Anal. Chem.

2000, 72, 4235-4241

Comparison of Packed-Column Supercritical Fluid

Chromatography-Tandem Mass Spectrometry with
Liquid Chromatography-Tandem Mass
Spectrometry for Bioanalytical Determination of
(R)- and (S)-Ketoprofen in Human Plasma Following
Automated 96-Well Solid-Phase Extraction
Steven H. Hoke, II,*,† J. David Pinkston,‡ Ruth E. Bailey,† Suzanne L. Tanguay,† and
Thomas H. Eichhold†
Health Care Research Center, The Procter & Gamble Company, P.O. Box 8006, Mason, Ohio 45040, and Miami Valley
Laboratories, The Procter & Gamble Company, P.O. Box 538707, Cincinnati, Ohio 45253

The popularity of packed-column supercritical fluid, sub- (R,S)-2-(3-Benzoylphenyl)propionic acid (ketoprofen, kt; Figure
critical fluid, and enhanced fluidity liquid chromatogra- 1), is a potent nonsteroidal, anti-inflammatory drug that is com-
phies (pcSFC) for enantiomeric separations has increased monly formulated as a tablet for oral administration or, in some
steadily over the past few years. The addition of a countries, as a topical gel or cream. In both cases, the active is
significant amount (typically 20-95%) of a viscosity formulated as a racemic mixture. It is well known that the (S)
lowering agent, such as carbon dioxide, to the mobile form of kt contains the intrinsic pharmacologic activity,1 and for
phase provides a number of advantages for chiral separa- this reason, it is desirable to determine the plasma concentrations
tions. For example, higher mobile-phase flow rates can of the individual enantiomers. The estimated bioavailability of
often be attained without a concomitant loss in chromato- orally dosed ketoprofen is g92%,2 and the maximum plasma
graphic efficiency since diffusion coefficients, and opti- concentrations reach low-microgram per milliliter levels after a
mum velocities, are typically higher in pcSFC. Ultratrace typical 25-mg dose.2,3 With peak concentrations in that range, LC
enantioselective quantitation of drugs in biomatrixes is with UV detection has been established to provide adequate
an ideal application for these chromatographic attributes. sensitivity for obtaining full pharmacokinetic (PK) curves.4-10
To demonstrate the utility of this approach, a pcSFC
Typical chiral LC-UV methods reported in the literature have
tandem mass spectrometry (pcSFC-MS/MS) method was
analysis times of approximately 10-25 min and lower limits of
compared to a LC-MS/MS method for quantitation of the
quantitation (LLOQ) of 25 ng/mL. Other approaches for bioana-
(R)- and (S)-enantiomers of ketoprofen (kt), a potent
lytical determination of ketoprofen are GC/MS based with
nonsteroidal, anti-inflammatory drug, in human plasma.
reported 10-20-min analysis times.11-15 One of these methods14
After preparation using automated solid-phase extraction
in the 96-well format, kt enantiomers were separated on
(1) Hutt, A. J.; Caldwell, J. Clin. Pharmacokinet. 1984, 9, 371-373.
a Chirex 3005 analytical column using isocratic condi- (2) Jamali, F.; Brocks, D. R. Clin. Pharmacokinet. 1990, 19, 197-217.
tions. Validation data and study sample data from patients (3) Foster, R. T.; Jamali, F.; Russell, A. S.; Alballa, S. R. J. Pharm. Sci. 1988,
dosed with either orally or topically administered keto- 77, 70-73.
(4) Carr, R. A.; Caille, G.; Ngoc, A. H.; Foster, R. T. J. Chromatogr., B: Biomed.
profen were generated using both pcSFC and LC as the Sci. Appl. 1995, 668, 175-181.
chromatographic methods to compare and contrast these (5) Lovlin, R.; Vakily, M.; Jamali, F. J. Chromatogr., B: Biomed. Sci. Appl. 1996,
analytical approaches. Generally, most analytical at- 679, 196-198.
(6) Grubb, N. G.; Rudy, D. W.; Hall, S. D. J. Chromatogr., B: Biomed. Sci. Appl.
tributes, including specificity, linearity, sensitivity, ac- 1996, 678, 237-244.
curacy, precision, and ruggedness, for both of these (7) Boisvert, J.; Caille, G.; McGilveray, I. J.; Qureshi, S. A. J. Chromatogr., B:
methods were comparable with the exception that the Biomed. Sci. Appl. 1997, 690, 189-193.
(8) Yagi, M.; Shibukawa, A.; Nakagawa, T. Chem. Pharm. Bull. 1990, 38, 2513-
pcSFC separation provided a roughly 3-fold reduction in 2517.
analysis time. A 2.3-min pcSFC separation and a 6.5-min (9) Wanwimolruk, S.; Wanwimolruk, S. Z.; Zoest, A. R. J. Liq. Chromatogr.
LC separation provided equivalent, near-baseline-resolved 1991, 14, 3685-3694.
(10) Rifai, N.; Lafi, M.; Sakamoto, M.; Law, T. Ther. Drug Monit. 1997, 19,
peaks, demonstrating a significant time savings for analy- 175-178.
sis of large batch pharmacokinetic samples using pcSFC. (11) Leis, H. J.; Leis, M.; Windischhofer, W. J. Mass Spectrom. 1996, 31, 486-
492.
(12) Alkatheeri, N. A.; Wasfi, I. A.; Lambert, M. J. Vet. Pharmacol. Ther. 1999,
* Corresponding author: (e-mail:) hoke.sh@pg.com.) 22, 127-135.
† Health Care Research Center.
(13) Gonzalez, G.; Ventura, R.; Smith, A. K.; de la Torre, R.; Segura, J. J.
‡ Miami Valley Laboratories.
Chromatogr., A 1996, 719, 251-264.

