Determination of Carprofen in Blood by Gas Chromatography Chemical Ionization Mass Spectrometry B. J. Hodshon, W. A. Garland, C. W. Perry? and G. J. Bader' Department of Biochemistry and Drug Metabolism, Hoffmann-La Roche Inc., Nutley, New Jersey 071 10, USA
A gas chromatographic chemical ionization mass spectrometric assay has been developed to measure carprofen in blood. The method features the addition of either a structural or stable isotope analog internal standard to plasma prior to a simple benzene extraction at pH 4.5. The residue, after removal of the benzene, is methylated with ethereal diazomethane. Following evaporation of the methylating solvents, a portion of the reconstituted residue is analyzed by gas chromatography mass spectrometry. The mass spectrometer is set to monitor in the gas chromatographic effluent the [MH]+ ions of carprofen methyl ester and the methyl ester of the internal standard generated by isobutane chemical ionization. Assay sensitivity is 5 pmol ml-'. When 200 pmol ml-' samples are analyzed using a stable isotope analog as the internal standard, the precision and accuracy are both 4%. Using a structural analog as the internal standard, the assay was neither as precise nor as accurate.
INTRODUCTION
Carprofen (1, Table 1) is currently under clinical investigation for use as an anti-inflammatory agent. The compound has been found to possess analgesic, antipyretic and anti-inflammatory properties similar to indomethacin and aspirin but with fewer adverse side effects.' Procedures have been reported for the quantitation of 1 in blood using thin-layer chromatography (TLC), high performance liquid chromatography (HPLC) with either ultraviolet or fluorescence
Table 1. Chemical structures and their designations
detection and gas chromatography (GC)with electron capture (EC) d e t e ~ t i o n . ' ~ A mass spectrometric assay for 1 was required to confirm the specificity of the HPLC procedure, to assay samples containing low levels of carprofen and to apply certain stable isotope procedures to studies of the bioavailability and metabolism of ~arprofen.'.~ Initially an assay was developed which used a structural analog of carprofen, compound 3, as internal standard. Later, compound 2, a stable isotope analog of carprofen, was used for this purpose. Data gathered from the analysis of the same sample using either 2 or 3 as internal standard permitted a comparison of the assay precision and accuracy when either a structural or stable isotope analog was used as internal standard.
EXPERIMENTAL H Compound
R
Designation
d, /-2-(6-Chloro-SH-carbazol2 4 )propanoic acid
1
d, /-2-(6-Chloro-9H-carbazol2-~1)-3,3,3 trideuterio propanoic acid 3-(6-Chloro-SH-carbazol-2-yl) propanoic acid 2-(6-Chloro-9H-carbazol-2-yl) propanedioic acid dirnethyl ester 2-Trideuteriomethyl-2-(6-
2
C2H3 I -CH--COzH
3
-CH2-CH2-C02H
4
-CH-(C02CH3)2
I
-CH -C02H
5
chloro-9H-carbazol-2-yl)
propanedioic acid dirnethyl ester (6-Chloro-SH-carbazoI-2-yl)
CZH3
I
-C-(C02CH3)2 6
acetic acid methyl ester
' Department of Chemical Research.
Instrumentation
CH3
-CH2-C02-CH3
A Finnigan 3200 mass spectrometer was interfaced with the manufacturer's Model 9500 gas chromatograph and Promin@ selected ion monitor. A glass column (5 ft X 2 mm i.d.) was packed with 3% OV-17 on 100/120 mesh Gas Chrom Q purchased from Applied Science Laboratories. Before use, the column was conditioned overnight at 295 "Cwith a nitrogen flow of 10 ml min-'. GC operating temperatures were: column oven, 280285 "C; injection port, 300 "C; separator oven and transfer line, 250°C. Isobutane was used as both GC carrier gas and CI reactant gas. The ion source was operated without external heating at a pressure of 0.5 Torr. The ion source operating parameters were as follows: lens, -15 V; ion energy, +1OV; electron energy, -175 V; filament emission, 1 mA. The voltage across the continuous dynode electron multiplier was 2000 V. Each channel of the Promin@was operated at
CCC-0306-042X/79/0006-0325$03.00 BIOMEDICAL MASS SPECTROMETRY, VOL. 6, NO. 8, 1979 325
B. J. HODSHON, W. A. GARLAND, C. W. PERRY AND G. J. BADER
100 ms dwell time, 0.5 Hz frequency response and a sensitivity of lo9 V/A. Ion chromatograms were recorded on a four channel Rikadenki pen recorder.
