Journal of Pharmaceutical and Biomedical Analysis 150 (2018) 427–435
Contents lists available at ScienceDirect
Journal of Pharmaceutical and Biomedical Analysis
j ournal homepage: www.elsevier.com/locate/jpba
Bionalytical validation study for the determination of unbound
ambrisentan in human plasma using rapid equilibrium dialysis
followed by ultra performance liquid chromatography coupled to
mass spectrometry
a a a,∗ a
Soledad Garcia-Martínez , Estitxu Rico , Enriqueta Casal , Alba Grisalena˜ ,
a a a b a
Eider Alcaraz , Nicholas King , Nerea Leal , Iker Navarro , Miguel Ángel Campanero
a
Bioanalytical services, Dynakin, S.L, Bizkaia Technology Park, 48160 Derio, Bizkaia, Spain
b
Noorik Biopharmaceuticals, Lange Gasse 15, 4052 Basel, Switzerland
a r t i c l e i n f o a b s t r a c t
Article history: Ambrisentan is a highly selective endothelin-1 type A receptor antagonist indicated for use in the
Received 10 July 2017
treatment of pulmonary hypertension. In this study an assay was developed and validated for the quan-
Received in revised form
tification of total and unbound (free) concentrations of ambrisentan in human plasma. Plasma samples
13 December 2017
were dialysed against phosphate buffered saline in a rapid equilibrium dialysis device to obtain dialysate
Accepted 15 December 2017
and plasma for unbound and total ambrisentan, respectively. Subsequently, ambrisentan and deuterated
Available online 16 December 2017
ambrisentan (internal standard) were extracted from plasma or plasma dialysate by solid-phase extrac-
tion and separated by ultra performance liquid chromatography using on a reversed-phase C column.
Keywords: 18
Detection was conducted with a tandem mass spectrometer with an electrospray ionization source and
Unbound ambrisentan
analysed in positive ion mode with multiple reaction monitoring. Calibration curves were generated over
Total ambrisentan
Rapid equilibrium dialysis (RED) a linear concentration range of 0.1–200 ng/mL in plasma and 0.1–10 ng/mL in plasma ultrafiltrate; with a
Ultra performance liquid recovery for ambrisentan of 69.4% and 77.5%, respectively. This assay has been shown to be reproducible
chromatography–tandem mass and sensitive. The lower limit of quantification in both cases was 0.1 ng/mL; reaching a sensitivity not
spectrometry (UPLC–MS/MS)
previously described in the literature. The inter- and intra-batch precision and accuracy were in both
cases ≤±15%. The procedure was applied to assess total and free plasma concentrations of ambrisentan
in healthy volunteers. Plasma protein binding of ambrisentan was approximately 99%.
© 2017 Elsevier B.V. All rights reserved.
1. Introduction AMB is highly protein-bound (98.8%), and is eliminated predom-
inantly by nonrenal mechanisms [2,3]. Therefore, the dose of AMB
Endothelin-1 is a potent vasoconstrictor and is mediated by two does not need adjustment in individuals with mild to moderate
types of endothelin receptors including type A and type B. Over- renal impairment. There are currently no data on the use of AMB in
expression of endothelin-1 has been demonstrated in patients with patients with severe renal impairment.
pulmonary arterial hypertension [1]. Ambrisentan (AMB), a highly Plasma protein binding of a drug can vary in various pathological
selective endothelin-1 type A receptor antagonist, is currently and physiological states, acting as a potential source of variability
approved in the United States of America, Europe and different in clinical efficacy. Hepatic impairment and nephritic syndrome are
countries for the treatment of this vascular pathology. AMB blocks associated with hypoalbuminemia which could potentially lead to
the effects of endothelin-1 and thus decreases blood pressure in altered protein binding levels of a drug.
the lungs. The thickening of blood vessels in the lungs and heart is Determination of plasma protein binding of drugs is critical in
also inhibited by AMB. the drug development since only the free drug concentration is
considered as the active fraction, because it is able to pass through
membranes to reach tissues and penetrate into cells to exert its
effect. Plasma protein binding of a drug is clinically important, since
it has profound effects on both the pharmacokinetics and pharma-
∗
Corresponding author. codynamics, as well as the safety margin [4,5]. The study of free drug
E-mail address: [email protected] (E. Casal).
https://doi.org/10.1016/j.jpba.2017.12.030
0731-7085/© 2017 Elsevier B.V. All rights reserved.
428 S. Garcia-Martínez et al. / Journal of Pharmaceutical and Biomedical Analysis 150 (2018) 427–435
concentrations can be of great utility whenever plasma protein used to prepare quality control (QC) samples, calibration standards,
binding changes occur, improving the therapeutic drug monitor- zeros and blank samples in plasma and plasma ultrafiltrate, respec-
ing approach based on total drug concentrations and contributing tively. Individual human plasma and plasma ultrafiltrate requested
to a better individualization of drug dosage regimen. Therefore, for the assessments of selectivity and matrix effect were provided
determination of unbound AMB concentrations in plasma may be by Sera-Lab Ltd. and BBI Solutions (Le Perray, France). Human whole
helpful in better characterizing the pharmacokinetics of the AMB blood was collected (by Laboratorio Axpe, Spain) from healthy
in a variety of clinical settings. donors, into collection tubes pre-treated with K2-EDTA.