10.1021/ac000068x CCC: $19.00 © 2000 American Chemical Society Analytical Chemistry, Vol. 72, No. 17, September 1, 2000 4235
Published on Web 07/22/2000
 meric separations has increased steadily over the past few
years.21-26 While these three techniques are different, they are
closely related and require the same instrumentation. All three
will be referred to as “pcSFC” in this report. The addition of a
significant amount, typically 20-95%, of a viscosity lowering agent,
such as carbon dioxide, to the mobile phase provides a number
of advantages for chiral separations.27,28 Lower viscosity allows the
use of higher mobile-phase flow rates (i.e., faster analyses) and/
or longer columns (i.e., higher efficiencies29) than when traditional
liquid mobile phases are used. Compared to liquid chromatogra-
phy, these higher mobile-phase flow rates can often be attained
without a concomitant loss in chromatographic efficiency since
diffusion coefficients, and optimum velocities, are typically higher
with pcSFC. While CO2 itself is relatively nonpolar, mixtures of
CO2 and polar organic solvents retain the polarity and solvating
power of the polar organic until significant levels (40-60%) of CO2
are added.30 These mixtures can often, though not universally,
be used to replace existing chiral LC mobile phases, without loss
of the chiral recognition mechanism of the analyte. However, each
separation must be tested on an individual basis.31
Although rarely performed, the ultratrace quantitation of drugs
Figure 1. Full-scan product ion spectra of (A) ketoprofen and (B)
in biological fluids is an ideal application of this separation
SIL ketoprofen. The asterisk denotes the chiral center.
technique when coupled with tandem mass spectrometric detec-
tion. Previous reports have described the development and survey
is stereoselective but requires derivatization and has a LLOQ of application of the pneumatically assisted electrospray source for
1 ng/mL for each enantiomer. pcSFC-MS/MS that was used in the present study.32,33 Here, a
Because the circulating levels of kt after topical administration thorough assessment of the “real world” utility of the pcSFC-
are much lower than those observed following an oral dose, the MS/MS technique for high-throughput bioanalytical quantitation
sensitivity of the methodologies listed above is not adequate to of drugs is provided, using enantiospecific quantitation of keto-
define PK curves following a topical dose. For this reason, a profen as an example. This is accomplished by validation and
method was developed to lower the limit of quantitation using application of a pcSFC-MS/MS method for quantitation of the
liquid chromatography coupled with tandem quadrupole mass (R)- and (S)-enantiomers of ketoprofen in human plasma and
spectrometry (LC-MS/MS).16 The resulting method has a LLOQ comparison of the results of this method to results generated with
of 50 pg/mL per enantiomer and relatively short analysis time of an earlier LC-MS/MS method.
6.5 min/sample. In addition, automation of the sample preparation
