution with analyte-free human serum, and calibrators for free phenytoin were prepared using a 1 + 99 dilution (v + v) of the spike solution with 20 % DMSO (v + v). The resulting calibration ranges for total and free phenytoin were 0.640–48.0 μg/mL and 0.0800–4.80 μg/mL, respectively.

Furthermore, four levels of matrix-based quality control (QC) samples were prepared using additional independent stock solutions for total and free phenytoin. The concentrations of QC sample levels were set at four critical control points: above the limit of quantification; below the therapeutic range; within the therapeutic reference range; and at the laboratory alert level. The resulting QC levels for total and free phenytoin had concentrations of 1.50, 6.00, 15.0, and 25.0 μg/mL, and 0.120, 0.400, 1.60, and 3.60 μg/mL, respectively. To prepare the QC levels, phenytoin was added to analyte-free human serum; to prepare the free phenytoin QC levels, phenytoin was spiked in analyte-free human serum ultrafiltrate. The ultrafiltrate was obtained by ultrafiltration of analyte-free human serum using an Amicon^® filter (Cat. No. UFC903024; Merck, Darmstadt, Germany). The filtration process involved centrifugation at 1,000 rcf and 25 °C for 180 min.

Internal standard solution

[²H₁₀]-phenytoin was dissolved in an appropriate amount of DMSO to obtain a 1 mg/mL stock solution, which was stored at −20 °C until further use. ISTD working solutions were freshly prepared on each day of sample preparation by 2-fold dilution of the ISTD stock solution, first with DMSO, then with Milli-Q^®-water to produce concentrations of 2.00 and 0.06 μg/mL for total and free phenytoin, respectively.

Sample preparation: total phenytoin assay

In a 2 mL tube (Eppendorf, Hamburg, Germany), 100 μL of ISTD working solution was pipetted, followed by the addition of 50 μL of the sample specimen (native sample/calibrator/QC). To precipitate proteins, 1,000 μL of 75 % methanol in Milli-Q^® water (v + v) was added. After centrifugation, 10 μL of the supernatant was diluted in two steps. In step 1, it was diluted at a ratio of 1 + 99 (v + v) using mobile phase A. In step 2, it was further diluted at a ratio of 1 + 7 (v + v) using mobile phase A.

Sample preparation: free phenytoin assay

Five hundred μL of native serum sample was pipetted into an ultrafiltration device (Merck Millipore Centrifree^® Ultracel^® PL Regenerated Cellulose, Art. No. 21986643122), which was then subjected to centrifugation at 1,000 rcf and 25 °C for a period of 30 min. In a separate 2 mL tube, 100 μL of ISTD working solution was mixed with 50 μL of the filtered sample (calibrator/QC/sample ultrafiltrate) and 1000 μL of 10 % methanol in Milli-Q^®-water (v + v) with 0.1 % acetic acid (mobile phase A). Finally, the solution was further diluted with mobile phase A at a ratio of 1 + 99 (v + v).

Liquid chromatography–mass spectrometry

Measurements for total and free phenytoin were performed on an Agilent 1290 Infinity II LC system (Santa Clara, CA, USA) equipped with a binary pump, a vacuum degasser, an autosampler at 7 °C, and a column compartment. Phenytoin was separated using an Agilent Zorbax Eclipse XDB-C8 column (100 × 3 mm, 3.5 µm, Santa Clara, CA, USA), which was kept at 40 °C in the column oven compartment. The eluents consisted of 10 % methanol in Milli-Q^®-water (v+) + 0.1 % acetic acid (mobile phase A) and 95 % methanol in Milli-Q^®-water (v + v) (mobile phase B). Measurements were performed with a flow rate of 0.6 mL/min, using a gradient over 10 min. A portion of the sample solution (5 μL) was injected, and a diverter valve used to reduce contamination of the MS, switching the eluent flow into waste from 0.0 to 0.5 min and from 6.0 to 10.0 min.

