Talanta

Effective quantification of 11 tyrosine kinase inhibitors and caffeine in
human plasma by validated LC-MS/MS method with potent phospholipids
clean-up procedure. Application to therapeutic drug monitoring

ABSTRACT
Therapeutic drug monitoring (TDM) help to improve treatment efficacy and safety. Therefore, a simple and
sensitive liquid chromatography-tandem mass spectrometry (LC-MS/MS) method was developed and validated
for the simultaneous monitoring of 11 tyrosine kinase inhibitors (TKIs) in human plasma. TKIs included in the
assay are used in the treatment of chronic myeloid leukemia (CML: imatinib, dasatinib, nilotinib, bosutinib,
ponatinib), polycythemia vera (ruxolitinib), chronic lymphocytic leukemia (ibrutinib) and rheumatoid arthritis
(filgotinib, tofacitinib, baricitinib, peficitinib). Caffeine was also included in the method. Caffeine increases the
acidity of the stomach and decreases its pH as well as is a competitive inhibitor of cytochrome P450 isoenzymes.
Thus, it may influence absorption and metabolism of some TKIs, by modifying their plasma levels. The analytes
of interest and their stable isotope-labeled internal standards were extracted from 200 μL of human plasma.
Microelution-solid phase extraction (μ-SPE) was optimized for method validation and compared to simple
protein precipitation (PPT). A gradient elution on a Poroshell 120 EC-C18 column at 60 °C and a flow rate of
0.5 mL/min was applied for analyte separation. The analytical run lasted 8 min and it was followed by a re￾equilibration time of 4 min. Dynamic multiple reaction monitoring scan in the positive ionization mode was
applied to improve method sensitivity. Endogenous plasma phospholipids can strongly affect MS analysis.
Hence, the monitoring of endogenous phospholipids was included in the assay. Full validation of the method was
achieved, including tests of precision, accuracy, trueness, linearity, extraction recovery, matrix effect, process
efficiency, stability, sensitivity (with excellent LLOQs), selectivity, identity confirmation and carry-over effect.
Regarding sample cleanup, more than 91% of early eluting and more than 96% of late eluting endogenous
phospholipids were eliminated by μ-SPE when compared to PPT. This method enables the simultaneous plasma
monitoring of 11 TKIs and caffeine and ensures high effectiveness in phospholipids elimination. The present
approach is currently used in our clinical practice, being applied to TDM of dasatinib, imatinib, nilotinib and
ponatinib. TKIs plasma monitoring helps to individualize dose adjustment and manage adverse effects in CML
patients.
1. Introduction
The introduction of tyrosine kinase inhibitors (TKIs) in the clinical
practice revolutionized cancer treatment by converting some previously
deadly malignancies into chronic diseases [1]. Although the efficacy is
the most relevant aspect of TKI treatment, medication safety is also
taken into consideration as the TKI therapy is potentially lifelong and
multiple TKIs are available [2]. While the drugs are administrated or￾ally, complex steps in drug absorption and metabolism are responsible
for large inter- and intraindividual variability of TKI plasma levels.
Consequently, this can lead to serious adverse events. Moreover, poly￾morphisms in cytochrome P450 complex (CYP) and ATP-binding

https://doi.org/10.1016/j.talanta.2019.120450

Received 5 August 2019; Received in revised form 30 September 2019; Accepted 4 October 2019
cassette (ABC) drug transporter, food effect, treatment adherence, in￾teracting co-medication or environmental factors, play a crucial role in
final drug plasma concentrations in patients [3,4]. Therefore, due to the
high inter- and intraindividual variability of TKI levels and their ther￾apeutic and toxic effects, these drugs are candidates for therapeutic
drug monitoring (TDM) [5]. The clinical benefit of TKI TDM is high￾lighted in numerous studies in order to adjust the dose and decrease
adverse events for each patient [5–9].
Even though TKIs are mostly used for cancer treatment, they are
utilized for the treatment of other conditions, such as autoimmune
diseases. Therefore, we included TKIs used in the treatment of chronic
myeloid leukemia (CML): imatinib (IMA), dasatinib (DASA), nilotinib
(NILO), bosutinib (BOSU), ponatinib (PONA); polycythemia vera: rux￾olitinib (RUXOLI, specific janus type 1 and 2 kinase [JAK 1/2] in￾hibitor); chronic lymphocytic leukemia: ibrutinib (IBRU, Bruton’s tyr￾osine kinase inhibitor) and rheumatoid arthritis: filgotinib (FILGO,
specific JAK1 inhibitor), tofacitinib (TOFACI, specific JAK 1/3 in￾hibitor), baricitinib (BARICI, specific JAK1/2 inhibitor) and peficitinib
(PEFICI, JAK3 inhibitor) in the present assay.
TKIs are predominantly metabolized by liver CYP enzymes. The 3A4
isoform (CYP3A4) is the most important metabolic route for the ma￾jority of TKIs. The CYP1A2 isoenzyme is responsible for IMA and NILO
metabolism with a minor contribution [10]. Additionally, caffeine
(CAFF) is a weak inhibitor of the CYP1A2 enzyme, thus it may interfere
with TKI metabolism. CAFF maximum plasma concentration (Cmax) is
reached within 1–2 h [11], being close to the Cmax of DASA (1 h) and
IMA (2 h). Moreover, CAFF induces the production of gastric acid [12]
and therefore decreases the gastric pH. It is widely known that even
relatively small change in gastric pH can have a large impact on drug
absorption. As TKI absorption is more efficient at low pH, CAFF could
increase their absorption by diminishing the pH. Accordingly, CAFF
might influence the safety and efficacy of a prescribed medication and
therefore it was added to the present approach.
Numerous analytical methods monitoring TKIs in plasma are re￾ported in the literature [13–17]. The objective in a modern clinical
laboratory is to increase the number of drugs analyzed in a single
sample preparation process and single run being cost effective and time
saving. Although methods quantifying 14 to 17 TKIs simultaneously
exist [14,16], there is no liquid chromatography-tandem mass spec￾trometry (LC-MS/MS) method for simultaneous plasma monitoring of
FILGO, BARICI and PEFICI in humans. Currently, the most frequently
applied technology for drug plasma monitoring is LC-MS/MS. In spite of
many advantages of LC-MS/MS, matrix effect (ME) [18,19] caused by
matrix components may compromise method credibility. As ME could
cause less efficient extraction recovery (RE) and negatively affect the
overall process efficiency (PE, a combination of ME and RE) [20], it is
extensively studied in our laboratory during method development and
validation.
Endogenous blood plasma phospholipids, proteins, salts, other
blood plasma components and exogenous substances are known to alter
the ionization process and the chromatographic separation of analytes
[21,22]. Endogenous phosphatidylcholines (PC) [23] are present in
high (approximately 70%), while lysophosphatidylcholines (LPC) [24]
in low abundance (8%) in total blood plasma. Polar hydrophilic phos￾pholipid phosphate groups are known to elute at the beginning of the
chromatogram on the reversed-phase LC column, while nonpolar and
hydrophobic fatty acid chains elute later with a higher percentage of
organic solvent, both compromising target compound elution [21].
Therefore, an appropriate selection of sample cleanup method is of
great importance. To date, protein precipitation (PPT) [15–17], liquid￾liquid extraction (LLE) [25] or the combination of phospholipid re￾moval products with LLE [26] were used by several authors as TKI
sample preparation method. Nevertheless, PPT is most likely to cause
ion suppression in electrospray ionization (ESI), since it does not re￾move all endogenous compounds that interfere with ESI-LC-MS/MS
analysis. On the contrary, LLE is characterized by clean extracts,
nonetheless requires multiple extraction steps to increase analyte re￾covery [18]. Solid phase extraction (SPE) is less time consuming and
requires less solvent volume. Moreover, microelution-SPE (μ-SPE) offers
analyte enrichment and allows smaller sample quantity. For the above
mentioned reasons, we chose SPE as sample cleanup procedure while
comparing it to simple PPT extraction.