was performed using robotic solid-phase extraction (SPE) in the EXPERIMENTAL SECTION
96-well plate format, which allowed the processing of a 96-well Chemicals and Reagents. (R,S)-Ketoprofen was purchased
plate every 50 min. Though the LC-MS/MS methodology was from the United States Pharmacopeial Convention (Rockville, MD)
quite reliable, accurate, and precise, it possessed the drawback while the internal standard, [13C1,2H3]-(R,S)-ketoprofen (SIL kt)
that the separation still required a 6.5-min analysis time to yield was synthesized at P&G Pharmaceuticals (Norwich, NY). Blank
adequate resolution of the enantiomers. While that throughput human plasma was obtained from volunteers at Procter & Gamble
represented an advance over existing methodologies, it was still (Mason, OH) or from Golden West Biologicals (Temecula, CA).
much slower than analysis times typically used for racemic
(21) Phinney, K. W.; Sander, L. C.; Wise, S. A. Anal. Chem. 1998, 70, 2331-
bioanalytical quantitation which range from a few minutes to less 2335.
than 1 min/analysis.17-20 (22) Maftouh, M. Spectra Anal. 1997, 26, 25-28.
The popularity of packed-column supercritical fluid, subcritical (23) Smith, R. In Supercritical Fluid Chromatography with Packed Columns,
Techniques and Applications; Anton, K., Berger, C., Eds.; Chromatography
fluid, and enhanced fluidity liquid chromatographies for enantio- Science Series 75; Marcel Dekker: New York, 1998; pp 223-249.
(24) Sun, Q.; Olesik, S. V. Anal. Chem. 1999, 71, 2139-2145.
(14) Jack, D. S.; Rumble, R. H.; Davies, N. W.; Francis, H. W. J. Chromatogr., B: (25) Terfloth, G. LC-GC 1999, 17, 400-405.
Biomed. Sci. Appl. 1992, 584, 189-197. (26) Wolf, C.; Pirkle, W. H. LC-GC 1997, 15, 352-363.
(15) DeJong, E. G.; Kiffers, J.; Maes, R. A. A. J. Pharm. Biomed. Anal. 1989, 7, (27) Anton, K., Berger, C., Eds. Supercritical Fluid Chromatography with Packed
1617-1622. Columns, Techniques and Applications; Chromatography Science Series 75;
(16) Eichhold, T. H.; Bailey, R. E.; Tanguay, S. L.; Hoke, II, S. H. J. Mass Marcel Dekker: New York, 1998.
Spectrom. 2000, 35, 504-511. (28) Berger, T. Packed Column SFC, RSC Chromatography Monographs; The
(17) Janiszewski, J.; Schneider, R. P.; Hoffmaster, K.; Swyden, M.; Wells, D.; Royal Society of Chemistry: Cambridge, 1995.
Fouda, H. Rapid Commun. Mass Spectrom. 1997, 11, 1033-1037. (29) Berger, T. A.; Wilson, W. H. Anal. Chem. 1993, 65, 1451-1455.
(18) Davies, I. D.; Allanson, J. P.; Causon, R. C. J. Chromatogr., B: Biomed. Sci. (30) Yuan, H.; Olesik, S. V. Anal. Chem. 1998, 70, 1595-1603.
Appl. 1999, 732, 173-184. (31) Bargmann-Leyder, N.; Tambute, A.; Caude, M. Chirality 1995, 7, 311-
(19) Joyce, K. B.; Jones, A. E.; Scott, R. J.; Biddlecombe, R. A.; Pleasance, S. 325.
Rapid Commun. Mass Spectrom. 1998, 12, 1899-1910. (32) Pinkston, J. D.; Baker, T. R. Rapid Commun. Mass Spectrom. 1995, 9, 1087-
(20) Zweigenbaum, J.; Heinig, K.; Steinborner, S.; Wachs, T.; Henion, J. Anal. 1094.
Chem. 1999, 71, 2294-2300. (33) Baker, T. R.; Pinkston, J. D. J. Am. Soc. Mass Spectrom. 1998, 9, 498-509.