Phenytoin was detected in multiple-reaction monitoring mode using an AB Sciex Triple Quad 6500+ or Q-Trap 6500+ mass spectrometer operating in negative electrospray ionization mode. An ion spray voltage of −3,500 V and a temperature of 600 °C resulted in the most abundant signals. Nitrogen gas was used as curtain gas, collision gas, ion gas source 1, and ion gas source 2, set at 35 psi, 10 psi, 60 psi, and 50 psi, respectively. A collision cell entrance potential of −10 V, a collision exit potential of −7 V, a declustering potential of −60 V, and a dwell time of 50 ms was applied for all mass transitions.

The quantifier transition (phenytoin m/z 251.0 to 102.0) served as the basis of the quantitative method and was associated with a correspondent transition of the ISTD [²H₁₀]-phenytoin (m/z 261.0 to 106.0). The additional qualifier transition (phenytoin m/z 251.0 to 208.1 and [²H₁₀]-phenytoin m/z 261.0 to 218.1) allowed to check for possible interferences in the clinical samples. The quantifier/qualifier ratios of neat system suitability test (SST) samples were compared to the quantifier/qualifier ratio of a clinical sample. If they did not deviate by more than ±20 %, interferences were excluded.

System suitability test

An SST was performed prior to each analysis to check the sensitivity of the system, chromatographic performance, and possible carry-over effects. Therefore, a phenytoin stock solution (1 mg/mL in DMSO) was freshly diluted with mobile phase A to produce two samples (1 and 2), with phenytoin concentrations corresponding to the processed calibrator levels 1 and 8. The analytical data from experiments were only used if the SST was passed. Pass criteria for the SST were signal intensity, retention time and carry over. The SST signal-to-noise ratio for the quantifier transition of Sample 1 had to be ≥30. The retention times of SST samples 1 and 2 had to be within 4.0 min±0.5 min. Carry-over effects were examined by injecting SST sample 2 followed by two blank injections. The analyte peak area in the following blank injection had to be ≤20 % of the analyte peak area of SST Sample 1. This carry-over criterion also applied to all further blanks in the measurement campaign.

Calibration and structure of analytical series and data processing

Data processing was performed using AB Sciex Analyst software (version 1.6.3 or higher) with the IntelliQuant algorithm. Phenytoin and [²H₁₀]-phenytoin peaks were integrated within a retention time window of 4.0 min±0.5 min using a smoothing factor of 3, a noise percent of 90 %, and a base sub-window of 0.50 min. Eight-point calibration was used to measure the calibrator levels in increasing concentration at the beginning and end of the analytical series. Both calibrator injections were used to generate the final calibration function. The calibration function was obtained by linear regression of the area ratios of the analyte and ISTD (y) against the analyte concentration (x), resulting in the function: y=x × A + b.

Selectivity/specificity

Selectivity was determined by spiking phenytoin and [²H₁₀]-phenytoin into analyte-free human serum, TDM-free serum, and native Li-heparin plasma at a concentration of 4.00 μg/mL (total phenytoin) and 0.600 μg/mL (free phenytoin). To examine possible interfering matrix signals for the analyte quantifier and qualifier transitions, analyte-free matrices were checked at the expected retention time. In addition, analyte-free matrices were spiked with ISTD to evaluate whether any residual unlabeled analyte remained in the ISTD. The amount of unlabeled analyte in the ISTD was not supposed to exceed 20 % of the lower limit of the measuring interval (LLMI), which corresponded to the lowest calibrator level.

Matrix effects

To assess matrix effects (ME), a qualitative post-column infusion experiment was performed by infusing a solution of 25 ng/mL phenytoin in mobile phase A/mobile phase B 1 + 1 (v + v), at a flow rate of 7 μL/min, via a T-piece into the HPLC column effluent. Thereafter, processed matrix samples (analyte-free human serum, TDM-free serum, and native plasma matrices [Li-heparin, K₂-EDTA. and K₃-EDTA]) were injected. Any decrease or increase of the MS output at the expected retention time would indicate a matrix component-mediated effect on the ionization.

Furthermore, calibration curves were compared in terms of slopes and coefficients of determination R² to exclude MEs. For total phenytoin, calibration curves were freshly prepared in mobile phase A as phenytoin is not well soluble in water, analyte-free human serum, TDM-free serum, or native Li-heparin plasma. For free phenytoin, calibration curves were compared in 20 % DMSO in Milli-Q^® water (v + v), analyte-free human serum ultrafiltrate, TDM-free serum ultrafiltrate, and native Li-heparin plasma ultrafiltrate [48].