According to our knowledge, none of the existing LC-MS/MS
methods for quantification of TKIs in human plasma included FILGO,
BRICI, PEFICI or CAFF in a single run nor compared the endogenous
phospholipids’ elimination efficacy of different extraction methods.
Therefore, the aim of the present study was to develop a sensitive LC￾MS/MS method for the simultaneous determination of 11 TKIs (PEFICI,
TOFACI, BARICI, IMA, FILGO, RUXOLI, DASA, BOSU, NILO, PONA and
IBRU) and CAFF. Moreover, the present approach was applied to TDM
of CML patients.
2. Material and methods
2.1. Chemicals and reagents
IMA mesylate, DASA, NILO, BOSU, PONA, IBRU, TOFA and BARICI
as well as stable isotope-labeled ISs (SIL-ISs) [2H8]-imatinib (IMA-D8),
[2H8]-dasatinib (DASA-D8), and [13C, 2H3]-nilotinib (NILO-13C-D3),
[2H9]-bosutinib (BOSU-D9), [2H8]-ponatinib (PONA-D8), [2H5]-ibru￾tinib (IBRU-D5) and [13C3, 15 N]-tofacitinib (TOFACI-13C3–15N) were
supplied by Alsachim (Illkirch-Graffenstaden, France). CAFF, PEFICI
and RUXOLI were purchased from MedChemExpress Europe
(Sollentuna, Sweden). Ultrapure water was adquired from a Milli-Q®
Water Purification System (Millipore-Ibérica, Madrid, Spain).
Acetonitrile (ACN), methanol (MeOH), orthophosphoric acid and am￾monia-hydroxide solution (HPLC grade) were obtained from SYMTA
(Madrid, Spain). Formic acid was purchased from Sigma-Aldrich
(Madrid, Spain). All chemicals were of analytical or LC-MS grade. For
the preparation of calibration and validation standards human plasma
samples were provided by the Transfusion Center of “Comunidad
Autónoma de Madrid” (Madrid, Spain).
2.2. Standard solutions
Stock solutions of calibrators (CALs) and quality controls (QCs) of
all analytes and SIL-ISs were prepared by dissolving an accurately
weighed quantity in MeOH to obtain a concentration of 1 mg/mL.
Working solutions of each analyte were prepared by independent di￾lutions from each stock solution at the following concentrations:
0.1 mg/mL, 0.01 mg/mL and 0.001 mg/mL. All solutions were pre￾served at −80 °C.
CAL concentrations were calculated based on the therapeutic ranges
of each drug. Eight CALs were prepared from independent dilutions of
each stock solution and spiked into blank plasma samples. Four QCs
(lower limit of quantification, LLOQ; low QC; medium QC and high QC)
were prepared in the same fashion. QC concentrations are shown in
Supplementary Table 1. According to the recommendations for bioa￾nalytical method validation of the US Food and Drug Administration
(FDA) [27], the European Medicines Agency (EMA) [28], and the In￾ternational Conference on Harmonisation (ICH) [29], a blank and a
zero (blank containing IS) plasma sample without adding the drugs
were included in the analysis. All CALs, QCs, and ISs were stored at
−80 °C prior to their use.
2.3. LC-MS/MS system
Samples were analyzed using an HPLC system composed of a 1200
Series separation module (Agilent Technologies, Madrid, Spain) cou￾pled to a triple quadrupole mass spectrometer (Agilent Technologies
6410B), equipped with electrospray ionization (ESI). The Agilent
MassHunter Workstation Data Acquisition software was used to control
D. Koller, et al.
2
the instrument. Chromatographic separation was performed using a
mobile phase composed of water (solvent A) and ACN (solvent D)
(82:12, v/v), each containing 0.1% formic acid with a flow rate of
0.5 mL/min on a Poroshell 120 ECC18 column (2.1 × 75 mm, 2.7 μm;
Agilent Technologies) held at 60 °C. The mobile phase pH was adjusted
with 25% ammonium hydroxide solution to reach the final pH of 2.0.
The following gradient conditions were applied to analyte separation:
0–0.9 min, 82% (A) and 18% (B); 0.9–3.0 min, gradually increasing
eluent B to 35%; 3.0–8.0 min, gradually increasing eluent B to 99%;
8.0–8.1 min gradually decreasing eluent B to 18%; followed by a re￾equilibration to initial solvent composition for 4.0 min (8.1–12.0 min).
This amounts to a total time of approximately 12.0 min between in￾jections. A volume of 5 μL was injected to the LC-MS/MS instrument
and the samples were kept at 20 °C prior to injection.
Analytes were quantified in the positive and dynamic multiple re￾action monitoring (dMRM) mode. The MS conditions were set as fol￾lows: nitrogen was used as nebulizing (40 psi) and drying gas (10 L/
min) at 350 °C; capillary voltage was set at 3.5 kV (kV). The MS colli￾sion gas was highly pure N2 (> 99.9995). Retention time (tR), frag￾mentor voltage (volt, V) and collision energy (electrovolt, eV) were
optimized under selected ion monitoring (SRM) mode for all the com￾pounds (Table 1). A confirmation transition for each analyte was also
monitored for more reliable results. In order to confirm successful en￾dogenous phospholipid elimination, mass to charge (m/z) 184 > 184
and 104 > 104 as common in-source collision-induced dissociation
(CID) ion fragments produced by endogenous phospholipids were
added to the method [30]. Moreover, m/z 524.4 > 184.1,
524.4 > 104.1, 496.4 > 184.1 and 496.4 > 104.1 produced by late
eluting phospholipids were included in the assay. The integration peak
area of the MRM transitions was calculated using MassHunter Work￾station Quantitative Analysis software for each analyte (Agilent Tech￾nologies, Madrid, Spain).
2.4. Sample cleanup
Two different sample preparation methods, SPE and PPT were
tested for analytes and their ISs extraction from human plasma. The
main objective of the present experiment was to choose the best ex￾traction method for the target compounds and clean up endogenous
phospholipids in the most efficient manner.
PPT extraction method: 300 μL of plasma sample was spiked with
10 μL ISs mix and 1200 μL precipitating agent, ACN with 0.1% formic
acid (4:1, v/v). The mixture was centrifuged at 14000 rpm (16947 g) at
4 °C for 5 min. Subsequently, the supernatant was collected and eva￾porated using a concentrator (5301, Eppendorf, Germany) at 45 °C for
2 h. Finally, the dry residue was reconstituted with 300 μL of 5%
NH4OH in MeOH/water reconstitution solution (1:1, v/v). Finally, five
μL was injected into the LC-MS/MS system.
SPE method: PRiME μ-SPE MCX (Mixed-mode Cation exchange
sorbent for bases) was applied as alternative sample preparation pro￾cedure. μ-SPE MCX includes both reversed-phase hydrophilic–lipophilic
sorbent and ion-exchange functionality for orthogonal sample pre￾paration. The SPE method consisted of three steps: sample loading,
washing and elution. Firstly, a mixture of 10 μL ISs mix along with
200 μL of 5% orthophosphoric acid in water was added to 300 μL of
plasma sample. The sample was loaded (2 × 255 μL) into the Oasis
PRiME μ-SPE MCX 96-well μElution Plate (Waters, Barcelona, Spain).
After the loading step, the first washing step was performed using
400 μL (2 × 200 μL) 100 mM ammonium formate + 2% formic acid in
water solution, followed by the second washing step with 400 μL
(2 × 200 μL) of 100% MeOH solution. Consequently, the compounds
were eluted with 100 μL (2 × 50 μL) of 5% NH4OH in MeOH solution
(1:1, v/v), and 100 μL (1 × 100 μL) of water, resulting in a 200 μL final
eluate volume. After each step, a vacuum of between 10 and 15 mm Hg
was applied until the wells were dry. The eluate was collected in a
1000 μL 96-well plate (Agilent Technologies, Santa Clara, USA) and
Table 1
Relevant LC-MS/MS conditions. Retention time (tR) in minutes (min), selectivity factor (Rs), qualification and confirmation transitions (m/z) used for selected
reaction monitoring (SRM) of 11 tyrosine kinase inhibitors, their stable isotope-labeled internal standards, caffeine and phospholipids.