4236 Analytical Chemistry, Vol. 72, No. 17, September 1, 2000

Sodium heparin was used as the anticoagulant with both sources H2O/MeOH, and was maintained at 0.5 mL/min. The makeup
of plasma. Methanol for SPE was purchased from J. T. Baker flow for this interface between the pcSFC and mass spectrometer
(Phillipsburg, NJ), and formic acid (98%) for SPE was obtained can be used to enhance ionization and to control the pressure
from EM Science (Gibbstown, NJ). Ammonium acetate, methanol drop across the column.34 In this work, the flow and dimensions
(HPLC grade), and formic acid (Reagent grade) for preparation of the interface transfer line32,33 were such that the postcolumn
of the mobile phases were purchased from J. T. Baker. Carbon pressure was maintained at a level to ensure the CO2/MeOH
dioxide for pcSFC was obtained from Air Products and Chemicals, mixture was one phase throughout the column.34 The makeup
Inc. (Allentown, PA). flow was delivered by a model D series 260 syringe pump (Isco,
Standard Solutions, Calibration Standards, and Quality Lincoln, NE). For both separation schemes, the entire chromato-
Control (QC) Samples. Standard solutions of (R,S)-ketoprofen graphic effluent was passed into the mass spectrometer interface
were prepared (as the racemate) at 0.01, 0.10, 1.0, 10, and 100 for nebulization/ionization and subsequent detection.
µg/mL in 80:20 H2O/MeOH. For each solution, 10, 20, or 50 µL The mass spectrometer used with both modes of separation
was added to 1 mL of plasma to prepare calibration standards at was a PE Sciex API III+ (Thornhill, ON, Canada). For the LC
0.1, 0.2, 0.5, 1.0, 2.0, 5.0, 10, 20, 50, 100, 200, 500, 1000, 2000, and separation, the mass spectrometer was operated in the TurboIon-
5000 ng/mL (as the racemate). Quality control samples were Spray configuration, consisting of the articulated IonSpray inlet
prepared similarly at levels of 1.0, 10, 200, and 1000 ng/mL (as used in conjunction with the heated TurboProbe desolvation unit.
the racemate). SIL ketoprofen solution was prepared at 2 µg/mL In the case of the pcSFC, a modified TurboIonSpray source was
in 80:20 H2O/MeOH, and a 25-µL aliquot was spiked into 1 mL of used which allowed the addition of a makeup flow using a tee
plasma for calibration standards, QC samples and study samples junction. The details of this modified source are described
to yield a final concentration of 50 ng/mL. Further details elsewhere.32,33 No sheath flow liquid was used in this work. For
regarding standards and QC samples are provided elsewhere.16 both effluents, the TurboProbe temperature and nitrogen gas flow
Sample Preparation and Solid-Phase Extraction. Prior to rate were 450 °C and 8 L/min, respectively, and the nebulizer
analysis, ketoprofen was isolated from the sample matrix by gas pressure was 60 psi (nitrogen).
performing automated SPE in the 96-well format. The sample Protonated analyte ions were generated using ESI and orifice
preparation procedures were performed using a Biomek 2000 potentials of 3800 and 65 V, respectively. Collisional activation was
equipped with a vacuum manifold and a Multimek 96 channel achieved using argon as the collision gas at a thickness of 270 ×
pipettor, both from Beckman Coulter, Inc. (Fullerton, CA). SPE 1013 molecules/cm2 and a collision energy of 17 eV. The selected
was performed using Oasis HLB sample extraction plates (Waters reaction monitoring (SRM) transition m/z 255-209 was monitored