To exclude MEs, mean slopes of calibrator sample sets (n=2 preparations of each set) in different matrices were required to be comparable and 95 % confidence intervals (CIs) should overlap. Furthermore, the coefficient of determination for the individual tested slopes needed to exceed 0.999. In addition to these slope experiments, matrix-based calibrator samples were measured against calibrator sets prepared in mobile phase A and in 20 % DMSO solution. Recoveries were reported as the percentage of recovery of the measured concentration relative to the nominal concentration.

A comparison of absolute peak areas for analyte and ISTD was performed by spiking analyte and ISTD solution in the four above-mentioned matrices after protein precipitation/ultrafiltration for three levels spread over the working range (6.00, 20.0, 40.0 μg/mL, and 0.400, 1.60, and 3.60 μg/mL for total and free phenytoin, respectively) [49]. Ion enhancement and suppression were evaluated by comparing peak areas of analyte, ISTD, and area ratios of matrix samples against neat samples. Samples were prepared in five replicates. Percentage deviation was required to be within the range of 100±10 % and balanced by the ISTD.

Linearity

In addition to the calibrator samples, four spiked samples were prepared to expand the calibration ranges by ±20 %, with final concentration ranges from 0.500–60.0 μg/mL and 0.0640–6.00 μg/mL for total and free phenytoin, respectively. Coefficients of determination and residuals were determined and required to be R²≥0.999 and randomly distributed.

Furthermore, the linearity of the method was proven based on the recovery of nine serially diluted samples. To produce these samples, calibrator level 1 was mixed with level 8 with the following shares: 9 + 1 (v + v), 8 + 2 (v + v), 7 + 3 (v + v), 6 + 4 (v + v), 5 + 5 (v + v), 4 + 6 (v + v), 3 + 7 (v + v), 2 + 8 (v + v), and 1 + 9 (v + v). These sample pools were required to show linear dependency, with a coefficient of determination of ≥0.999. Recoveries were reported as the percentage of recovery for each measured concentration relative to the nominal concentration of the sample pools.

Lower limit of measuring interval and limit of detection

To demonstrate accuracy, trueness, and precision at the LLMI, spiked, analyte-free human serum samples and samples spiked in 20 % DMSO were prepared with a concentration corresponding to the lowest calibrator levels for total phenytoin (0.640 μg/mL) and free phenytoin (0.0800 μg/mL), respectively. These samples were prepared 5-fold to ensure reliable results.

In addition, the limit of detection (LOD) was estimated using the approach described by Armbruster et al. [50]. To do this the limit of blank (LOB) was calculated using 10 independent matrix blank samples as follows: LOB=mean_blank + 1.645(SD_blank). The LOD was then estimated using 10 replicates of calibrator level 1, used as the low concentration sample: LOD=LOB + 1.645(SD_{low concentration sample}).

Accuracy, trueness, and precision

A 5-day validation experiment, previously described in Taibon et al. [51], was performed to evaluate the precision and accuracy of the developed methods. Total method variability was estimated using an analysis of variance (ANOVA)-based variance component analysis, including variability components, such as between-injection variability, between-preparation variability, between-calibration variability, and between-day variability.

On each day, four spiked analyte-free human serum samples, covering the measuring range (1.50, 10.0, 20.0, and 40.0 μg/mL for total phenytoin and 1.20, 4.00, 16.0, and 24.0 μg/mL for free phenytoin), assuming a free analyte fraction of approximately 10 % in the matrix, were prepared, as well as two native patient serum samples, at concentrations close to the medical decision point. Additionally, Li-heparin plasma samples were spiked with the same concentration ranges for total phenytoin.

For both individual measurement parts (A and B), all samples were prepared in triplicate and injected twice, resulting in a total of 12 measurements per day and 60 measurements over the course of 5 days. This validation scheme was carried out in parallel by two operators, with one person responsible for part A and another responsible for part B. For each part, independent calibration curves were generated and used for quantitative analysis. Data were evaluated using Biowarp, an internal statistic program based on the VCA Roche Open-Source software package in R [52].