The study of unbound concentration requires two steps. Firstly
the separation of the unbound and bound fractions. There are a 2.2. Preparation of calibration standards, quality control samples
number of commonly used methods available for the separation and internal standard solutions
of the unbound and bound fractions; these include ultrafiltra-
tion, ultracentrifugation, gel filtration, capillary electrophoresis A stock solution of 1.0 mg/mL AMB in dimity lsulphoxide was
and equilibrium dialysis. Extensive reviews of all the techniques used for preparation of calibration standards and QC samples. The
can be found in the literature [6,7]. However, rapid equilibrium solubility of AMB in different solvents was initially evaluated (data
dialysis (RED) has long been considered the gold standard method not shown). The results demonstrated that the AMB is soluble in
since non-specific adsorption of drugs to the device and membrane dimetyl sulphoxide, acetonitrile and acetonitrile:water (1:1, v:v)
has less impact than other techniques [8–10]. In addition, is the and slightly soluble in water and 0.1% formic acid in acetoni-
most amendable to automation, because it can be implemented trile:water (2:8, v:v). The stock (1 mg/mL in dimetyl sulphoxide)
◦
using simple liquid handling techniques. was stored protected from the light at −20 C and it was stable for
The second step in the determination of unbound concentra- at least eighteen days.
tion of a drug requires a methodology to separate and quantify the Two sets of intermediate calibration solutions were prepared by
drug; usually via chromatographic techniques coupled to different diluting of the stock solution in acetonitrile:water (1:1, v:v). One
types of detectors. A thorough literature survey has revealed that set (10000, 9000, 2500, 250, 50, 25, 10 and 5 ng/mL) was prepared
the methods currently available for analysis of AMB include ultra- for determination of total AMB in plasma. Another set (500, 450,
violet in pharmaceuticals and related substances [11–13] or mass 250, 200, 50, 25, 10 and 5 ng/mL) was prepared for determination
spectrometry detection in biological samples or pharmaceutical of AMB in plasma ultrafiltrate. These intermediate calibration solu-
◦
degraded samples [14–18]. None of these methods have been used tions were stored protected from the light at −20 C and these were
to quantify the unbound AMB fraction. In addition, these methods stable for at least seven days.
are not sufficiently sensitivity to measure the concentration of the The intermediate calibration solutions were used to prepare cal-
unbound fraction. ibration curves in plasma at concentrations of 200, 180, 50, 5, 1, 0.5,
The aim of this study was to develop a RED based assay 0.2 and 0.1 ng/mL and in plasma ultrafiltrate at concentrations of
followed by a solid phase extraction (SPE) ultra performance liq- 10, 9, 5, 4, 1, 0.5, 0.2 and 0.1 ng/mL. The calibration curves were
uid chromatography tandem mass spectrometry (UPLC–MS/MS) freshly prepared on the day of sample processing.
to measure the unbound AMB fraction in human plasma. This Two sets of intermediate QC solutions (stable for at least seven
◦
methodology required to perform the validation of two indepen- days at −20 C) were prepared by diluting of the stock solution
dent methods: (1) determination of total AMB concentration in (1 mg/mL) in acetonitrile: water (1:1, v:v). One set (160 g/mL and
human plasma and (2) determination of unbound AMB concentra- 10000, 8000, 500, 15 and 5 ng/mL) was prepared to evaluate total
tion in human plasma dialysate. Both quantification methods were AMB in plasma. Another set (8000, 500, 400, 100, 15 and 5 ng/mL)
based on the same SPE and UPLC–MS/MS method. was prepared to evaluate AMB in plasma ultrafiltrate. These sets
In this paper we describe a validated bioanalytical assay for were subsequently used to prepare QC samples (dilution quality
the quantification of AMB in plasma and dialysate. In addition, we control, upper limit of quantification (ULOQ), high quality control
present data from a pharmacokinetics study in which AMB was (HQC), medium quality control (MQC), low quality control (LQC)
administrated to healthy volunteers via intravenous infusion. This and lower limit of quantification (LLOQ)) in plasma at concen-
proposed new route of administration for AMB could be useful in trations of 3200, 200, 160, 10, 0.3 and 0.1 ng/mL and in plasma
situations in which the normal oral administration route is not ultrafiltrate at concentrations of 160, 10, 8, 2, 0.3 and 0.1 ng/mL,
appropriate. respectively. QC samples were stored in propylene tubes and frozen
◦
at -80 C until use, except those used for to evaluate the long term
◦
stability at −20 C. These QC samples were analysed following one
2. Experimental freeze thaw cycle to represent an analytical sample undergoing
initial analysis. The dilution QC samples were diluted with blank
2.1. Chemicals and reagents matrix before processing (factor dilution 1 in 20 for plasma and 1
in 10 for plasma ultrafiltrate).
2
AMB, chemically known as (2S)-2-[(4,6-dimethylpyrimidin- For the IS, a 1.0 mg/mL stock solution of [ H10]-AMB was pre-
2-yl)oxy]-3-methoxy-3,3-diphenyl-propanoic acid (>99% purity), pared in dimetyl sulphoxide and stored protected from the light
◦
was provided by TLC Pharmachem (Vaughan, Canada) and at −20 C; this solution was stable for at least forty days. An inter-
2
the internal standard (IS) [ H10]-ambrisentan (>99% purity) by mediate working solution (1 g/mL) of IS was prepared by further
Alsachim (Strasbourg, France). Dimetyl sulphoxide was obtained dilution with acetonitrile:water (1:1, v:v); this solution was stable
◦
from Sigma-Aldrich (St. Louis, MO, USA). Phosphate-buffered saline at −20 C for at least seven days. A working solution of IS (1 ng/mL)
(dialysis buffer containing 0.1 M sodium phosphate and 0.15 M in 2% formic acid was prepared just prior to use and was not stored.
sodium chloride; pH 7.2), methanol, acetonitrile and formic acid This solution was used for the plasma assay and for the dialysate
were provided by Thermo Fisher Scientific (Waltham, MA, USA). assay.
All mobile phase solvents were LC/MS grade. Milli-Q water was
obtained in-house using a Millipore system (Bedford, MA, USA). 2.3. Extraction in plasma and plasma ultrafiltrate or dialysed
Two pools of drug-free (blank) human plasma and blank human
plasma ultrafiltrate, with K2-EDTA as the anticoagulant, were pur- For determination of total and unbound AMB fraction, 100 L
chased from Sera-Lab Ltd. (Sussex, England). These pools were of plasma or plasma ultrafiltrate/dialysate (calibration standards,
S. Garcia-Martínez et al. / Journal of Pharmaceutical and Biomedical Analysis 150 (2018) 427–435 429
QC samples, zeros, blank samples or real samples) were mixed ultrafiltrate sample processed without AMB and but with IS) and
for 10 min with 400 L IS solution (1 ng/mL) in the wells of a 96- at least one “blank” sample (plasma or plasma ultrafiltrate sample
deep-well polypropylene plate. After the addition of the IS, plasma processed without AMB and without IS) after the highest calibra-
samples were processed by SPE using Oasis HLB 96-well plates tion sample. In the case of “accurate and precision” batches six
(30 mg sorbent per well, 30 m particle size; Waters Corp., Mil- replicates of each QC sample (LLOQ, LQC, MQC, HQC and ULOQ)
ford, MA, USA). The extraction plates were first conditioned with were included.