Corp., Milford, MA) with each column containing 30 mg of for detection of (R)-kt and (S)-kt, while the SRM transition m/z
259-213 was monitored for SIL (R)-kt and SIL (S)-kt. Dwell time
sorbent. HTS deep well tubes were purchased from Matrix
for each transition was 400 ms for the LC analysis and 200 ms for
Technologies Corp. (Lowell, MA). Plasma samples, calibration
the pcSFC analysis because the pcSFC peaks were considerably
standards, and QC samples were prepared for analysis as
sharper than the LC peaks.
described elsewhere.16
Quantitation of (R)- and (S)-kt. Peak areas for the chro-
LC- and pcSFC-MS/MS Instrumentation. The pcSFC
matographic peaks were determined using the PE-Sciex software
solvent delivery system was composed of a Gilson (Middletown,
package, MacQuan version 1.4. Calibration curves for (R)-kt were
WI) modular system, that included a model 308 control pump,
constructed by plotting peak area ratios of (R)-kt/SIL (R)-kt versus
designed to deliver CO2, two model 306 auxiliary pumps, for the
(R)-kt concentrations and fitting these data to a 1/x2 linear
delivery of conventional organic and aqueous mobile phases, a
regression plot. For the LC data, the calibration curve was
model 811C dynamic mixer, a model 821 pressure regulator, a
segmented into two separate curves ranging from 0.05 to 100 ng/
model 831 temperature regulator (column compartment), and a
mL and from 100 to 2500 ng/mL to obtain better accuracy of
model 234 autosampler. The system was configured using a series
calibration standards and quality control samples. The pcSFC
of two- and three-way valves such that it could be converted in
calibration range was also segmented into two curves, the first
5-10 min from LC to SFC mode and vice versa.
ranging from 0.05 to 2.5 ng/mL and the second from 2.5 to 2500
The packed chiral stationary phase used with both modes of
ng/mL, again for better accuracy. Calibration of the instrumenta-
separation was Chirex 3005 (Phenomenex, Torrance, CA) consist-
tion for (S)-kt was performed analogously.
ing of (R)-1-naphthylglycine and 3,5-dintrobenzoic acid. The
Pharmacokinetic Study. Four male volunteers received a 25-
dimensions of the guard and analytical columns used for the LC
mg topical dose of ketoprofen using Oruvail Gel. This product,
separation were 2 × 30 mm and 2 × 250 mm, respectively, while
which is commercially available in the United Kingdom, was
the dimensions for the guard and analytical columns used for the applied on the bicep region of the left arm and remained on the
pcSFC work were 4 × 50 mm and 4 × 250 mm, respectively. skin for the duration of the sampling period. Four additional
The mobile phase for the LC analysis was 30 mM ammonium volunteers received a 25-mg oral dose of ketoprofen, in the form
acetate adjusted to pH 3.5 with formic acid in 5:95 H2O/MeOH of two 12.5-mg Actron caplets. For both dose forms, venous blood
delivered at 0.5 mL/min. The mobile phase for the pcSFC analysis (10 mL) was drawn prior to dosing and at 20, 40, and 80 min and
consisted of a mixture of carbon dioxide/methanol (45:55) that 2.5, 4.5, 8.5, 12.33, 16, and 24 h post-treatment. Plasma was
was maintained at a rate of 5.0 mL/min. For the pcSFC work, a harvested and stored frozen at -70 °C in polypropylene cryovials
makeup flow was added to the effluent after the separation but until the time of analysis.
prior to the MS analysis. The makeup flow consisted of 30 mM
ammonium acetate, adjusted to pH 3.5 with formic acid in 5:95 (34) Chester, T. L.; Pinkston, J. D. J. Chromatogr., A 1998, 807, 265-273.