Trueness and accuracy for total phenytoin were assessed using four spiked analyte-free human serum and native Li-heparin plasma samples with concentrations of 1.50, 10.0, 20.0, and 40.0 μg/mL. Dilution integrity was assessed using two spiked analyte-free human serum samples at concentration levels of 60.0 μg/mL and 100 μg/mL.

For free phenytoin, four concentration levels in analyte-free human serum, analyte-free human serum ultrafiltrate, and Li-heparin plasma ultrafiltrate were prepared. To create these, spiked, analyte-free human serum samples were prepared at four concentration levels (1.20, 4.00, 16.0, and 24.0 μg/mL), assuming a free analyte fraction of approximately 10 % in the matrix. The nominal concentration was determined (n=3 preparations and two injections) by independent measurement of the free and bound fractions of phenytoin. In addition, analyte-free human serum and Li-heparin ultrafiltrate samples were prepared at the following concentrations: 0.168, 0.560, 2.24, and 3.36 μg/mL. To show dilution integrity, a concentrated, spiked, analyte-free human serum sample (56.0 μg/mL) was filtered and then diluted with analyte-free human serum ultrafiltrate. The dilution was required to show a linear dependency (coefficient of determination of 0.998).

All samples were prepared in triplicate for each of part A and part B (n=6 measurements) on one day. Accuracy was determined as the percentage recovery of the measured concentration relative to the spiked concentration, while trueness was reported as the percentage recovery of the mean measured concentration relative to the spiked concentration.

Sample stability

The stability of the processed samples in the autosampler was evaluated at 7 °C after 7 days. Recoveries were calculated by comparing the measured value with freshly prepared samples.

Stability of calibrator and control material stored at −20 °C was evaluated for a period of four weeks for the total phenytoin assay (analyte-free, human serum calibrators ranged from 0.640–48.0 μg/mL, and analyte-free, human serum QC levels from 1.50–25.0 μg/mL). Stability at −20 °C for calibrator and QC material for the free phenytoin assay was tested over a period of 65 days (calibrators in 20 % DMSO ranged from 0.0800–4.80 μg/mL, and QC levels prepared in analyte-free, human serum ultrafiltrate ranged from 0.120–3.60 μg/mL).

Recoveries were calculated by comparing the measured values for stored samples with those from with freshly prepared samples. The total error (TE) was used as the acceptance criterion, set to ±5 %. Stability was guaranteed for x–1 day for a measurement interval of 2–28 days (x) and for y–1 week for a measurement interval of >4 weeks (y).

Equivalence of results between independent laboratories

A method comparison study was undertaken to assess agreement of the candidate RMP between two independent laboratories (laboratory 1 Dr. Risch, Buchs SG, and laboratory 2 Roche Diagnostic GmbH, Penzberg). For total phenytoin this study included 146 samples (65 native serum patient samples, 41 native plasma samples, 10 patient pool samples, and 30 spiked samples) for total phenytoin. For free phenytoin, a total of 85 samples were measured, of which 55 (28 native serum patient samples, 17 native plasma samples, and 10 patient pool samples) were the same as for total phenytoin, and 30 were additional spiked serum samples. All patient samples were collected in accordance with the Declaration of Helsinki.

Additionally, a 3-day precision experiment was performed at laboratory 2 using the same spiked, analyte-free, human serum samples as in laboratory 1. The LC-MS system and laboratory equipment used were similar in both laboratories. For gravimetrical procedures, Site 2 utilized an ultra-microbalance XP6U/M (Mettler Toledo) and aluminum weighing boats. Calibrators were prepared independently in both laboratories as described in Supplementary Materials 1 and 2.

Uncertainty of measurements

MU was determined according to the GUM [47] and the approach adopted by Taibon et al. [51] where the following parameters were considered: purity of the reference material (based on the certificate); weighing of the analyte; preparation of stock, working, spike and calibrator solutions; preparation of the ISTD solution; sample preparation of the calibrators; measurement of the calibrators; generation of the calibration curve; preparation and measurement of unknown samples; and evaluation of the sample results. To assess MU in calibrator preparation, a type B evaluation was performed, while all other aspects were evaluated as type A. Total MU was then estimated by combining type A and B uncertainties. Further information on the estimation of type B uncertainty is provided in Supplementary Material 4.