Abbreviations: BARICI: Baricitinib; (Ba-Im): BARICI-IMA Rs; BOSU: Bosutinib; BOSU-D9: Bosutinib-D9, IS; (Bo-N): BOSU-NILO Rs; CAFF: caffeine; (C-Pe): CAFF￾PEFICI Rs; DASA: Dasatinib; DASA-D8: Dasatinib-D8, IS; (D-Bo): DASA-BOSU Rs; FILGO: Filgotinib; IBRU: Ibrutinib; IBRU-D5: Ibrutinib-D5, IS; (Im-F): IMA-FILGO
Rs; IS: Internal Standard; IMA: Imatinib; IMA-D8: Imatinib-D8, IS; LPC: Lysophosphatidylcholine; LPC 16: Lysophosphatidylcholine (16:0); LPC 18:
Lysophosphatidylcholine (18:0); NILO: Nilotinib; NILO-13C-D3: Nilotinib-13C-D3, IS; (N–Po): NILO-PONA Rs; PC-Phosphatidylcholine; PEFICI: Peficitinib; (Pe-T):
PEFICI-TOFACI Rs; PONA: Ponatinib; PONA-D8: Ponatinib-D8, IS; (Po-Ib): PONA-IBRU Rs; RUXOLI: Ruxolitinib; (R-D): RUXOLI-DASA Rs; (T-Ba): TOFACI-BARICI Rs;
TOFACI: Tofacitinib; TOFACI-13C3–15 N: Tofacitinib-13C3–15 N, IS; *8.88 (Po-Ib): PONA-IBRU Rs, the same for PONA and IBRU.
D. Koller, et al.
3
5 μL was directly injected into the LC-MS/MS system without eva￾poration and reconstitution steps. SPE procedure was chosen for sample
cleanup method in method validation.
2.5. Method validation
The validation of this method was based on FDA [27], EMA [28],
ICH [29] and partially on IUPAC [31], SANTE [32] and European
Commission Decision 2002/657/EC [33] guidelines. Robustness, line￾arity, selectivity, LLOQ, identity confirmation, precision and accuracy,
trueness, matrix effect, extraction recovery, process efficiency and
stability tests were performed to validate the present assay.
2.5.1. Robustness
Robustness is the ability of an analytical method to remain un￾affected by small variations in method parameters and influential en￾vironmental factors [29]. Mobile phase pH, column type and tem￾perature as well as laboratory room temperature were tested for
method robustness.
2.5.2. Linearity and lower limit of quantification (LLOQ)
Calibration curve range (see Table 2) was designed to cover the
plasma therapeutic range. The linearity of the assay was investigated by
analyzing eight CALs in six independent analytical runs. To meet the
validation criteria, the error of accuracy and relative standard deviation
(RSD, %) should not exceed 15% for each CAL. Calibration curve re￾gression calculation was performed by linear weighted least-squares
method with a weighting factor of 1/x, 1/x2 or 1/y, depending on the
analyte. Regression model adequacy was tested as suggested by IUPAC
guideline [31] by comparing the Lack-of-Fit test to pure error variances
at a 95% confidence level. The F value is used to accept or reject the
null hypothesis. To meet the acceptance criteria for the Lack-of-Fit test
(accept the null hypothesis), the Fcalculated value should be lower than
the Ftabulated value (see Table 2).
LLOQ was defined as the lowest quantifiable concentration. The
followed equation was used to estimate LLOQ: LLOQ = 10*(Sy,x/slope)
where Sy,x is the standard deviation (SD) of residuals in the calibration
curve area close to limit of detection (LOD).
2.5.3. Selectivity, specificity, confirmation of identity
Selectivity (specificity) refers to the ability of a method to measure
the amount of the analyte that is claimed to be measured [34]. Method
selectivity was ascertained for each validation series by analyzing 8
different blank plasma samples from human donors and a zero plasma
sample (consisting of a blank sample with IS) to exclude any inter￾ference with endogenous and/or exogenous substances. A method is
considered selective (specific), when the signal from the blank sample
at analyte’s tR is less than 20% of the LLOQ for the analyte and less than
5% for the IS. Moreover, we evaluated the selectivity of the method by
chromatographic resolution (Rs) between the analyte and the closest
eluting peak. A resolution of at least 1.5 is sought by AOAC [35] while
1.0 is the minimum useable separation. Nevertheless, FDA requires Rs
of minimum 2.0 [27]. In addition, tR-, relative tR-, and ion ratio
identity confirmation (based on qualifier ratio) according to SANTE
[32] and European Commission Decision 2002/657/EC [33] guidelines
were estimated. The following acceptance criteria were defined: tR
difference among extracted analyte and neat solution of the analyte
should be lower than 0.1 min [32], relative tR difference between ex￾tracted analyte and neat solution of the analyte should be lower than
2.5% [33], and ion ratio difference between CALs and QCs (samples)
should not differ more than 30% [32].
2.5.4. Precision (repeatability, intermediate precision), trueness and
accuracy
Accuracy and precision (repeatability, intermediate precision) were
evaluated using 4 QCs (‘LLOQ’, ‘low QC, ‘medium QC’ and ‘high QC’,
Supplementary Table 1) analyzed in a single analytical run on a single
day and 4 analytical runs from 4 different days. Precision was calcu￾lated as the mean concentration (ng/mL) ± SD and RSD (%). Accuracy
(%bias) was evaluated as the measured value divided by the nominal
value. Acceptance criteria are defined at RSD < 15% except for LLOQ
(RSD < 20%). Trueness understood as the systematic error of a mea￾surement system [36]. Trueness was calculated from spiking studies
[34]: a known amount of analyte was added to a known amount of
matrix. Four pre-extracted QC samples (analyte added to the biological
matrix before extraction, Pre) were analyzed from 4 months in com￾parison with spiking samples (analyte added to the biological matrix
after extraction, Post) as reference. Zeta-score was used to compare a
test value (Pre) to a reference value (Post). Values less than 2 were
recognized as satisfactory, those between 2 and 3 were defined as
questionable and those higher than 3 were considered unsatisfactory.
Additionally, we performed tests for incurred sample reanalysis (ISR).
The difference between the initial and the repeated concentration
should be less than ± 20% of the mean value for at least 67% of the ISR
samples [27].
2.5.5. Extraction recovery, matrix effect and process efficiency
RE, ME and PE were quantitatively estimated as described in the
literature
RE is generally calculated as the ratio of the analyte response added
to the biological matrix before extraction procedure (Pre) and the
analyte response added to the biological matrix after the extraction
process (Post). RE can be calculated as: a) relative: using the con￾centration ratio or b) absolute: using the XIC peak area ratio. The fol￾lowing equation was used to
ME is generally calculated as the ratio of the analyte response in the
presence of biological matrix added after extraction process (Post) and
the analyte response in absence of matrix, the neat solution of the
analyte. [5% NH4OH in MeOH/water reconstitution solution (1:1, v/v)]
without undergoing the extraction process. In the same fashion, abso￾lute and relative ME values were calculated. The equation below was
PE refers to the joint effect of possible analyte losses during sample
preparation and ionization suppression/enhancement in the ion source.
PE values were calculated as the ratio of the analyte response between
Table 2
Calibration range, correlation coefficient (R2) and Ftabulated and Fcalculated values
obtained from Lack-of-Fit test for 11 tyrosine kinase inhibitors and caffeine. To
consider the calibration model linear, the Fcalculated value should be lower than
Ftabulated value.