Analytical Chemistry, Vol. 72, No. 17, September 1, 2000 4237

 Method Comparison. For both LC and pcSFC validation and
study sample analysis, the following analytical attributes were
examined and compared: (1) specificity, (2) linearity, (3) sensitiv-
ity, (4) accuracy, (5) precision, (6) ruggedness, (7) sample
throughput, and (8) utility for analysis of PK study samples. For
the purpose of comparing pcSFC to LC accuracy with actual study
samples, pooled samples were prepared by mixing four 250-µL
aliquots from samples collected at the same time after administra-
tion of a given dose to obtain 1-mL aliquots. These pooled samples
were then prepared for SPE as described above and each resulting
extract was analyzed by both LC-MS/MS and pcSFC-MS/MS.

RESULTS AND DISCUSSION

Mass Spectra. The product ion MS/MS spectra of protonated
kt and SIL kt are shown in Figure 1A and B, respectively. By
evaluating all of the available transitions for quantitation, it was
determined that monitoring the loss of the acidic functionality from
the protonated molecular ion provided the best sensitivity and
selectivity in the plasma matrix and was used for method
development. While this is not a particularly selective transition,
it was shown to provide good results for quantitation of (R)- and
(S)-kt in human plasma. Figure 2. pcSFC chromatograms obtained during method develop-
Separation Optimization. Chromatographic conditions for ment to optimize the separation of the (R)- and (S)-enantiomers of kt
with flow rates of (A) 5 mL/min with a 50:50 CO2/MeOH mobile phase,
both modes of separation were investigated to optimize the
(B) 10 mL/min using a mobile phase consisting of 50:50 CO2/MeOH,
sensitivity, specificity, and resolution of the two enantiomers as and (C) 10 mL/min using a mobile phase of 70:30 CO2/MeOH.
well as sample throughput. For the LC separation, normal-phase
options were explored including the use of a Chiralpak AD column
(Chiral Technologies, Exton, PA) with a hexane/2-propanol/TFA conditions did sacrifice sensitivity (Figure 2B,C). The loss in
mobile phase. These conditions produced a very nice separation sensitivity is likely due to a decrease in the ionization efficiency
with a resolution (Rs) of 2.33. The sensitivity, however, was very caused by reducing the MeOH content. By using 40% MeOH and
poor using this mobile phase and the mass spectrometer signal a 10 mL/min flow rate, a good compromise among resolution,
was unstable without the addition of a makeup flow. Therefore, throughput, and sensitivity was achieved. The elution of both
chiral separations with reversed-phase conditions were explored. enantiomers occurred in less than 1.1 min with a peak intensity
An excellent separation was obtained with a Chirex 3005 4.0-mm- of 16,825 and a resolution of 1.14. These conditions resulted in a
diameter column with a resolution of 2.09. Similar conditions using further 2-fold increase in throughput (versus the 5 mL/min pcSFC
a Chirex 3005 2.0-mm-diameter column provided a ∼4-fold separation) for a total 6-fold increase in throughput versus the
increase in sensitivity and faster run times, but resulted in reduced LC-MS/MS conditions. However, at the time of the original
resolution of 1.42. A practical compromise between sensitivity, comparison of the pcSFC and LC methods, the pcSFC instrumen-
resolution, and analysis time resulted in the selection of the 2.0- tation used for these studies was not capable of providing flow
mm column at a flow rate of 0.5 mL/min for the LC conditions. rates above 5 mL/min. Therefore, the maximum available flow, 5
The pcSFC separation of (R)- and (S)-kt was attempted with a mL/min, was utilized with a 55% MeOH mobile phase for the
variety of conditions and columns. Various combinations of CO2 method comparison work reported here. This separation, like the
and MeOH were used with Chiralcel OD, Chiralcel OJ, and LC separation, provided a good balance among sensitivity, resolu-
Chiralpak AD columns (Chiral Technologies), but no acceptable tion, and sample throughput.
separation of the kt enantiomers was obtained. Also, the 2.0-mm Specificity. Specificity was evaluated for both modes of
Chirex 3005 column was used with pcSFC and produced a separation and it was determined that, with MS/MS detection and
separation with a 45% valley between the enantiomers (Rs ) 0.78). SPE cleanup of the human plasma matrix, there was very rarely
This result was judged unacceptable for further quantitative an interference observed in the retention window of either analyte
method development. The 4.0-mm column, however, did produce with either separation methodology. A slight disturbance was
a very good separation. Figure 2A shows the separation with a observed just after the void volume as shown in Figure 3. As for
resolution of 1.47 achieved with a 5 mL/min flow rate and 50% enantiomeric specificity, typical chromatographic resolution was
MeOH in CO2 in less than 2 min. monitored for each separation mode and found to be slightly better
More recent results have been obtained using 10 mL/min for pcSFC versus LC. Although, as stated in the method optimiza-
pump heads on the Gilson pcSFC instrument. Under the condi- tion section, resolution can be improved at the cost of sensitivity
tions described above for the 5 mL/min result, but using a 10 and sample throughput.
mL/min flow rate, a separation could be obtained in <0.9 min, Linearity. The quantitative range for both modes of separation
but with lower resolution at 1.09. By decreasing the MeOH was shown to extend from 0.05 to 2,500 ng/mL, representing
concentration to 30% with a flow rate of 10 mL/min, a near-baseline nearly 5 orders of magnitude. To avoid signal saturation due to
separation (Rs ) 1.48) was obtained in less than 1.5 min, but these instrument limitations, and to preserve linearity of response, it
4238 Analytical Chemistry, Vol. 72, No. 17, September 1, 2000
 Table 1. Average QC Recoveries for 4 Days of
Validation Obtained Using LC-MS/MS

concn mean no. of

analyte (ng/mL) recovery (%) RSD (%) replicates
(R)-kt 500 102.9 7.6 12
100 99.7 6.0 12
5 99.2 4.7 12
0.5 102.2 4.5 12

(S)-kt 500 101.2 5.8 12

100 101.9 5.7 12
5 99.9 5.6 12
0.5 98.7 4.8 12

Table 2. Average QC Recoveries for 3 Days of

Validation Obtained Using pcSFC-MS/MSa

concn mean no. of

analyte (ng/mL) recovery (%) RSD (%) replicates
(R)-kt 500 103.7 3.8 12
100 99.7 3.9 12
5 105.7 7.2 12
0.5 102.8 4.3 11
(S)-kt 500 103.1 3.4 12
100 98.6 4.1 12
5 103.0 6.4 12
0.5 105.0 5.6 11
a One QC sample at the 0.5 ng/mL level was excluded because of