To meet RMP requirements, the target uncertainty for phenytoin was set at <3.2 % (k=2). For free phenytoin, an RMP target uncertainty of <9.1 % (k=2) was determined.

Traceability to SI units

Traceability to the SI unit of mass (kilogram), the most important parameter for a reference measurement method, was established by the utilization of methyl-3,5-dinitrobenzoate as the qNMR ISTD, which is directly traceable to NIST PS1 (primary qNMR standard). Furthermore, traceability to the SI unit of amount of substance (mole) was also established via NIST PS1. Since Planck’s constant is currently considered the main parameter for defining a kilogram, as is Avogadro’s constant for a mole, qNMR methodology is well suited to fulfill both traceability chains. Additionally, as per the latest IUPAC Technical Report, qNMR represents a potential primary RMP, ideally suited to the characterization of primary reference materials [46]. Six individual experiments (detailed in Figure S2 and Table S1 in Supplementary Material 3), involving six individual weighings, yielded a final absolute content value of 99.7±0.4 % (k=2).

Selectivity/specificity

The developed gradient, combined with a reversed phase column (Agilent Zorbax Eclipse XDB-C8), allowed the selective and specific measurement of phenytoin (see Figures 1A–C and 2A–C). Specificity was determined by analyzing analyte-free human serum, TDM-free serum, and native Li-heparin plasma. No signals were observed at the expected retention time (4.0 min). Moreover, no residual analyte was found in the ISTD-spiked matrix samples. Furthermore, samples measured within the method comparison study showed no interferences in the retention time window for phenytoin.

Figure 1:

Total phenytoin LC-MS/MS derived analytical readouts. Analyte on the left-hand side, ISTD on the right-hand side. (A) Chromatogram of a matrix blank. (B) The lowest calibrator level peak (0.640 μg/mL) spiked in analyte-free human serum. (C) Patient pool (n≥5, 1.44 μg/mL). ISTD, internal standard; LC-MS/MS, liquid chromatography-tandem mass spectrometry.

Figure 2:

Free phenytoin LC-MS/MS derived analytical readouts. Analyte on the left-hand side, ISTD on the right-hand side. (A) Chromatogram of a matrix blank. (B) The lowest calibrator level peak (0.0800 μg/mL) spiked in 20 % DMSO (v+v). (C) Patient pool (n≥5, 2.82 μg/mL). ISTD, internal standard; LC-MS/MS, liquid chromatography-tandem mass spectrometry.

Matrix effects

Matrix-dependent effects were eliminated by the use of a sample preparation protocol involving a substantial dilution after protein precipitation. This was shown by the post-column infusion experiment, where no change in the ionization field was observed at the expected retention times in analyte-free human serum, TDM-free serum and plasma matrices (K₂-EDTA plasma, K₃-EDTA plasma, and Li-heparin plasma).

Experiments were performed to compare slopes in different concentration ranges for total and free phenytoin. Mobile phase A was selected as a matrix-free comparator matrix for the concentration range of total phenytoin. For the concentration range of free phenytoin, a solution of 20 % DMSO in water was used, which was also the matrix used to produce calibrators for the assay.

The mean slopes for total phenytoin were found to be 0.31 (95 % CI: 0.30–0.31) in analyte-free human serum, 0.30 (95 % CI: 0.30–0.31) in TDM-free serum, 0.30 (95 % CI: 0.30–0.31) in native plasma, and 0.30 (95 % CI: 0.29–0.30) in mobile phase A. All experiments showed least square regression coefficients of determination of ≥0.999. Using the mobile phase calibrator set as a standard and all other samples as independent controls (n=48), the mean recovery was found to be 102 % with a standard deviation of 2.7 % and 95 % CI: 101–103 %.