200 L of methanol and then equilibrated with 200 L of water.
Subsequently, the diluted samples with the IS (500 L) were loaded 2.5.1. Selectivity
to the SPE wells. The SPE wells were then washed with 200 L of The selectivity in both methods (plasma or plasma ultrafiltrate)
5% methanol in water and finally each well was eluted under vac- was determined using drug-free plasma or plasma ultrafiltrate
uum with 100 L of methanol and the eluents were collected into from six different individuals (including four non-haemolysed and
a clean 96-well collection plate. The wells of the collection plate two haemolysed samples). Haemolysed samples were prepared by
◦
were subsequently dried under nitrogen flow at 40 C. The result- adding haemolysed whole blood to the blank samples to produce
ing residues were reconstituted each with 50 L of 0.1% formic acid a 2% haemolysed whole blood solution. The selectivity was tested
in acetonitrile:water (2:8, v:v) and mixed for 10 min 5 L of sample both with and without the addition of IS. The selectivity samples
were injected onto the UPLC–MS/MS system. were extracted as described in Section 2.3 and analysed using the
method described in Section 2.4. The selectivity was evaluated by
2.4. UPLC–MS/MS conditions comparing the area quantified on the area of the chromatogram
corresponding to the analyte and IS in the selectivity samples to
Chromatographic separation was carried out using an Agilent the mean area of the LLOQ samples (six replicates).
(Palo Alto, CA, USA) UPLC system consisting of an Agilent 1290
binary pump, an Agilent 1290 infinity standard autosampler and 2.5.2. Matrix effect on ionization
an Agilent 1290 infinity thermostated column compartment. The The effect of matrix was evaluated for AMB in plasma and plasma
◦ ◦
cooled autosampler was set to 4 C and the column oven to 30 C. ultrafiltrate from six different individuals. Each set of samples
×
A Kinetex EVO C18 column (50 2.1 mm, 1.7 m) provided by Phe- included: non-haemolysed blank samples (four plasma samples)
nomenex (Torrance, USA) was used for separation of AMB. A binary and haemolysed blank samples (two samples). For each method,
gradient method was developed using solvent A (0.1% formic acid two sets of these blank matrix samples were processed as described
in water) and solvent B (0.1% formic acid in acetonitrile). The gra- in Section 2.3 and the final extracts, before injecting into the UPLC
dient was as follows: from 80 to 5% A in 3.5 min, constant at 5% system, were spiked with AMB and IS equivalent in concentra-
A for 0.5 min, back to 80% A in 0.1 min, and constant at 80% A for tion to that which would have be found in a sample assuming
0.9 min. The flow rate was 0.25 mL/min. The injection volume was 100% recovery in the final extract. One set of blank matrix was
set at 5 L. reconstituted with the low QC spiking solution. Another set was
The UPLC system was coupled to an Agilent 6495 triple reconstituted with a high QC spiking solution. Both spiking solu-
quadrupole mass spectrometer for the detection of the AMB and tions were prepared in 0.1% formic acid in acetonitrile:water (2:8,
2
internal standard [ H10]-AMB. The mass spectrometer was oper- v:v). To determine matrix effect, peaks areas of blank samples
ated in electrospray positive mode (ESI + ). Data was acquired using spiked with the spiking solutions were compared to peak areas of
the following settings: capillary voltage 6000 V, drying gas tem- six spiked solutions (without extracted matrix) at corresponding
◦
perature 125 C, gas flow 20 L/h, nebulizer pressure 60 psi, sheath concentrations in solution (0.1% formic acid in acetonitrile:water
◦
gas temperature 300 C, sheath gas flow 8 L/h, high pressure ion (2:8, v:v), representing 100% (no matrix effect).
funnel RF voltage 100 V and low pressure ion funnel RF voltage The matrix factor (MF) was used to assess relative matrix effects.
2
60 V. Detection of AMB and [ H10]-AMB was achieved by multiple
reaction monitoring (MRM) at m/z transitions of 379.1 > 303.2 and
Area of analyte in the presence of matrix
=
389.2 > 313.2, respectively. The dwell time was set to 200 ms, the MF
Mean area of analyte in the spiking solution
fragmentator 380 V and the collision energy to 8 V.
Agilent MassHunter data acquisition for triple quadruple mass
Area of IS in the presence of matrix
spectrometer (version B 07.01) was used for data acquisition and ISmatrixfactor =
Mean area of IS in the spiking solution
Agilent MassHunter quantitative software (version B 07.01) was
used for data processing.
Analyte matrix factor
IS normalized matrix factor =
IS matrix factor
2.5. Specific assay validation
To quantify the unbound AMB fraction in human plasma was 2.5.3. Recovery
necessary to validate two independent bioanalytical methods. One The recovery of AMB in plasma and plasma ultrafiltrate was
a method for the determination of total AMB in plasma and the evaluated by processing LQC, MQC and HQC samples (six replicates
other a method to quantify AMB in dialysed plasma. Both meth- of each) in the normal manner. The peak areas obtained from these
ods were validated according to the FDA and EMA guidelines for samples were then compared with the peak areas obtained from
validation of bioanalytical assays [19,20] for linearity, sensitivity, processed blank matrix samples that had been spiked post extrac-
accuracy and precision, selectivity, matrix effect, recovery, dilution tion to equivalent levels assuming 100% recovery (six replicates per
integrity and stability. Both methods were validated by analysis level).
of QC samples prepared as previously described and included in
an analytical batch. Each analytical batch was performed on a 96- 2.5.4. The stability of AMB in different conditions
well plate and included: conditioning samples (at least eight), a 2.5.4.1. Whole blood stability. Whole blood (using K2-EDTA as the
calibration curve (in the range of 0.1–200 ng/mL for plasma or anticoagulant) from an individual human was spiked with AMB
0.1–10 ng/mL for plasma ultrafiltrate), one set of duplicate QC sam- (100 ng/mL). Aliquots of the whole blood were centrifuged after
◦
ples in plasma or plasma ultrafiltrate at three concentration levels 0, 30, 60, 90 and 120 min post equilibration (15 min at 37 C) to
(LQC, MQC and HQC), at least one “zero” sample (plasma or plasma produce plasma, which was then frozen prior to analysis. This was
430 S. Garcia-Martínez et al. / Journal of Pharmaceutical and Biomedical Analysis 150 (2018) 427–435
Table 1
were extracted (Section 2.3) and finally analysed by UPLC–MS/MS
Flow rages of ambrisentan for groups A and B.