Materials
Nanograde benzene, methanol and chloroform were obtained from Mallinckrodt Chemicals. One molar borate buffer, pH 10, and 5 M acetic acid buffer, p H 4.5 were prepared according to standard method^.^ Ethanol free ethereal diazomethane was prepared from Diazald@ (Aldrich Chemical Co.) using the manufacturer's suggested procedure. Twenty ml culture tubes (Kimble No. 45066-A) with Teflon@lined screw caps were used for sample extraction. Fifteen ml screw cap centrifuge tubes (Kimble 45 161) were used for final solvent evaporation. Extractions were performed by shaking the tubes for 20 min on an IEC No. 106234 bottle shaker. Centrifugations were performed at 10°C for 10 min at 1500Xg using a Damon CRU-500 refrigerated centrifuge. The solvents were removed at 60 "C under a gentle stream of nitrogen (N-Evap@, Organomation Assoc.). Aqueous solutions of 1, 2 and 3 were prepared separately. To the appropriate amount of compound, enough 1M, pH 10, borate buffer was added to maintain the compound in solution. The basic solution was titrated to p H 7.6 using 0.5 N HCI and the resulting solution was brought to volume with water. Using this solvation procedure, the following solutions were prepared for use as standards: 500 pmol per 100 pl of 1, 1000 pmol per 100 pl of 2 and 1000 pmol per 100 p1 of 3. Compounds 1 and 3 were obtained from Dr W. Scott, Hoffmann-La Roche. Compound 4 was prepared by methanolysis of the diethyl ester obtained as a gift from Dr W. Harry Mandeville, Hoffmann-La Roche. The synthesis of the deuterated internal standard, compound 2 from compound 4 is described below. d, 1-2-(6-Chloro-9H-carbazol-2-yl) propanoic acid (2). Sodium (253 mg, 11milli-equivalents) was dissolved in 10 ml of methanol and asolution of 4 (3.315 g, 10 mmol) in 60 ml of dioxane was added. The clear yellow solution was distilled to remove methanol until the boiling point of the distillate stabilized at 101 "C, at which point 4 was apparently in the form of its insoluble sodium salt. The suspension was cooled to 20 "C, trideuteriomethyl iodide [1.739 g, 0.763 ml, 12 mmol (Stohler Isotopes)] was added, and the mixture was stirred at room temperature for 22 h. Acetic acid (721 mg, 0.684 ml, 12 mmol) was added to neutralize the base. Additional dioxane was added and the mixture was filtered. Evaporation of the filtrate left a solid residue (3.54 g) which was crystallized from 15 ml of methanol to give a first crop, 2.37 g, m.p. 133-149 "C and a second crop, 194 mg, m.p. 125-142°C. These two crops were combined and recrystallized from toluene + hexane to afford 2.21 g of 5, m.p. 147-154 "C, which contained about 10-20% of a less polar impurity which was later identified as the N-methyl homolog of 5. Most of this material was used in the next step to prepare 2. The remainder was recrystallized three times from methanol, with attendant loss of material, to afford white crystals, 123 mg, m.p. 158-159 "C, which still included about 1-2% of the N-methyl homolog of 5. 326 BIOMEDICAL MASS SPECTROMETRY, VOL. 6, NO. 8, 1979
The diester 5 (1.744 g, 5 mmol) was dissolved in 25 ml of boiling methanol and 12.5 ml of 1N aqueous sodium hydroxide was added dropwise, causing 5 to precipitate as a fine suspension. With continued refluxing the precipitate dissolved, but within 5-10 min the boiling s o h tion became cloudy again. The mixture was refluxed overnight. A little water was added to redissolve a small amount of solid which had deposited and the clear hot soIution was treated with charcoal and filtered hot. The filtrate was carefully acidified (gas evolution) with 1 N hydrochloric acid, brought to a boil, and acetic acid added dropwise at the boiling point until all of the solids redissolved. Upon cooling the solution and seeding with 1, the crude product crystallized and was collected by filtration and washed with water. Recrystallization from toluene afforded two crops, 349 mg, m.p. 197-203 "C, and 228mg, m.p. 193-204"C, both still impure. Recrystallization from aqueous acetic acid and then from ethyl acetatethexane failed to remove all of a minor impurity (which was later identified as 6), but one small second crop (49 mg) of 2 was obtained which was >98% pure. The m.p. and IR and UV spectra of this compound were identical to that of authentic carprofen and its 'H NMR spectrum was consistent with that of the desired compound. By GCMS analysis, the compound was >99% *H3and showed chromatographic properties identical to authentic carprofen. Assay procedure