The mean slopes for free phenytoin were found to be 0.93 (95 % CI: 0.91–0.95) in analyte-free human serum ultrafiltrate, 0.92 (95 % CI: 0.92–0.93) in TDM-free serum ultrafiltrate, 0.93 (95 % CI: 0.93–0.93) in native Li-heparin plasma, and 0.94 (95 % CI: 0.93–0.96) in 20 % DMSO. All experiments showed least square regression coefficients of determination of ≥0.999. Using the 20 % DMSO phase calibrator set as a standard and all other samples as independent controls (n=48), the mean recovery was found to be 99 % with a standard deviation of 2.7 % and a 95 % CI: 98–100 %.

Furthermore, analyte and ISTD peak areas and area ratios from spiked matrix samples were compared to spiked neat samples. Analyte peak areas for total phenytoin ranged from 98–103 %, ISTD peak areas from 96–104 %, and area ratios from 99–102 % (Table 1). Analyte peak areas for free phenytoin ranged from 99–105 %, ISTD peak areas from 98–105 %, and area ratios from 99–102 % (Table 2).

Linearity

Linearity of the method was proven by analyzing calibration curves covering ±20 % of the measuring range (n=6, sample preparations). Residuals were randomly distributed in a linear regression model for total and free phenytoin. Coefficients of determination were ≥0.999 for all individual calibrations for total and free phenytoin. Applying the preferred linear regression model, samples 1–11 were evaluated and showed a linear dependence with a coefficient of determination of 0.999 for total and free phenytoin. The relative deviation ranged from −1.0 to 3.3 % and −1.6 to 2.5 %, and the coefficient of variation (CV) was determined to be ≤3.9 and ≤2.4 % for total and free phenytoin, respectively.

Lower limit of measuring interval and limit of detection

The LLMI was determined using spiked samples at the lowest calibrator levels (0.640 and 0.0800 μg/mL) for total and free phenytoin. Relative bias showed deviations of −1.3 and 2.7 %, and CVs of 2.7 and 0.3 %, for total and free phenytoin, respectively. LODs were estimated to be 0.217 and 0.0237 μg/mL for total and free phenytoin, respectively.

Accuracy, trueness, and precision

The total method variability was estimated using an ANOVA-based variance component analysis (VCA). Repeatability CVs, including variability components, such as between-injection variability and between-preparation variability, ranged from 1.7–3.8 % for total phenytoin over all concentration levels, and were independent of the matrix (Table 3). Comparable repeatability was shown for free phenytoin with values ranging from 1.4–3.7 % (Table 4). Intermediate-precision CVs, including additional variability components, such as between-calibration variability and between-day variability, were found to be <3.8 and <3.7 % for total and free phenytoin, respectively (Tables 3 and 4).

Accuracy for total phenytoin was assessed using four spiked analyte-free human serum and Li-heparin plasma samples, including three preparations of each operator (in total, n=6 preparations). The relative mean bias for total phenytoin ranged from −2.7–0.3 % for analyte-free human serum samples and 0.0–1.1 % for Li-heparin plasma samples (Table 5); high concentration samples showed a mean bias of −0.6 and 0.3 % for total phenytoin (Table 5).

To determine the amount of free phenytoin in serum samples, free and bound fractions were determined. The following concentrations of serum sample were used to determine the trueness of the free phenytoin assay: 0.162±0.005, 0.542±0.002, 2.31±0.02, and 3.59±0.05 μg/mL. Biases for free phenytoin ranged from −3.5 to −2.7 % for spiked analyte-free human serum samples, from 2.5–4.1 % for spiked analyte-free human serum ultrafiltrate samples, and from 2.8–4.0 % for spiked Li-heparin plasma ultrafiltrate samples (Tables 6 and 7). For free phenytoin, the highest concentration sample showed a linear dependence (r²=0.998) after serial dilution, and a CV of less than 2.3 %.

Stability

Processed samples were found to be stable at 7 °C for 6 days, with mean recoveries of 103 and 98 % for total and free phenytoin, respectively. The stability of spiked, frozen analyte-free human serum samples stored at −20 °C was verified for 28 days for total phenytoin, with a mean recovery of 100 %. The stability of phenytoin at −20 °C in 20 % DMSO and in analyte-free human serum ultrafiltrate was verified for 64 days, with a mean recovery of 101 %.