(Section 2.4).
Time (hours) Group A Flow Group B Flow RED was conducted to obtain the unbound fraction of AMB in
rate (g/h) rate (g/h)
plasma samples. The single-use plate RED inserts contained dialy-
0–4 75 300 sis membrane with a molecular weight cut-off of approximately
4–12 44 200 8000 Da (ThermoFisher Scientific, Waltham, MA, USA). The pro-
12–16 100 500
cedure for RED was as follows. Aliquots of plasma (300 L) and
16–24 75 75
dialysis buffer (buffer solution similar to plasma) (500 L) were
added to sample chamber and buffer chamber, respectively on the
RED device. Using the appropriate amount of buffer is essential to
performed at room temperature. The derived plasma samples were
maintain the liquid level in both chambers. The loaded dialysis plate
analysed in triplicate and the different time points were compared.
was sealed and placed on a shaker at approximately 250 rpm and
◦
◦ ◦ incubated at physiological temperature (37 C) for 4 h. Both the vol-
−
2.5.4.2. Bench top, long term stability at 20 C and −80 C and freeze
ume ratio and the incubation conditions established in the dialysis
thaw stability. The stability of AMB in human K2-EDTA plasma or
process were performed according to the manufacturer of the RED
plasma ultrafiltrate was investigated under different conditions. ◦ ◦
device. The dialysate was stored at −80 C ± 10 C prior to assay.
Two levels of QC samples (LQC and HQC; six replicates per level)
The dialysate samples (100 L) were quantified in a batch contain-
prepared in matrix were used to evaluate the stability of AMB under
ing: calibration samples (in the range of 0.1–200 ng/mL), duplicate
different conditions. The conditions evaluated were: (1) stability at
◦ ◦ QC samples at three concentration levels (0.3, 10 and 160 ng/mL),
room temperature for 24 h, (2) stability at −20 C ± 5 C for 35 days,
◦ ◦ zero samples and a blank sample. Then, the samples were extracted
(3) stability at −80 C ± 10 C for 35 days and (4) stability following
and analysed by UPLC–MS/MS in the same manner as the plasma
four freeze thaw cycles. In this case, the QC samples were frozen at
− ◦ ◦ samples.
80 C ± 10 C for a minimum of 12 h and then defrosted at room
In this study, incurred sample reanalysis (ISR) was performed
temperature for 2 h.
on 40 samples for the total concentration of AMB and 16 samples
for the free concentration. The samples were selected using the
2.5.4.3. Processed sample integrity. The stability of processed sam-
following criteria: (1) samples were above LLOQ, (2) samples cov-
ples (prepared using both methods; plasma and ultrafiltrate) was
ered the whole range of doses included in the study, (3) samples
assessed by the injection of an “accuracy and precision” batch after
◦ were reanalysed within the stability period and (4) samples were
5 days storage at nominally 10 C.
from the maximum plasma concentration and from the elimination
phase.
2.5.4.4. Re-injection stability. The stability of AMB in re-injected
samples was assessed by the re-injection of a previously accepted
“accuracy and precision” batch following three days storage under
3. Results and discussion
the same conditions as expected in the autosampler (nominal ◦
10 C). 2
3.1. Characterization of ambrisentan and [ H10]-ambrisentan by
UPLC–MS/MS
2.6. Application of the assay
The chromatographic conditions described in this analytical
In this study, the objective was to characterise the pharmacoki- procedure were achieved after investigating different columns and
netic profile of total and unbound AMB in healthy volunteers when several mobile phases. Symmetrical and sharp peak for AMB was
administered AMB via intravenous. The AMB was administered observed for using an EVO C18 column and a mixture of acetonitrile
over a 24 h period using two different flow rates profiles. A total −water (in both cases with formic acid 0.1%) as mobile phase. Under
of 20 volunteers were included in 2 study groups (group A: n = 10, chromatography conditions the retention times of AMB and the IS
total dose = 1.652 mg of AMB; group B: n = 10, total dose = 7.800 mg). were 2.8 min (Fig. 1). The total run time was only 5 min. According
Both groups were infused with a 5 mg/mL solution but the flow rate to the full-scan ESI (+) mass spectra, the protonated molecule ion
+ + 2
was altered to change the infusion rate. The infusion profiles for [M + H] m/z 379.1 for AMB and [M + H] m/z 389.2 for [ H10]-AMB
groups A and B are presented in Table 1. were selected as the precursor ions to obtain the product ions. The
The study complied with the recommendations of the 18th most sensitive mass transition from the precursor ions to the prod-
World Health Congress (Helsinki) [21] and all applicable amend- uct ion were m/z 379.1 → 303.2 and m/z 389.2 → 313.2. AMB in the
ments. The study also complied with the international and local ESI + mode shows a selective loss of the CH3OH followed by CO2,
applicable laws, regulations and guidelines and was approved by resulting in the product ion m/z 303 [14] (see Fig. 1).
the corresponding regional ethics committee.