Figure 3. Chromatographic profiles of plasma blanks and 50 pg/ a batch building error.
mL spiked plasma calibration standards generated using (A) LC-
MS/MS and (B) pcSFC-MS/MS for the determination of (R)- and (S)-
kt.
Sensitivity. The sensitivity of both methodologies was evalu-
was necessary to dilute the higher calibration standards, QCs, ated by spiking known amounts of kt into blank plasma. Figure 3
and study samples. This large dynamic range enabled the shows the plasma blanks generated using both methodologies
generation of PK curves from both topical and oral samples in a and the LLOQ plasma spike of 50 pg/mL. That level corresponds
single batch. Using single-batch analysis minimized any analytical to ∼10 pg on column for each enantiomer. Similar signal-to-noise
bias that might have been introduced by using one method for ratios are observed for both separation methodologies despite the
the oral PK samples and a separate method for the topical PK fact the pcSFC chromatograms were generated using a 4.0-mm
samples. column versus the 2.0-mm column used with the LC separation.
Though a single, 1/x2 weighted calibration curve covering the The average recoveries for the 50 pg/mL (R)- and (S)-kt calibra-
entire calibration range provided acceptable linearity, accuracy tion standards using the LC method were 98.0 (2.2) and 94.8%
was improved by segmenting the calibration curve into a low and (7.8%), respectively, with relative standard deviations (RSDs)
a high portion. For the LC data, one curve for each analyte was displayed in parentheses. For the pcSFC method, the recoveries
constructed using calibration standards from 0.05 to 100 ng/mL, of the 50 pg/mL calibration standards were 98.8 (3.6) and 103.2%
while the second segment ranged from 100 to 2500 ng/mL. All (5.8%) for (R)- and (S)-kt, respectively.
curves used 1/x2 linear regression. Typically, recoveries of the Precision and Accuracy. Precision and accuracy were evalu-
calibration standards were within (15% of theoretical. For the ated by spiking known amounts of kt into blank human plasma
calibration curves used in this comparison, the range of correlation at four levels throughout the quantitative range of the methodol-
coefficients for the lower portions of the curves extended from ogy. Table 1 displays the results for kt recovery using the LC-
0.9942 to 0.9991 while the correlation coefficients for the upper MS/MS methodology and Table 2 displays the results using the
portions of the curves ranged from 0.9968 to 0.9997. For the pcSFC pcSFC-MS/MS methodology. For the LC methodology, the
calibration, one curve was constructed using calibration standards average accuracy ranged from 98.7 to 102.9%, with variability
from 0.05 to 2.5 ng/mL, while the second curve ranged from 2.5 (RSD) ranging from 4.5 to 7.6%. For the pcSFC methodology, the
ng/mL to 2500 ng/mL; all pcSFC curves were also weighted with average accuracy ranged from 98.6 to 105.7%, with variability
1/x2 and recoveries of standards were typically within (15% of ranging from 3.4 to 7.2%. Generally, similar results for accuracy
theoretical. The range of correlation coefficients for the lower and precision of spiked plasma samples were obtained using both
portions of the curves used in these studies ranged from 0.9952 modes of separation.
to 0.9991, while the correlation coefficients for upper curves Ruggedness. The ruggedness of both methods was evaluated
ranged from 0.9981 to 0.9996. For both methods, correlation by analyzing batches of bioanalytical samples and noting problems
coefficients for linear regression typically exceeded 0.995, with that occurred using normal method conditions. Both methods
intercepts near zero. were capable of analyzing batches containing a minimum of 150-
Analytical Chemistry, Vol. 72, No. 17, September 1, 2000 4239
 Figure 5. Plots of (A) (R)-kt and (B) (S)-kt plasma levels versus
time after topical and oral administration of kt and comparing the LC
and pcSFC quantitative results.
Figure 4. Comparison of (A) LC and (B) pcSFC for analysis of a
PK study sample collected at 2.5 h after topical administration of
Oruvail gel. Approximately 100 pg/mL of each enantiomer was present
in this sample. The major difference is that the pcSFC separation Pharmacokinetic Study. Representative PK samples col-
can be performed in about one-third the time that it takes to perform lected after topical and oral administration of kt were prepared
the LC analysis. and analyzed by both modes of chromatographic separation to
further investigate the utility of the pcSFC approach and to
compare the pcSFC quantitative results with LC results for actual
200 samples prior to noting a slight loss of chromatographic
study samples. Chromatograms comparing the LC and pcSFC
resolution or occasionally a slight increase in back pressure. In
analysis of a plasma sample collected 2.5 h after topical admin-