Equivalence of results between independent laboratories

A method comparison study for total phenytoin was conducted using 146 anonymized patient samples, comprising 65 serum, 41 plasma, 30 spiked, and 10 pooled samples. Seven samples fell below the LLMI and were excluded from the analysis. In the study of free phenytoin, a total of 85 samples were evaluated, including 28 serum, 17 plasma, 30 spiked, and 10 pooled samples. Twenty-one samples exceeded the ULMI, three were below the LLMI, and two samples were missing from Site 1. Consequently, these samples were excluded from the evaluation.

Passing-Bablok regression analysis for total phenytoin yielded a slope of 1.00 (95 % CI: 1.00–1.01) and an intercept of −0.01 (95 % CI: −0.09 to 0.03), with an strong Pearson correlation coefficient of 0.998. For free phenytoin, the analysis resulted in a slope of 1.06 (95 % CI: 1.04–1.07) and an intercept of 0.00 (95 % CI: −0.02 to 0.02), alongside an equally confirming Pearson correlation coefficient of 0.999 (Figures 3 and 4).

Figure 3:

Results from the patient sample-based total phenytoin method comparison study performed between two independent laboratories. (A) Passing–Bablok regression plot, including the Pearson regression analysis, for the method comparison study of the RMP (n=139 samples) between the independent laboratories (Site 1: Risch; Site 2: Roche). Passing–Bablok regression analysis resulted in a regression equation with a slope of 1.00 (95 % CI: 1.00–1.01) and an intercept of −0.01 (95 % CI: −0.09 to 0.03). The Pearson correlation value was ≥0.998. (B) Bland–Altman plot for the method comparison study of the RMP (n=139 samples) between two independent laboratories (Laboratory 1: Risch site, Laboratory 2: Roche site). The interlaboratory measurement bias was −0.2 % (95 % CI: −0.8 to 0.3), and the 2S interval of the relative difference was 6.5 % (lower limit 95 % CI: −7.7 to −5.8 %, upper limit 95 % CI: 5.3–7.2 %). CI, confidence interval; RMP, reference measurement procedure; SD, standard deviation.

Figure 4:

Results from the patient sample-based free phenytoin method comparison study performed between two independent laboratories. (A) Passing–Bablok regression plot, including the Pearson regression analysis for the method comparison study of the RMP (n=59 samples) between the independent laboratories (Site 1: Risch; Site 2: Roche). Passing–Bablok regression analysis resulted in a regression equation with a slope of 1.06 (95 % CI: 1.04–1.07) and an intercept of 0.00 (95 % CI: −0.02 to 0.02). The Pearson correlation value was ≥0.998. (B) Bland–Altman plot for the method comparison study of the RMP (n=59 samples) between two independent laboratories (Laboratory 1: Risch site, Laboratory 2: Roche site). The interlaboratory measurement bias was 5.1 % (95 % CI: 4.2–6.1 %) and the 2S interval of the relative difference was 6.8 % (lower limit 95 % CI: −3.3 to −0.1 %, upper limit 95 % CI: 10.4–13.5 %). CI, confidence interval; RMP, reference measurement procedure; SD, standard deviation.

Bland-Altman analysis revealed a mean bias of −0.2 % (95 % CI: −0.8 to 0.3 %) and 2S agreement of 6.5 % (lower limit 95 % CI: −7.7 to −5.8 %, upper limit 95 % CI: 5.3–7.2) for total phenytoin. For free phenytoin, the analysis demonstrated a mean bias of 5.1 % (95 % CI: 4.2–6.1 %) and 2S agreement of 6.8 % (lower limit 95 % CI: −3.3 to −0.1 %, upper limit 95 % CI: 10.4–13.5 %).

Furthermore, performance evaluation at Site 2 (n=36) confirmed intermediate precision performance with values of ≤2.4 % for total phenytoin and ≤3.2 % for free phenytoin.

Uncertainty of results

Calibration level uncertainties were evaluated as type B, while precision experiments for sample preparation and measurement were assessed as type A uncertainties. MU (k=2) for single serum sample measurements were 3.5–5.1 %, and 1.8–3.3 % for target value assignment (n=6) (Tables 8 and 9). For free phenytoin, uncertainties ranged from 4.1–7.9 % for single measurements and 2.0–2.7 % for target value assignment, meeting uncertainty requirements (Tables 10 and 11).