Blood samples, using K2-EDTA as the anticoagulant, were taken
from the volunteers at the following time points: pre-dose (–1 h), 2, 3.2. Validation
4, 6, 9, 12, 14, 16, 18, 21, 24, 26, 28, 32, 36, 42, 48 and 72 h post-dose
for the quantification of total AMB. The unbound plasma concen- In this study, we develop a RED assay followed by SPE and
trations of AMB were determined at the following time points: UPLC–MS/MS to measure the unbound AMB fraction in human
pre-dose (–1 h), 4, 12, 16, 24 and 36 h post-dose. Following the plasma. This methodology required the validation of two inde-
collection, whole blood samples were centrifuged (at 2000g for pendent methods: (1) determination of total AMB concentration
ten min) and, once centrifuged, all plasma samples were frozen in human plasma and (2) determination of unbound AMB concen-
◦ ◦
at −80 C ± 10 C prior to assay. tration in human plasma ultrafiltrate. The Table 2 summarizes the
Initially, the determination of the total concentration of AMB results of both methods.
was carried out. Plasma samples (100 L) were included in a batch Acceptance criteria were established for the overall analytical
containing: calibration samples (in the range of 0.1–200 ng/mL), run. For the calibration samples were that the back-calculated con-
duplicate QC samples at three concentration levels (0.3, 10 and centrations of 75% (8 out of 10) standards should fall within ±15%
160 ng/mL), zero samples and a blank sample. Then, the samples (± 0% at the LLOQ) of the nominal value. Values falling outside these
S. Garcia-Martínez et al. / Journal of Pharmaceutical and Biomedical Analysis 150 (2018) 427–435 431
Fig. 1. Representative multiple reaction monitoring chromatograms of ambrisentan (m/z 379 > 303) and its internal standard (m/z 389 > 313) at the lower limit of quantification
(LLOQ; 0.1 ng/mL) in human plasma (A) and plasma ultrafiltrate (B). Representative chromatograms of blank plasma (C) and plasma ultrafiltrate (D) without internal standard.
The proposed pattern of fragmentation for ambrisentan is shown on top.
limits were excluded. In addition the coefficient of determination 3.2.1. Evaluation of the linearity of the calibration curve
2
(r ) of the calibration curve should be >0.9800. Calibration curves (in the range 0.1–200 ng/mL for plasma and
Acceptance criteria for the QC samples were that the measured 0.1–10 ng/mL for plasma ultrafiltrate) were prepared in blank
concentrations of at least 67% (4 out of 6) of the QC samples should matrix for each batch (a total seven calibration curves by each
fall within ±15% of the nominal value. Two out of six QC sam- method). All the calibration curves in this study were freshly pre-
ples could fall outside this limit as long as they were both not at pared on the day of sample processing. Each batch contained at
the same nominal concentration. Additionally, acceptance crite- least one zero sample and at least one blank sample. These zero
ria for the individual assessments were also established. These are and blank samples were used for evaluating possible interferences,
described in detail in each Section. without including them in the calibration curve. The calibration
2
models were assessed to be linear regression with 1/ weighting
432 S. Garcia-Martínez et al. / Journal of Pharmaceutical and Biomedical Analysis 150 (2018) 427–435
Table 2
the previously fixed (1.0 ng/mL, [15]) in biological samples with a
Summary of method validation data for the measurement of total ambrisentan (in
similar run time.
plasma) and free ambrisentan (in plasma ultrafiltrate).
The ULOQ or the highest concentration that can be measured
Species Human
with acceptable precision and accuracy were also evaluated. The
Matrix K2-EDTA plasma and plasma
ULOQ of these assays was 200 ng/mL for plasma and 10 ng/mL for
ultrafiltrate
plasma ultrafiltrate. In both methods, acceptable values of precision
Analyte Ambrisentan
2
≤ ±
Internal standard [ H10]-ambrisentan (CV% 15%) and accuracy (bias within 15%) at the ULOQ were
Calibration range 0.1–200 ng/mL (plasma) and 0.1– demonstrated (Table 3).
10 ng/mL (plasma ultrafiltrate)
The carryover was assessed by injecting four blank plasma sam-
Lower limit of quantification 0.1 ng/mL
ples after the injection of the highest calibration sample. The area
Sample volume 100 L
responses in the blank samples were compared to the mean area
Method outline SPE followed by detection with
UPLC–MS/MS response of the LLOQ QCs in the same batch. No carryover was
Calibration model Linear regression
detected for the AMB or IS in plasma and plasma ultrafiltrate, with
2
Weighting factor 1/
no peaks found with response >20% of the corresponding mean
Within batch accuracy and Requirements fulfilled
precision response of the analyte in the LLOQ QCs, and no peaks found with
Between batch accuracy and Requirements fulfilled response >5% of the corresponding mean response of the IS in the
precision LLOQ QCs.
Selectivity Negligible
Matrix effect on ionization Negligible
3.2.3. Selectivity
Recovery ambrisentan 69.4% (plasma) 77.5% (plasma
ultrafiltrate) The selectivity in both methods (plasma or ultrafiltrate) was
Carryover Negligible
evaluated by comparing chromatograms of blank samples and
Dilution integrity Determinations after 20 fold dilution
plasma or plasma ultrafiltrate spiked with AMB at LLOQ level. No
(in plasma) and 10 (ultrafiltrate)
◦ ◦ endogenous components interfering with the detection of AMB
Stock solution stability At least 18 days at −20 C ± 5 C
(1 mg/mL) were found in the chromatograms of blank samples. Representative
◦ ◦
− ±
Working solution stability At least 7 days at 20 C 5 C MRM chromatograms of blank plasma and ultrafiltrate without the
◦ ◦
Isotopic purity of the internal At least 40 days at −20 C ± 5 C
addition of the IS are included in Fig. 1C and D, respectively.
standard (1 mg/mL)
◦ ◦
Internal standard working At least 7 days at −20 C ± 5 C
solution (1 g/mL) stability 3.2.4. Matrix effect on ionization
Long term stability at At least 35 days
The determination of ionization suppression (or enhancement)
− ◦ ◦
80 C ± 10 C
by matrix extracts is important as this phenomenon can severely
Long term stability at At least 35 days
◦ ◦ affect the sensitivity and robustness of the analytical method. Sam-
−20 C ± 5 C
Processed samples integrity At least 5 days ples of matrix from different sources were analysed. The matrix
Re-injection stability At least 3 days effect was evaluated by comparing the peak area ratio in the post-
Freeze thaw stability At least 4 cycles
extracted plasma samples at two concentration levels (LQC and
Bench top stability At least 24 h at room temperature
HQC) with neat standard solutions at the same concentration levels.