Abbreviations: BARICI: Baricitinib; BOSU: Bosutinib; CAFF: caffeine; DASA:
Dasatinib; FILGO: Filgotinib; IBRU: Ibrutinib; IMA: Imatinib; NILO: Nilotinib;
PEFICI: Peficitinib; PONA: Ponatinib; RUXOLI: Ruxolitinib; TOFACI:
Tofacitinib.
D. Koller, et al.
plasmas spiked before the extraction process (Pre) and the analyte re￾sponse in neat solution. Similar to RE and ME, absolute and relative
values were determined. For the evaluation of PE, the following equa￾tion was applied [37]:
PE Pre Neat solution (%) 100/ =
In each test, eight repetitions at low and high QC were analyzed in
blank plasma samples from 8 different human donors for all analytes.
RE, ME and PE values, must not be 100% but it is acceptable between
80 and 120% and the RSD should not exceed 15% for all QCs (except
for LLOQ, when 20% is allowed).
2.5.6. Stability
Stability was evaluated using four replicates at low and high QC
level for: a) extracted samples (autosampler stability) b) unextracted
samples (room temperature, +4 °C and −80 °C stability) c) short-term
conditions (72 h at room temperature, 72 h at +4 °C and 3 freeze-thaw
cycles at −80 °C stability) and d) long-term conditions (2 months at
−80 °C stability). Analyte stability is acceptable between 80 and 120%
with an RSD less than 15% for the studied QCs. The concentration ratio
between nominal value and different storage conditions value was used
to determine analyte stability.
2.5.7. Carry-over
The carry-over effect described in FDA [27] and EMA [28] guide￾lines is understood as a sample contamination causing analyte peaks to
re-appear in later runs (analyte peaks in blank runs). Here, the carry￾over effect was tested by injecting a blank sample after the highest CAL.
If the peak area of the blank sample is higher than 20% of the LLOQ
peak area and 5% of the IS peak area, there is a significant carry-over
effect. ACN/water solution (50:50, v/v) was applied for needle wash
between injections to prevent the mentioned phenomenon.
2.5.8. Application of the method to TDM samples
The present assay was successfully applied to TDM. Hereby, we
determined the through plasma concentration (Cmin) and Cmax of 260
plasma samples from 142 CML patients: DASA (n = 71), IMA (n = 61),
NILO (n = 8) and PONA (n = 2). Blood samples were collected in 3 mL
ethylenediaminetetraacetic acid potassium salt (EDTA-K2) tubes at
steady state, just before the next administration (Cmin) and 1 (DASA), 2
(IMA), 3 (NILO) or 4 h (PONA) after drug administration (Cmax).
Interpatient RSD (%) was evaluated for each drug and their Cmin and
Cmax values, respectively. The informed consent was obtained from all
the patients and approved by the local Ethics Committee (Clinical
Research Ethics Committee of “Hospital Universitario de la Princesa,”Madrid, Spain).
2.5.9. Statistics
Microsoft Excel (Microsoft® Office® 2003, Microsoft Corp., USA)
was utilized for data analysis (mean values and SDs). As the IUPAC
validation guideline suggests [31], Lack-of-Fit test was used to evaluate
linearity at a 95% confidence level. Zeta-score test was employed for
trueness evaluation.
3. Results and discussion
3.1. LC-MS/MS optimization and method robustness
pKa values for the majority of TKIs included in the present method
were similar ranging from 6.30 to 8.43. However, BARICI, FILGO,
RUXOLI and PEFICI present lower pKa values between 3.91 and 5.70.
CAFF has a completely different pKa value (strongest basic) of −0.92.
All these differences made it challenging to optimize a single extraction
method for all the compounds. Water solubility varied for all TKIs,
ranging from 0.002 to 0.02 mg/mL for IMA, DASA, BOSU, NILO, PONA
and IBRU while it was higher for PEFICI, TOFACI, BARICI and RUXOLI
(ranging from 0.116 to 0.357 mg/mL). Moreover, CAFF has higher
water solubility than the other included compounds with the value of
11 mg/mL. Therefore, the elution was performed in two steps: firstly an
organic solvent at high pH was applied (to elute water insoluble com￾pounds) and finally only water was used to elute CAFF due to its high
water solubility.
Mobile phase pH, column type and its temperature as well as la￾boratory room temperature were tested for method robustness.
Initially, two different columns were compared: Zorbax SB-C18 column
(2.1 × 50 mm, 1.8 μm; Agilent Technologies) and Poroshell 120 EC C18
column (2.1 × 75 mm, 2.7 μm; Agilent Technologies). Finally, due to its
superficially porous particles and therefore high resolution efficiency at
low pressure, the Poroshell column was chosen. To optimize chroma￾tographic conditions, we tested different analytical column tempera￾tures (40, 50 and 60 °C), buffer compositions (4 mM ammonium for￾mate vs 0.1% formic acid), buffer pH (2 and 3.2), mobile phase flow
rates (0.5 and 0.6 mL/min) and variety of gradients. As a result, final
chromatographic conditions were set as follows: 0.1% formic acid at pH
2.0 as aqueous buffer, flow rate of 0.5 mL/min and the column tem￾perature of 60 °C. By changing mobile phase conditions (from 4 mM
ammonium formate to 0.1% formic acid and pH from 3.2 to 2) we
improved method sensitivity (DASA LLOQ 0.75 ng/mL vs 0.38 ng/mL)
compared to our previous assay [39]. Different laboratory room tem￾peratures were tested (20, 22 and 25 °C). The best results for method
reproducibility were achieved at 20 °C. Thus, the method validation
was performed at this temperature.
The total run time including chromatographic separation (8.0 min)
and re-equilibrating steps (4.0 min) was of 12.0 min. Despite the ex￾istence of analytical methods with shorter run times [13,14], we found
it essential to elute endogenous phospholipids outside of the elution
profile of the analytes (Fig. 1). To prevent ME and LC-MS/MS instru￾ment contamination [40], we used a small sample injection volume.
Five μL of the extract was injected into the HPLC system. To avoid
carry-over phenomenon in subsequent injections, a mixture of organic￾aqueous solvent (ACN/water, 50:50; v/v) was used for washing be￾tween injections. Fig. 1 shows total ion chromatograms (TIC, Panel A)
and extracted ion chromatograms (XIC, Panel B) of PEFICI, TOFACI,
BARICI, IMA, FILGO, RUXOLI, DASA, BOSU, NILO, PONA, IBRU, CAFF
and plasma phospholipids.
Concerning MS/MS conditions, ESI in positive mode was selected
for scanning all analytes. As their polarity was positive, dMRM scan
mode was applied to significantly improve the analytes’ peak shape and
method selectivity. Table 1 displays relevant LC-MS/MS characteristics
and Supplementary Fig. 1 depicts mass spectrums of 11 TKIs, CAFF and
endogenous phospholipids. Fragmentation patterns for all compounds,
except for FILGO and phospholipids which are not available, were
confirmed by DrugBank LC-MS/MS spectra [41].
Crosstalk phenomenon in the collision cell during MS/MS detection
occurs when product ions from one SRM transition are scanned out
during another transition [42]. For this reason, crosstalk was taken into
account during method optimization. Since CAFF and IBRU generate
the same product ion (m/z 138.2), they may undergo crosstalk phe￾nomenon. Thus, apart from SIL-ISs, we used chromatographic separa￾tion (CAFF tR = 0.620 min vs IBRU tR = 6.712 min) to prevent cross￾talk effect. Similarly, BARICI and RUXOLI detection is based
on almost identical product ions (m/z 186.0 vs 186.1). Therefore, they
were successfully separated on the analytical column (BARICI
tR = 1.313 min vs RUXOLI tR = 4.406 min) to ensure collision cell
emptying between transitions. Qualifier ratio was used as well to ensure
the reliability of the analyte detection. Additionally, we analyzed tR-,
relative tR-, and ion ratio-based identity confirmation. tR differences
between extracted analytes and neat solution of the analytes were lower
than 0.1 min in all cases. When analyzing tR-ion ratio-based identity
confirmation, it was lower than 2.5%. Relative tR-identity confirmation
and the ion ratio difference between CALs and QCs did not differ more
than 30%.