all cases, the original chromatographic performance was recovered istration are shown in Figure 4. The signal-to-noise ratios of both
by simply replacing the guard column. With the very minimal the (R)- and (S)-kt peaks are very similar for both the LC and
problems observed with either method and the easy restoration pcSFC analysis of this sample containing ∼100 pg/mL of each
of performance by replacing the guard column, the overall enantiomer. These chromatograms again illustrate the dramatic
ruggedness of both techniques is acceptable for bioanalytical decrease in analysis time that can be achieved when pcSFC is
analyses. used for chiral bioanalytical analyses. Slightly better resolution
Sample Throughput. The sample throughput of the LC-MS/ was also exhibited with the pcSFC methodology.
MS methodology is much greater than that of the typical LC- The PK plots for (R)-kt and (S)-kt comparing the LC and pcSFC
UV methods reported in the literature4-10 (10-25 min/sample) quantitative results and showing the plasma levels of (R)- and (S)-
largely due to the enhanced specificity of the tandem mass kt measured after topical and oral administration are provided in
spectrometric detection. This specificity allowed analysis times Figure 5A and B, respectively. Excellent correlation of the results
to be reduced to 6.5 min from injection to injection. The use of from both methods was achieved. The deviation of each pcSFC
the pcSFC separation improved the injection-to-injection cycle time result from the corresponding LC result was calculated. For (R)-
to 2.3 min/samplesa factor of ∼3 times better than the LC-MS/ kt, the average of the absolute percent deviations was 3.8% with
MS separation and up to 10-fold better than the LC-UV methods. a maximum deviation of 10.9%. The average of the absolute
With the faster flow rates available using the 10-mL pump heads, deviations of the pcSFC values from the LC values for (S)-kt was
another 2-fold reduction of the analysis time is possible. That 2.5% with a maximum deviation of 11.9%.
results in a total 6-fold throughput improvement versus the LC- From a PK viewpoint, the topical kt levels were below the
MS/MS approach, even when using the larger 4.0-mm diameter LLOQ until the 2.5-h time point and then continued to increase
column versus the 2.0-mm column used with the LC-MS/MS throughout the entire sampling period. The oral dose yielded a
work. Of course, the throughput of all these methods could be maximum concentration at ∼2.5 h, nearly 4 orders of magnitude
increased even further by using other approaches such as column higher than its respective topical plasma concentration. Interest-
switching. ingly, the plasma concentrations from oral and topical administra-
4240 Analytical Chemistry, Vol. 72, No. 17, September 1, 2000
tion were nearly identical for the 16-h time point. The levels of MS/MS approach for most bioanalytical attributes, but has the
(R)-kt compared to (S)-kt show slight differences throughout the advantage of a 3-fold, and if faster flow rates are used, up to 6-fold,
24-h time course for topical samples;16 however, there are clearly higher sample throughput. This increase in throughput for the
higher levels of (S)-kt relative to (R)-kt by the end of the 24-h instrumental portion of the assay is important for the effective
period following the oral dose. These findings agree with previous utilization of capital intensive mass spectrometry instrumentation.
reports.2,3 To further increase pcSFC bioanalytical throughput for pharma-
ceutical analyses, the use of other techniques such as column
CONCLUSIONS switching could be employed. It may also be possible to use the
A pcSFC-MS/MS method was developed and validated for pcSFC-MS/MS approach for nonchiral bioanalytical analyses
quantitation of kt in human plasma from 0.05 to 2500 ng/mL; this
either with or without column switching so that the analysis times
method was shown to be valuable for measuring kt concentrations
of those gradient or isocratic separations could also be reduced.
in human plasma following oral and topical administration. The
pcSFC-MS/MS methodology for the chiral determination of kt
was found to be comparable to the LC-MS/MS approach in terms ACKNOWLEDGMENT
of specificity, linearity, sensitivity, accuracy, precision, and rug- The authors thank Ramona Quintanilla for the synthesis of the
gedness, but had an advantage of a total analysis time of ∼1/3 of stable-labeled internal standards used for these studies.
the time for the LC-MS/MS methodology at 2.3 min/sample. It
is ∼10-fold faster than LC-UV methods and has an LLOQ of 500-
fold lower than UV methods for the determination of kt in plasma. Received for review January 24, 2000. Accepted June 2,
The application of pcSFC separation coupled with tandem mass 2000.
spectrometric detection is comparable to the widely used LC- AC000068X

Analytical Chemistry, Vol. 72, No. 17, September 1, 2000 4241