Whole Blood Stability At least 120 min at room temperature
The ion suppression or enhancement of the signal in terms of IS-
normalized matrix factor was found to be at LQC level 0.954 ± 3.0%
2 and 0.992 ± 0.6% and at HQC level 0.919 ± 4.7% and 0.988 ± 0.6%
and the r for each curve were greater than 0.9800. For both meth-
for AMB in plasma and plasma ultrafiltrate, respectively. Therefore,
ods, the results from the calibration standards were found to be
acceptable precision (CV of IS-normalized matrix factor ≤15%) was
precise (coefficient of variation or CV ≤ 6.8%) and acceptable (bias
demonstrated for both methods and the results indicated that any
within ±4.0%).
matrix effect on the ionization of AMB was consistent and did not
compromise the quantification of AMB in plasma or ultrafiltrate
3.2.2. Accuracy and precision, sensitivity, lower and upper limit
plasma.
of quantification and carryover
A total of three “accuracy and precision” batches for method
3.2.5. Recovery
were processed. Each “accuracy and precision” batch included: con-
The overall recovery for AMB and the IS was calculated by com-
ditioning samples, a calibration curve in matrix, at least one zero
paring the peak-area ratios of spiked samples before and after the
sample, four blank samples after the highest calibration sample and
extraction in different lots of plasma and plasma ultrafiltrate at
six replicates of each QC sample (LLOQ, LQC, MQC, HQC and ULOQ)
three QC levels (LQC, MQC, and HQC). The mean recovery of AMB
prepared in matrix. The results obtained for these QC samples were
was 69.4% and the IS normalised recovery had a CV% of 5.6%.
used to calculate the within and between-batch accuracy and pre-
In plasma ultrafiltrate the mean recovery of AMB was high
cision. Acceptable accuracy (bias within ±15% or ±20% at the LLOQ)
(77.5%) and consistent (CV% of 4.8%) and the recovery of the IS was
and precision (CV ≤ 15% or ≤20% at the LLOQ) for the determination
similar to AMB.
of total AMB in plasma and AMB in dialysate plasma were demon-
strated. The results from the within and between-batch accuracy
3.2.6. Dilution integrity
and precision measurements are reported Table 3.
The samples were analysed and the results (corrected for dilu-
These “accuracy and precision” batches were also used to eval-
tion) compared with the nominal concentration. The ability to
uate the limits of quantification of both methods. The sensitivity
perform accurate determinations after a 20-fold dilution for total
or the lowest concentration of a sample (LLOQ) that can be quan-
AMB in plasma and 10-fold dilution for AMB in plasma ultrafiltrate
tified with acceptable precision and accuracy were evaluated in
were demonstrated after dilution with blank matrix of six repli-
both methods. Acceptable precision (CV ≤ 20%) and accuracy (bias
cates of a dilution QC sample (3200 ng/mL in plasma and 160 ng/mL
within ±20%) at the LLOQ was demonstrated for AMB in plasma
in plasma ultrafiltrate). The results obtained were precise (3.6%
and for AMB in plasma ultrafiltrate (Table 3). Representative chro-
and 6.7%) and accurate (6% and -4% bias) for plasma and plasma
matograms from the LLOQ level in plasma and ultrafiltrate plasma
ultrafiltrate, respectively.
are shown in Fig. 1. The achieve LLOQ (0.1 ng/mL) was lower than
S. Garcia-Martínez et al. / Journal of Pharmaceutical and Biomedical Analysis 150 (2018) 427–435 433
Table 3
Within and between-batch precision and accuracy in quality control samples prepared in plasma and plasma ultrafiltrate.
Plasma
LLOQ LQC MQC HQC ULOQ
(0.1 ng/mL) (0.3 ng/mL) (10 ng/mL) (160 ng/mL) (200 ng/mL)
◦
Batch N 1
Within-batch mean (ng/mL) 0.103 0.291 10.0 173 210
Within-batch CV (%) 15.1 4.9 1.6 11.2 1.5
−
Within-batch bias (%) 3 3 0 8 5
n 6 6 6 6 6
◦
Batch N 2
Within-batch mean (ng/mL) 0.111 0.323 10.2 168 213
Within-batch CV (%) 2.3 2.0 1.0 2.3 2.2
Within-batch bias (%) 11 8 2 5 7
n 5* 6 6 6 6
◦
Batch N 3
Within-batch mean (ng/mL) 0.100 0.283 9.69 157 209
Within-batch CV (%) 3.8 1.7 1.2 2.6 2.3
Within-batch bias (%) 0 −6 −3 −2 5
n 6 6 6 5* 6
Between-batch mean (ng/mL) 0.104 0.299 10 167 210
Between-batch CV (%) 9.9 6.7 2.3 7.8 2.1
Between-batch bias (%) 4 0 −1 4 5
n 17 18 18 17 18
Plasma Ultrafiltrate
LLOQ LQC MQC HQC ULOQ
(0.1 ng/mL) (0.3 ng/mL) (2 ng/mL) (8 ng/mL) (10 ng/mL)
◦
Batch N 1
Within-batch mean (ng/mL) 0.0898 0.282 1.80 6.92 8.79
Within-batch CV (%) 2.1 3.8 2.3 2.7 3.7
Within-batch bias (%) −10 −6 −10 −14 −12
n 6 6 6 6 6
◦
Batch N 2
Within-batch mean (ng/mL) 0.0935 0.269 1.86 7.24 9.74
Within-batch CV (%) 0.3 0.7 4.0 1.9 9.1
Within-batch bias (%) −7 −10 −7 −10 −3
n 6 6 6 6 6
◦
Batch N 3
Within-batch mean (ng/mL) 0.0886 0.270 1.85 7.37 9.00
Within-batch CV (%) 4.7 2.3 1.0 1.5 1.8
Within-batch bias (%) −11 −10 −8 −8 −10
n 6 6 6 6 6
Between-batch mean (ng/mL) 0.0906 0.273 1.83 7.18 9.18
Between-batch CV (%) 3.6 3.3 2.9 3.3 7.3
Between-batch bias (%) −9 −9 −9 −10 −8
*A QC level was excluded according to Qtest.