D. Koller, et al.
5 3.2. Linearity and lower limit of quantification (LLOQ)
Calibration curve range (see Table 2) covers the plasma therapeutic
range. Fig. 1, Panel B depicts LLOQs for all 11 TKIs and CAFF.
Achieving low LLOQs is very important for drugs with low Cmin values,
such as for DASA or PONA. We acquired lower LLOQ for DASA than
those obtained by Haoulala et al. [17], De Francia et al. [15], Lankheet
et al. [43], improving our previous results [39]. Our LLOQ for DASA is
higher than the values achieved by Merienne et al. [14] and Bouchet
et al. [13] as UHPLC instruments were used in their studies. Of note, we
improved the LLOQ values for RUXOLI, TOFACI and PONA when
compared to the bibliography [14,16,44,45].
3.3. Selectivity, specificity, confirmation of identity
Qualifier ratio for all 11 TKIs and CAFF is shown in Table 1. Fig. 1
displays the total ion chromatogram (TIC) of all 11 TKIs and CAFF
which are eluted in a separated area than endogenous phospholipids.
Corresponding Rs for peak pairs is depicted in Fig. 1, Panel A and XIC at
LLOQ level are presented in Fig. 1, Panel B. Although complete chro￾matographic separation (Rs ≥ 1.5) was not achieved for some peak
pairs (CAFF-PEFICI, Rs = 0.64; PEFICI-TOFACI, Rs = 1.25; IMA-FILGO,
Rs = 0.77 and RUXOLI-DASA, Rs = 1.27), tR-, relative tR-, and ion
ratio identity confirmation tests were evaluated to confirm identity of
all compounds included in our analysis while fulfilling the regulatory
requirements.
3.4. Precision (repeatability, intermediate precision), trueness and accuracy
Results concerning precision and accuracy are presented in
Supplementary Table 1. In regard to reparability (1 day) and inter￾mediate precision (4 days), we accomplished an RSD lower than 11%
and pooled RSD of 12.94%, respectively. CAFF represented higher RSD
values due to the lack of SIL-ISs. Bias of accuracy was lower than 15%
for all compounds. Zeta-score values calculated for trueness of the
method were satisfactory, being lower than 2. Furthermore, method
reliability was verified by reanalysis of 28 incurred samples from a total
number of 260 samples. The results showed that for 96.43% of the
Fig. 1. Total ion chromatogram (TIC, Panel A) and extraction ion chromatograms (XIC, Panel B) of 11 tyrosine kinase inhibitors (PEFICI: peficitinib; TOFACI:
tofacitinib BARICI: baricitinib; IMA: imatinib; FILGO: filgotinib; RUXOLI: ruxolitinib; DASA: dasatinib; BOSU: bosutinib; NILO: nilotinib; PONA: ponatinib IBRU:
ibrutinib), caffeine (CAFF) and the stable isotope-labeled internal standard (SIL-IS), [13C3, 15 N]-tofacitinib (tofacitinib-13C3–15 N), [2H8]-imatinib (imatinib-D8),
[2H8]-dasatinib (dasatinib-D8), [2H9]-bosutinib (bosutinib-D9), [13C, 2H3]-nilotinib (nilotinib-13C-D3), [2H8]-ponatinib (ponatinib-D8), [2H5]-ibrutinib (ibru￾tinib-D5), as well as phospholipids lysophosphatidylcholine 16:0 (LPC 16:0) and lysophosphatidylcholine 18:0 (LPC 18:0) from human blank plasma at medium
quality control concentration (Panel A) and lower limit of quantification (LLOQ, Panel B). Additionally, Panel A displays chromatografic resolution (Rs) for eluted
peak pairs in order from 1 to 11. Retention times (tR) and concentration (ng/mL) values calculated from calibration curves are given for all analytes. The results are
presented as the percentage of counts (%) versus time in minutes (min). All chromatograms have been normalized to the largest peak.
D. Koller, et al. Talanta
6
reanalyzed samples, the difference between the repeated and original
concentration was within ± 9.61%. Only for 3.57% of samples, the
difference was higher than 20%; what still met the acceptance criteria.
3.5. Extraction recovery, matrix effect, process efficiency and endogenous
phospholipids elimination
RE, ME and PE are all related to trueness. In LC-MS analysis it is
useful to make a distinction between RE (what refers to the analyte
losses during sample extraction process) and PE (what describes the
analyte signal changes including the effects from sample extraction, RE
and analyte ionization/detection, ME) [37,46].
For the determination of relative and absolute RE, I and PE, two QC
levels— low and high— were applied. The results regarding RE, ME and
PE for SPE compared to PPT are depicted in Supplementary Fig. 2 and
2. The outcomes are represented as mean percentages of relative
(Supplementary Fig. 2, Panel A, C and Fig. 2, Panel A) and absolute
(Supplementary Fig. 2, Panel B, D and Fig. 2, Panel B) RE, ME and PE,
whereas RSD values are shown as error bars.
Relative RE values for SPE ranged from 82 to 114% (except for
TOFACI with the value of 120%) and the RSD was not higher than 13%.
Absolute RE values were obtained between 79 and 144%, with an RSD
within 22%. On the contrary, the relative mean RE with PPT were
accomplished within 80–124% for all compounds (RSD = 26%), while
absolute RE ranged between 60 and 102% with significantly higher
RSD of 39%. Inspecting the bibliographic data for relative SPE results,
we found that other authors [13,39,44,45] obtained similar results.
Absolute RE results are not usually shown in published methods.
Nevertheless, Merienne et al., displayed the absolute values of RE for
SPE, being between 20 and 30% — relatively lower than our values
(79–144%). PPT RE bibliographic results oscillated between 54 and
114% [15,17,43], being lower than the values we obtained.
Regarding ME, relative SPE values ranged between 85 and 118%
(RSD within 15%), greatly better than absolute ME ranging from 69 to
140% with an RSD of 23%. The relative ME for PPT were as followed:
73–136% and RSD within 17%, while absolute values were comprised
between 26 and 119% with an RSD of 32%. Relative ME calculated for
SPE in the literature ranges from 83 to 114% [13,14,44] while for PPT
varies from 73 to 126% [15–17,43], both analogous to our results.
With reference to PE, relative PE values fluctuated between 80 and
118% (RSD = 7%), while absolute PE ranged from 68 to 125% with an
RSD of 13%, except for CAFF representing a value of 132%
(RSD = 7%). However, the results obtained with PPT were certainly
poorer compared to SPE. Relative PPT-PE ranged between 80 and 161%
(RSD = 25%); absolute PE was relatively lower ranging from 20 to 71%
with a high RSD value of 37%. Regarding bibliographic data, there are
Fig. 2. Matrix effect and endogenous phospholipid elimination efficiency tests achieved with solid phase extraction (SPE) and protein precipitation (PPT) for 11
tyrosine kinase inhibitors (PEFICI: peficitinib; TOFACI: tofacitinib BARICI: baricitinib; IMA: imatinib; FILGO: filgotinib; RUXOLI: ruxolitinib; DASA: dasatinib; BOSU:
bosutinib; NILO: nilotinib; PONA: ponatinib IBRU: ibrutinib) and caffeine (CAFF). Matrix effect is expressed as relative (to their stable isotope-labeled internal
standard) (Panel A) and absolute (Panel B) values. Averaged data are given as the percentage of the mean matrix effect ± relative standard deviation (RSD, % shown
as error bar) of the total number of n = 8 for low and high quality controls (QCs). Endogenous phospholipids [early eluting lysophosphatidylcholine (LPC) and
phosphatidylcholine (PC) as well as late eluting lysophosphatidylcholine 16:0 (LPC 16:0) and lysophosphatidylcholine 18:0 (LPC 18:0)] removal from biological
matrix comparison between SPE and PPT is shown in Panel C. Data are expressed as mean percentages ± relative standard deviation (RSD, %) of 31 human plasma
samples. Paired t-test, one tailed was performed to compare phospholipids removal using PPT versus SPE, p < 0.001.