3.2.7. AMB in matrix: stability determinations 3.2.7.2. Bench top stability. The stability of AMB in matrix (plasma
Drug stability in a biological fluid is a function of the storage or plasma ultrafiltrate) at room temperature for 24 h was tested
conditions, the chemical properties of the drug, the matrix, and and demonstrated using LQC and HQC samples (six replicates per
the container system. The stability of the analyte should be evalu- level). The stability of AMB under these conditions was demon-
ated during sample collection and handling, after long-term (frozen strated for both methods. The precision (CV%) was less than 3.9%
at the intended storage temperature) and short-term (bench top, and the accuracy (expressed as bias) when compared with the nom-
room temperature) storage, and after going through freeze and inal concentration was less than ±14%.
thaw cycles and the analytical process. Conditions used in stabil-
◦ ◦
− −
ity experiments should reflect situations likely to be encountered 3.2.7.3. Long term stability at 20 C and 80 C. The stabilities of
during actual sample handling and analysis. AMB when stored LQC and HQC (prepared with plasma or plasma
◦ ◦ ◦ ◦
ultrafiltrate) at −20 C ± 5 C or −80 C ± 10 C were assessed for
3.2.7.1. Whole blood stability. The stability of AMB in whole blood 35 days. In both cases, the stability of AMB under these conditions
was assessed to find possible problems during the collection and was demonstrated (% CV was less than 7.8% and the bias was less
processing of samples. Aliquots of whole blood spiked with AMB than ±11%).
were centrifuged after 0, 30, 60, 90 and 120 min after the equi-
libration period. The response of the samples left for 30, 60, 90 3.2.7.4. Freeze thaw stability. The stability of AMB in human K2-
and 120 min were within ±6% of the sample after the equilibration EDTA plasma and plasma ultrafiltrate after four freeze thaw cycles
period. Therefore, the stability of AMB in human whole blood for was also demonstrated (% CV ranged from 1.1% to 4.1% and bias
−
120 min was demonstrated at room temperature. from 13% to 1%).
434 S. Garcia-Martínez et al. / Journal of Pharmaceutical and Biomedical Analysis 150 (2018) 427–435
360 Rate 1 Rate 2 Rate 3 Rate 4 340 (4 h) (8 h) (4 h) (8 h)
320
300 Group A 280
260 Group B
240
220
200
180
160
140
120
100
80
60
40
20
0 04812 16 20 24 28 32 36 40 44 48 52 56 60 64 68 72 Time (hours)
4.0 Rate 1 Rate 2 Rate 3 Rate 4 (4 h) (8 h) (4 h) (8 h) 3.5 Group A
Group B 3.0
2.5
2.0
1.5
1.0
0.5
0.0
-0.5
-1.0
0510 15 20 25 30 35
Fig. 2. Mean plasma concentration–time curves of total ambrisentan (A) and unbound ambrisentan (B) in human plasma after intravenous administration (Group A :
total dose = 1.652 mg; Group B : total dose = 7.800 mg). The continuous infusion rates used in each group are included in Table 1. Each point represents the mean ± S.D.
S. Garcia-Martínez et al. / Journal of Pharmaceutical and Biomedical Analysis 150 (2018) 427–435 435
3.2.7.5. Processed sample integrity. Processed sample integrity was References
◦
demonstrated after at least 5 days at nominally 10 C for both meth-
[1] A.H. Chester, M.H. Yacoub, The role of endothelin-1 in pulmonary arterial
ods. The precision of the QC samples was ≤ 4.9% (% CV) at all levels.
hypertension, Glob. Cardiol. Sci. Pract. 2 (2014) 62–78.
The within batch accuracy (expressed as bias) was ranged from
[2] US Food and Drug Administration Drug approval package, Letairis
−
14% to 3%. (ambrisentan NDA #022081). Medical Review, Part 1, 2012, Available from:
http://www.accessdata.fda.gov/drugsatfda docs/nda/2007/022081s000 MedR P1.
pdf. Accessed November 11, 2012.
3.2.7.6. Re-injection stability. The stability of AMB in processed
[3] S.M. Elshaboury, J.R. Anderson, Ambrisentan for the treatment of pulmonary
samples was assessed by re-injection of “accurate and precision” arterial hypertension: improving outcomes, Patient Prefer. Adher. 7 (2013)
401–409.
batches (in plasma or plasma ultrafiltrate) previously accepted
[4] G.L. Trainor, The importance of plasma protein binding in drug discovery,
containing calibration standards and QC samples (six replicates
Expert Opin. Drug Discov. 2 (2007) 51–64.
each of LLOQ, LQC, MQC, HQC and ULOQ) after three days stored [5] S. Schmidt, D. Gonzalez, H. Derendorf, Significance of protein binding in
◦
under the same conditions as expected in the auto sampler (10 C). pharmacokinetics and pharmacodynamics, J. Pharm. Sci. 99 (2010)
1107–1122.
Re-injection stability was demonstrated after at least 3 days at
◦ [6] K. Vuignier, J. Schappler, J.L. Veuthey, P.A. Carrupt, S. Martel, Drug-protein
nominally 10 C (% CV was less than 13% and bias less than ±10%).
binding: a critical review of analytical tools, Anal. Bioanal. Chem. 398 (2010)
53–66.
[7] S.M. Illamola, D. Hirt, J.M. Tréluyer, S. Urien, S. Benaboud, Challenges
3.3. Application of the method
regarding analysis of unbound fraction of highly bound protein antiretroviral
drugs in several biological matrices: lack of harmonisation and guidelines,
Here, we described the reiterated stepwise equilibrium dialysis Drug Discov. Today 20 (2015) 466–474.
[8] H. Rakhila, T. Rozek, A. Hopkins, S. Proudman, L. Cleland, M. James, M. Wiese,
method, its validation and use to determine the extent of binding
Quantitation of total and free teriflunomide (A77 1726) in human plasma by
of AMB to plasma proteins.
LC-MS/MS, J. Pharm. Biomed. Anal. 55 (2011) 325–331.
The use of the RED device for the measurement of plasma pro- [9] Y. Deng, H. Wong, R.A. Graham, W. Liu, H.S. Shen, Y. Shi, L. Wang, M. Meng, V.
Malhi, X. Ding, B. Dean, Determination of unbound vismodegib (GDC-0449)
tein binding provided a robust and accurate data with shorter
concentration in human plasma using rapid equilibrium dialysis followed by
preparation and dialysis times. The high membrane surface-to-
solid phase extraction and high-performance liquid chromatography coupled
volume ratio allowed rapid dialysis [22]. It is interesting to note that to mass spectrometry, J. Chromatogr. B Analyt. Technol. Biomed. Life Sci. 879
(2011) 2119–2126.
an excessively long incubation might promote compound insta-
[10] Y. Mano, K. Kusano, A validated LC-MS/MS method of total and unbound
bility or result in a volume increase of the plasma sample from
lenvatinib quantification in human serum for protein binding studies by
hydrostatic pressure. After the equilibrium was reached negligible equilibrium dialysis, J. Pharm. Biomed. Anal. 114 (2015) 82–87.
volume shifts were observed, negating the need to assess volume [11] M. Dousa,ˇ P. Gibala, Rapid determination of ambrisentan enantiomers by
shifts. enantioselective liquid chromatography using cellulose-based chiral
stationary phase in reverse phase mode, J. Sep. Sci. 35 (2012) 798–803.
The utility of this assay for the quantitative of total and unbound
[12] M.B.V. Narayana, K.B. Chandrasekhar, B.M. Rao, A validated specific
AMB in human plasma after intravenous administration dose stability-indicating RP-HPLC assay method for ambrisentan and its related
substances, J. Chromatogr. Sci. 52 (2014) 818–825.
(group A: total dose = 1.652 mg; group B: total dose = 7.800 mg) was
[13] J.K. Patel, N.K. Patel, Stability-indicating RP-HPLC method for the
demonstrated. A total of 360 plasma samples (20 subjects, 18 sam-
determination of ambrisentan and tadalafil in pharmaceutical dosage form,
ples per subject) were analysed for the determination of the total Sci. Pharm. 82 (2014) 749–763.
[14] N.R. Ramisetti, R. Kuntamukkala, LC-MS/MS characterization of forced
concentration of AMB in plasma. In addition, a total of 120 plasma
degradation products of ambrisentan: development and validation of a
samples (20 subjects, 6 samples per subject) were analysed for the
stability-indicating RP-HPLC method, New J. Chem. 38 (2014) 3050–3061.
concentration of unbound AMB. The mean plasma concentration- [15] R. Nirogi, V. Kandikere, P. Komarneni, R. Aleti, N. Padala, I. Kalaikadhiban,
LC-ESI–MS/MS method for quantification of ambrisentan in plasma and
time curves for total and unbound AMB are shown in Fig. 2. The
application to rat pharmacokinetic study, Biomed. Chromatogr. 26 (2012)
pre-doses concentrations were not quantifiable ( < 0.1 ng/mL). 1150–1156.
According to these results, unbound AMB represents 0.96% [16] Y. Yokoyama, M. Tomatsuri, H. Hayashi, K. Hirai, Y. Ono, Y. Yamada, K.
Todoroki, T. Toyo’oka, H. Yamada, K. Itoh, Simultaneous microdetermination
(range 0.83–1.1%) of the total level. These values are similar to the
of bosentan, ambrisentan, sildenafil, and tadalafil in plasma using liquid
values found in the literature.
chromatography/tandem mass spectrometry for pediatric patients with
In addition, the results obtained in the ISR met the acceptance pulmonary arterial hypertension, J. Pharm. Biomed. Anal. 89 (2014) 227–232.
criterion (the difference between the mean of two values obtained [17] O.K. Lukram, R. Sharma, High-performance liquid chromatography tandem
mass spectrometry method for quantification of endothelin receptor
was within 20% for 100% of the repeats).
antagonist drug, ambrisentan, in human plasma and its application in a
pharmacokinetic study, Biomed. Chromatogr. 28 (2014) 1147–1155.
4. Conclusions [18] Y. Enderle, L. Witt, H. Wilkens, E. Grünig, W.E. Haefeli, J. Burhenne,
Simultaneous quantification of endothelin receptor antagonists and
phosphodiesterase 5 inhibitors currently used in pulmonary arterial
A novel bioanalytical method, involving a simple and fast RED
hypertension, J. Pharm. Biomed. Anal. 143 (2017) 291–298.
assay followed by SPE–UPLC–MS/MS, for the measurement of the [19] Guidance for Industry, Bioanalytical Method Validation, U.S. Department of
Health and Human Services, Food and Drug Administration Centre for Drug
free AMB fraction in human plasma was successfully developed and
Evaluation and Research (CDER), Centre for Veterinary Medicine (CVM), 2001.
validated. This is the first report on the determination of free AMB
[20] Guideline on bioanalytical method validation, European Medicines Agency,
in plasma. In addition, an associated method for the quantifica- 2011, EMEA/CHMP/EWP/192217/2009.
[21] World Medical Association (WMA), WMA Declaration of Helsinki – Ethical
tion of the bound AMB concentration in plasma was validated with
Principles for Medicinal Research Involving Human Patients, 64th WMA
high sensitivity. The achieved LLOQ was of 0.1 ng/mL. The assay
General Assembly, Fortaleza, Brazil, 2013 (October).
described herein was suitable to characterise the pharmacokinetic [22] N.J. Waters, R. Jones, G. Williams, B. Sohal, Validation of a rapid equilibrium
dialysis approach for the measurement of plasma protein binding, J. Pharm.
profile of total and bound AMB in healthy volunteers when given
Sci. 97 (2008) 4586–4595.
AMB intravenously; a new route of administration. In the future,
this methodology could also have application in the analysis of
bound and free AMB fraction of plasma samples generate from multiple clinical studies.