7
only two methods evaluating relative PE as a test included in the
method validation [14,17], where it ranged from 83 to 115%.
On the whole, more preferable values for ME, RE and PE were ob￾tained when SIL-ISs were applied (relative) in comparison to absolute
values. SIL-ISs are known to compensate for analyte loss during ex￾traction process (RE) and for its suppression or enhancement during the
ionization process (ME). Notwithstanding, SPE results were similar, but
more satisfactory than PPT. Although PPT results were analogous, the
differences lie in the RSD values. RSD is greatly higher with PPT,
causing method irreproducibility and compromised precision and ac￾curacy.
After applying an appropriate type of SPE sorbent (hydrophilic-li￾pophilic with mixed mode component) for amphiphilic molecules, we
were able to remove more than 91% of early eluting phospholipids
(92% of lysophosphatidylcholine, LPC and 99% of phosphatidylcholine,
PC) and more than 96% of late eluting phospholipidic species (99% of
lysophosphatidylcholine 16:0, LPC 16:0 and 97% of lysopho￾sphatidylcholine 18:0, LPC 18:0; Fig. 2, Panel C) compared to PPT.
Based on that, we chose this procedure for method validation. As some
phospholipidic species persist in the sample, a good chromatographic
resolution was applied to separate phospholipids from the analytes of
interest. Target compounds were eluted within 0.620–6.712 min, out￾side of the phospholipids elution times (tR = 0.373 min and tR = 0.397
for early eluting phospholipids; tR = 8.164 min and tR = 8.962 for late
eluting phospholipid species). Endogenous phospholipid removal en￾sures method reproducibility, therefore it was an important step during
the optimization of our method. According to our best knowledge, no
other method evaluated endogenous phospholipid elimination effi￾ciency along with TKIs quantification.
Our results clearly show the importance of PE and the necessity of
phospholipid elimination while evaluating RE and ME during method
development and validation. Even though no significant ME could be
observed in the quantitative ME evaluation experiment, matrix com￾ponents influence the overall PE and endogenous phospholipids could
have an impact on chromatographic separation and ionization of target
compounds.
3.6. Stability
Analyte stability is crucial. Therefore stability should be ensured
during the whole analytical procedure. In the present analytical
method, we tested stability in unextracted sample and after sample
extraction along with short and long-term stabilities. Stability assay
results at low and high QCs are depicted in Supplementary Fig. 3, Panel
A, B, C, D, E. All analytes were stable in extracted and unextracted
samples at all investigated storage conditions (stabilities 80–120% and
RSD < 15%), except for IBRU after 48 h and 72 h at room temperature
(71 and 53%, respectively) and CAFF after 2-months storage at −80 °C
(67%), especially at low QC level. Based on our experience, most TKIs
are relatively stable. However, IBRU showed rapid degradation at room
temperature in plasma. IBRU was shown to be unstable in whole blood
and oral fluid under most conditions studied by Rood et al. [47].
Therefore, it requires special care during transportation at room tem￾perature.
3.7. Carry-over
In case carry-over effect is detected during analytical method opti￾mization, it should be corrected, as it may have a negative influence on
precision, accuracy and trueness of the assay. The response of the blank
sample at the tR of the analytes was < 18% of the corresponding peak
area of the LLOQ samples (except for PONA with the value of 24.34%)
and < 0.01% for IS peak area. Thus, carry-over met the acceptance
criteria for all compounds except for PONA. Consequently, as PONA
was present in all CALs and QCs, additional washing steps were applied
to reduce PONA carry-over phenomenon to less than 20%. Carry-over
phenomenon detected by other authors [14,43] was similar to ours,
except for Van Erp et al. [48], who managed to reduce the carry-over
effect to less than 10% for all analytes.
3.8. Application of the method to TDM samples
We were able to monitor plasma levels of 11 TKIs in a single ex￾traction process and single run, reducing the cost of the analysis and
saving time needed for sample processing. Data concerning TDM of
DASA, IMA, NILO and PONA are shown in Table 3. Plasma levels were
compared to bibliographic data published by Yu et al. [6], Rood et al.
[9] and Miura [8] to demonstrate the feasibility of the method. Plasma
concentrations are characterized by very high interpatient variability
ranging from 24% for NILO to 104% for DASA, similar to the results
obtained by Merienne et al. [14]. CAFF levels were not detectable in
Cmin samples. However, Cmax samples represent variable CAFF levels.
Further investigations should be performed in each patient separately
with and without CAFF uptake to evaluate possible CAFF impact on TKI
pharmacokinetics.
4. Conclusion
The proposed LC-MS/MS method is based on simultaneous and re￾liable measurement of 11 TKIs and CAFF in human plasma. Efficient
three step μ-SPE ensures excellent endogenous phospholipid elimina￾tion resulting in high extraction recoveries, no significant matrix effect
and exceptional overall process efficiency. The assay is distinguished by
its sensitivity, competitive LLOQs and calibration range, good analyte
stabilities and no significant carry-over effect. The present method is
cost effective, as it enables the measurement of 11 TKIs and CAFF in a
single run, thus it can be easily applied for routine TDM.
Declaration of conflicting interests
F. Abad-Santos has been consultant or investigator in clinical trials
sponsored by the following pharmaceutical companies: Abbott, Alter,
Bristol-Myers Squibb, Chemo, Cinfa, FAES, Farmalíder, Ferrer,
GlaxoSmithKline, Galenicum, Gilead, Janssen-Cilag, Kern, Normon,
Novartis, Servier, Silverpharma, Teva, and Zambon. The remaining
authors declare no conflicts of interest.
Acknowledgements
The authors are thankful to Sandra Salvador, Xavier Rodríguez and
Ignazio Garaguso (Waters Corporation) and Leticia Rodriguez as well as
Fernando Tamarit (Agilent Technologies) for their help and technical
assistance.
Table 3
Imatinib, dasatinib, nilotinib and ponatinib therapeutic drug monitoring results
of real samples obtained from 142 CML patients. The data are shown as plasma
through concentration (Cmin) and plasma maximum concentration (Cmax) (ng/
mL) ± standard deviation (SD) and interpatient relative standard deviation
Abbreviations: Cmax: plasma maximum (peak) concentration; Cmin: plasma
minimum (through) concentration; DASA: Dasatinib; IMA: Imatinib; NILO:
Nilotinib; PONA: Ponatinib.
D. Koller, et al. Talanta xxx (xxxx) xxxx
8 Appendix A. Supplementary data
Supplementary data to this article can be found online at https://
doi.org/10.1016/j.talanta.2019.120450.
Funding sources
This work was partially supported by Bristol-Myers Squibb (Madrid,
Spain) and the H2020 Marie Sklodowska-Curie Innovative Training
Network 721236 grant to Dora Koller.
References
[1] N. Widmer, C. Bardin, E. Chatelut, A. Paci, J. Beijnen, D. Levêque, G. Veal, A. Astier,
Review of therapeutic drug monitoring of anticancer drugs part two – targeted
therapies, Eur. J. Cancer 50 (2014) 2020–2036, https://doi.org/10.1016/j.ejca.
2014.04.015.
[2] J.L. Steegmann, M. Baccarani, M. Breccia, L.F. Casado, V. García-Gutiérrez,
A. Hochhaus, D.-W. Kim, T.D. Kim, H.J. Khoury, P. Le Coutre, J. Mayer,
D. Milojkovic, K. Porkka, D. Rea, G. Rosti, S. Saussele, R. Hehlmann, R.E. Clark,
European LeukemiaNet recommendations for the management and avoidance of
adverse events of treatment in chronic myeloid leukaemia, Leukemia 30 (2016)
1648–1671, https://doi.org/10.1038/leu.2016.104.
[3] H.-J. Klümpen, C.F. Samer, R.H.J. Mathijssen, J.H.M. Schellens, H. Gurney, Moving
towards dose individualization of tyrosine kinase inhibitors, Cancer Treat Rev. 37
(2011) 251–260, https://doi.org/10.1016/j.ctrv.2010.08.006.
[4] B.S. Kappelhoff, K.M.L. Crommentuyn, M.M.R. de Maat, J.W. Mulder,
A.D.R. Huitema, J.H. Beijnen, Practical guidelines to interpret plasma concentra￾tions of antiretroviral drugs, Clin. Pharmacokinet. 43 (2004) 845–853, https://doi.
org/10.2165/00003088-200544040-00009.
[5] P. Herviou, E. Thivat, D. Richard, L. Roche, J. Dohou, M. Pouget, A. Eschalier,
X. Durando, N. Authier, Therapeutic drug monitoring and tyrosine kinase in￾hibitors, Oncol. Lett. 12 (2016) 1223–1232, https://doi.org/10.3892/ol.2016.
4780.
[6] H. Yu, N. Steeghs, C.M. Nijenhuis, J.H.M. Schellens, J.H. Beijnen, A.D.R. Huitema,
Practical guidelines for therapeutic drug monitoring of anticancer tyrosine kinase
inhibitors: focus on the pharmacokinetic targets, Clin. Pharmacokinet. 53 (2014)
305–325, https://doi.org/10.1007/s40262-014-0137-2.
[7] G. Rosti, F. Castagnetti, G. Gugliotta, M. Baccarani, Tyrosine kinase inhibitors in
chronic myeloid leukaemia: which, when, for whom? Nat. Rev. Clin. Oncol. 14
(2017) 141–154, https://doi.org/10.1038/nrclinonc.2016.139.
[8] M. Miura, Therapeutic drug monitoring of imatinib, nilotinib, and dasatinib for
patients with chronic myeloid leukemia, Biol. Pharm. Bull. 38 (2015) 645–654,

https://doi.org/10.1248/bpb.b15-00103.

[9] J.J.M. Rood, J.H.M. Schellens, J.H. Beijnen, R.W. Sparidans, Recent developments
in the chromatographic bioanalysis of approved kinase inhibitor drugs in oncology,
J. Pharm. Biomed. Anal. 130 (2016) 244–263, https://doi.org/10.1016/j.jpba.
2016.06.037.
[10] R.W.F. van Leeuwen, T. van Gelder, R.H.J. Mathijssen, F.G.A. Jansman, Drug-drug
interactions with tyrosine-kinase inhibitors: a clinical perspective, Lancet Oncol. 15
(2014) e315–326, https://doi.org/10.1016/S1470-2045(13)70579-5.
[11] J.R. White, J.M. Padowski, Y. Zhong, G. Chen, S. Luo, P. Lazarus, M.E. Layton,
S. McPherson, Pharmacokinetic analysis and comparison of caffeine administered
rapidly or slowly in coffee chilled or hot versus chilled energy drink in healthy
young adults, Clin. Toxicol. 54 (2016) 308–312, https://doi.org/10.3109/
15563650.2016.1146740.
[12] K.I. Liszt, J.P. Ley, B. Lieder, M. Behrens, V. Stöger, A. Reiner, C.M. Hochkogler,
E. Köck, A. Marchiori, J. Hans, S. Widder, G. Krammer, G.J. Sanger, M.M. Somoza,
W. Meyerhof, V. Somoza, Caffeine induces gastric acid secretion via bitter taste
signaling in gastric parietal cells, Proc. Natl. Acad. Sci. U. S. A 114 (2017)
E6260–E6269, https://doi.org/10.1073/pnas.1703728114.
[13] S. Bouchet, E. Chauzit, D. Ducint, N. Castaing, M. Canal-Raffin, N. Moore, K. Titier,
M. Molimard, Simultaneous determination of nine tyrosine kinase inhibitors by 96-
well solid-phase extraction and ultra performance LC/MS-MS, Clin. Chim. Acta 412
(2011) 1060–1067, https://doi.org/10.1016/j.cca.2011.02.023.
[14] C. Merienne, M. Rousset, D. Ducint, N. Castaing, K. Titier, M. Molimard, S. Bouchet,
High throughput routine determination of 17 tyrosine kinase inhibitors by LC–MS/
MS, J. Pharm. Biomed. Anal. 150 (2018) 112–120, https://doi.org/10.1016/j.jpba.
2017.11.060.
[15] S. De Francia, A. D’Avolio, F. De Martino, E. Pirro, L. Baietto, M. Siccardi,
M. Simiele, S. Racca, G. Saglio, F. Di Carlo, G. Di Perri, New HPLC–MS method for
the simultaneous quantification of the antileukemia drugs imatinib, dasatinib, and
nilotinib in human plasma, J. Chromatogr. B 877 (2009) 1721–1726, https://doi.
org/10.1016/j.jchromb.2009.04.028.
[16] H.H. Huynh, C. Pressiat, H. Sauvageon, I. Madelaine, P. Maslanka, C. Lebbé,
C. Thieblemont, L. Goldwirt, S. Mourah, Development and validation of a si￾multaneous quantification method of 14 tyrosine kinase inhibitors in human plasma
using LC-MS/MS, Ther. Drug Monit. 39 (2017) 43–54, https://doi.org/10.1097/
FTD.0000000000000357.
[17] A. Haouala, B. Zanolari, B. Rochat, M. Montemurro, K. Zaman, M.A. Duchosal,
H.B. Ris, S. Leyvraz, N. Widmer, L.A. Decosterd, Therapeutic Drug Monitoring of
the new targeted anticancer agents imatinib, nilotinib, dasatinib, sunitinib,
sorafenib and lapatinib by LC tandem mass spectrometry, J. Chromatogr. B Anal.
Technol. Biomed. Life Sci. 877 (2009) 1982–1996, https://doi.org/10.1016/j.
jchromb.2009.04.045.
[18] A. Van Eeckhaut, K. Lanckmans, S. Sarre, I. Smolders, Y. Michotte, Validation of
bioanalytical LC–MS/MS assays: evaluation of matrix effects, J. Chromatogr. B 877
(2009) 2198–2207, https://doi.org/10.1016/j.jchromb.2009.01.003.
[19] B.K. Matuszewski, M.L. Constanzer, C.M. Chavez-Eng, Strategies for the assessment
of matrix effect in quantitative bioanalytical methods based on HPLC−MS/MS,
Anal. Chem. 75 (2003) 3019–3030, https://doi.org/10.1021/ac020361s.
[20] J.-F. Bienvenu, G. Provencher, P. Bélanger, R. Bérubé, P. Dumas, S. Gagné,
É. Gaudreau, N. Fleury, Standardized procedure for the simultaneous determination
of the matrix effect, recovery, process efficiency, and internal standard association,
Anal. Chem. 89 (2017) 7560–7568, https://doi.org/10.1021/acs.analchem.
7b01383.
[21] X. Guo, E. Lankmayr, Phospholipid-based matrix effects in LC–MS bioanalysis,
Bioanalysis 3 (2011) 349–352, https://doi.org/10.4155/bio.10.213.
[22] Y.-Q. Xia, M. Jemal, Phospholipids in liquid chromatography/mass spectrometry
bioanalysis: comparison of three tandem mass spectrometric techniques for mon￾itoring plasma phospholipids, the effect of mobile phase composition on phospho￾lipids elution and the association of phospholipids with matrix effects, Rapid
Commun. Mass Spectrom. 23 (2009) 2125–2138, https://doi.org/10.1002/rcm.
4121 , Accessed date: 21 August 2017.
[23] G.B. Phillips, J.T. Dodge, Composition of phospholipids and of phospholipid fatty
acids of human plasma, J. Lipid Res. 8 (1967) 676–681.
[24] P.G. Munder, M. Modolell, R. Andreesen, H.U. Weltzien, O. Westphal,
Lysophosphatidylcholine (lysolecithin) and its synthetic analogues.
Immunemodulating and other biologic effects, Springer Semin. Immunopathol. 2
(1979) 187–203, https://doi.org/10.1007/BF01891668.
[25] M. Birch, P.E. Morgan, S. Handley, A. Ho, R. Ireland, R.J. Flanagan, Simple meth￾odology for the therapeutic drug monitoring of the tyrosine kinase inhibitors da￾satinib and imatinib: therapeutic drug monitoring of dasatinib and imatinib,
Biomed. Chromatogr. (2012), https://doi.org/10.1002/bmc.2796 n/a-n/a.
[26] E. Kralj, J. Trontelj, T. Pajič, A. Kristl, Simultaneous measurement of imatinib, ni￾lotinib and dasatinib in dried blood spot by ultra high performance liquid chro￾matography tandem mass spectrometry, J. Chromatogr. B Anal. Technol. Biomed.
Life Sci. 903 (2012) 150–156, https://doi.org/10.1016/j.jchromb.2012.07.011.
[27] FDA, Bioanalytical method validation. Guidance for industry, https://www.fda.
gov/downloads/drugs/guidances/ucm070107.Pdf, (2018).
[28] EMA, Guideline on bioanalytical method validation, http://www.ema.europa.eu/
docs/en_GB/document_library/Scientific_guideline/2011/08/WC500109686.pdf,
(2011).
[29] ICH, International, Conference on harmonization of technical requirements for re￾gistration of pharmaceuticals for human use. Harmonized tripartite guideline, va￾lidation of analytical procedures: text and methology Q2, http://academy.gmp￾compliance.org/guidemgr/files/Q2(R1).PDF, (2005).
[30] J.L. Little, M.F. Wempe, C.M. Buchanan, Liquid chromatography-mass spectro￾metry/mass spectrometry method development for drug metabolism studies: ex￾amining lipid matrix ionization effects in plasma, J. Chromatogr. B Anal. Technol.
Biomed. Life. Sci. 833 (2006) 219–230, https://doi.org/10.1016/j.jchromb.2006.
02.011.
[31] IUPAC, Harmonized guidelines for single laboratory validation of methods of
analysis, http://iupac.org/publications/pac/pdf/2002/pdf/7405×0835.pdf,
(2002).
[32] SANTE, SANTE/11813/2017. EUROPEAN COMMISSION directorate general for
health and food safety. Implemented by 01/01/2018, https://ec.europa.eu/food/
sites/food/files/plant/docs/pesticides_mrl_guidelines_wrkdoc_2017-11813.pdf,
(2017).
[33] 2002/657/EC, COMMISSION DECISION implementing Council Directive 96/23/EC
concerning the performance of analytical methods and the interpretation of results,

https://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX:32002D0657

&from=SL, (2002).
[34] A. Kruve, R. Rebane, K. Kipper, M.-L. Oldekop, H. Evard, K. Herodes, P. Ravio,
I. Leito, Tutorial review on validation of liquid chromatography-mass spectrometry
methods: part I, Anal. Chim. Acta 870 (2015) 29–44, https://doi.org/10.1016/j.
aca.2015.02.017.
[35] AOAC, AOAC Guidelines for Single-Laboratory Validation of Chemical Methods for
Dietary Supplements and Botanicals, Official Methods of Analysis, nineteenth ed.,
(2012) Appendix K https://www.aoac.org/aoac_prod_imis/AOAC_Docs/
StandardsDevelopment/SLV_Guidelines_Dietary_Supplements.pdf.
[36] JCGM 200, International vocabulary of metrology-Basic and general concepts and
associated terms (VIM), 2008, 2012. https://www.bipm.org/utils/common/
documents/jcgm/JCGM_200_2012.pdf.
[37] P.J. Taylor, Matrix effects: the Achilles heel of quantitative high-performance liquid
chromatography–electrospray–tandem mass spectrometry, Clin. Biochem. 38
(2005) 328–334, https://doi.org/10.1016/j.clinbiochem.2004.11.007.
[38] B.K. Matuszewski, M.L. Constanzer, C.M. Chavez-Eng, Strategies for the assessment
of matrix effect in quantitative bioanalytical methods based on HPLC−MS/MS,
Anal. Chem. 75 (2003) 3019–3030, https://doi.org/10.1021/ac020361s.
[39] A. Wojnicz, B. Colom-Fernández, J.L. Steegmann, C. Muñoz-Calleja, F. Abad-Santos,
A. Ruiz-Nuño, Simultaneous determination of imatinib, dasatinib, and nilotinib by
liquid chromatography-tandem mass spectrometry and its application to ther￾apeutic drug monitoring, Ther. Drug Monit. 39 (2017) 252–262
[40] L.L. Jessome, D.A. Volmer, Ion Suppression : a Major Concern in Mass
Spectrometry, (2006).
[41] DrugBank, (n.d.)
[42] L.-P. Morin, J.-N. Mess, M. Furtado, F. Garofolo, Reliable procedures to evaluate
and repair crosstalk for bioanalytical MS/MS assays, Bioanalysis 3 (2011) 275–283,

https://doi.org/10.4155/bio.10.203.

[43] N.A.G. Lankheet, M.J.X. Hillebrand, H. Rosing, J.H.M. Schellens, J.H. Beijnen,
A.D.R. Huitema, Method development and validation for the quantification of da￾satinib, erlotinib, gefitinib, imatinib, lapatinib, nilotinib, sorafenib and sunitinib in
human plasma by liquid chromatography coupled with tandem mass spectrometry:
simultaneous determination of eight tyrosine kinase inhibitors for TDM, Biomed.
Chromatogr 27 (2013) 466–476, https://doi.org/10.1002/bmc.2814.
[44] C. Pressiat, H.-H. Huynh, A. Plé, H. Sauvageon, I. Madelaine, C. Chougnet, C. Le
Maignan, S. Mourah, L. Goldwirt, Development and validation of a simultaneous
quantification method of ruxolitinib, vismodegib, olaparib, and pazopanib in
human plasma using liquid chromatography coupled with tandem mass spectro￾metry, Ther. Drug Monit. 40 (2018) 337–343
[45] A.S. Abdelhameed, M.W. Attwa, A.A. Kadi, An LC-MS/MS method for rapid and
sensitive high-throughput simultaneous determination of various protein kinase
inhibitors in human plasma, Biomed. Chromatogr. BMC. 31 (2017)
[46] A. Kruve, R. Rebane, K. Kipper, M.-L. Oldekop, H. Evard, K. Herodes, P. Ravio,
I. Leito, Tutorial review on validation of liquid chromatography–mass spectrometry
methods: Part II, Anal. Chim. Acta 870 (2015) 8–28
[47] J.J.M. Rood, S. van Hoppe, A.H. Schinkel, J.H.M. Schellens, J.H. Beijnen,
R.W. Sparidans, Liquid Filgotinib chromatography-tandem mass spectrometric assay for the
simultaneous determination of the irreversible BTK inhibitor ibrutinib and its dihydrodiol-metabolite in plasma and its application in mouse pharmacokinetic studies, J. Pharm. Biomed. Anal. 118 (2016) 123–131,
[48] N.P. van Erp, D. de Wit, H.-J. Guchelaar, H. Gelderblom, T.J. Hessing, J. den
Hartigh, A validated assay for the simultaneous quantification of six tyrosine kinase
inhibitors and two active metabolites in human serum using liquid chromatography
coupled with tandem mass spectrometry, J. Chromatogr. B 937 (2013) 33–43