One ml of control blood, and the appropriate amount of 1 plus additional water to make l m l were added together for the standard curve samples. One ml of water was added to 1 m l of blood for the unknown biological samples. To all blood samples, standards and unknowns, were added 1000pmol of either 2 or 3 and the mixture was briefly vortexed. Three-tenths ml acetic acid buffer was added followed by 10 ml benzene; the tubes were capped and the samples were extracted. Eight ml of the benzene phase was transferred to a 15 ml conical extraction tube. Following evaporation of benzene, the residue was dissolved in 100 pl methanol. One ml ethereal diazomethane was added and the solution was allowed to stand at room temperature for 1h. This solution was evaporated and the residue was reconstituted in 50 pl chloroform. The mass spectrometer was set to monitor the [MH]+ ions corresponding to the methyl esters of 1 and 3 ( m / z 288), or 1 ( m / z 288) and 2 ( m / z 291). Two to lop1 of the final solution was injected. Approximately 60 s after injection the GC divert valve was turned off and at 75 s the ionizer was turned on. The retention time of 1-and 2-methyl esters under these conditions was approximately 140 s and the retention time of 3-methyl ester was approximately 180 s. For assays using 2 as internal standard the peak height ratios of m / z 288 to m / z 291 were calculated and converted to carprofen concentrations using a standard curve which was run with each experiment. The standard curve samples were prepared from spiked samples containing 0, 50, 100, 500, 1000, 2000 and 3000 pmol of 1 per ml blood and 1000 pmol of 2 per ml blood in each sample. The nonlinear standard curves were analyzed by a computer programmed for nonlinear @ Heyden & Son Ltd, 1979
DETERMINATION OF CARPROFEN IN BLOOD
288
"1
I
H
CH3
Mot. wt ~ 2 8 7
291
1t
+tJ Time
0 0 I
Mol.
wt =287 256 CH3 150
blood sam'ples from a subject 1 h after receiving a 100 mg oral dose of carprofen. The solid line is the tracing of rnlz 288 (carprofen and 3) and the dashed line the tracing of m l z 291 (2). Ion chromatogram (a) is the output recording using 2 as the internal standard in the assay. Ion chromatogram (b) is the output recording using 3 (smaller peak) as the internal standard in the assay. Both internal standards were added at a concentration of 1000 pmol. The carprofen level in this sample was approximately 1000 pmol.
I
..H
250
200
a 300
350
4 M
m/z Figure 1. lsobutane (0.5Torr) CI mass spectra of (a) carprofen ( l ) ,(b) the stable isotope internal standard (2) and (c) the structural analog internal standard (3).
regression analysis as described in the discussion. For assays using 3 as internal standard, the data from the standard curve samples were analyzed by linear regression analysis.
RESULTS AND DISCUSSION The isobutane CI mass spectra of the methyl esters of 1, 2 and 3 are shown in Fig. 1. As expected the spectra are quite simple consisting principally of [MH]+ molecular ions. The spectra of the methyl esters of 1 and 2 show small ions corresponding t o the loss of the elements of methyl formate. The spectrum of 3-methyl ester shows an ion corresponding to the loss of the elements of methanol. Loss of methyl formate from 3-methyl ester would generate an unstabilized primary carbonium ion, while the same loss from the methyl esters of 1 or 2 would yield a stabilized benzylic carbonium ion. Typical ion chromatograms from the analysis of the same blood sample using either 2 or 3 as the internal standard can be seen in Fig. 2. If 2 is used as the internal standard both the internal standard and carprofen elute @ Heyden & Son Ltd, 1979
Time
Figure 2. Ion chromatograms from the duplicate analysis of 1 ml
C ~ ~ C " , - C H * - IIc
50t
-
at the same time. This procedure has the theoretical disadvantage that two different ions must be monitored. It has been shown that such ion switching can introduce considerable error.' Using 3 as the internal standard the same ion can be monitored for both carprofen and the internal standard. This procedure has the disadvantage of requiring very exacting conditions in order to chromatographically separate carprofen and its internal standard. In this regard a large number of injections of an ethyl acetate extract of control blood were required at the beginning of each set of injections in order to decrease peak tailing and increase sensitivity. Since compounds 1 and 3 are not structurally identical, peak areas, calculated as the peak widths at half-height times peak heights, were used to determine ion intensities in assays using 3 as internal standard. Assay recoveries were determined from six, 1 ml, blood samples each spiked with 1000 pmol of 1 and six, 1 ml, blood samples each spiked with 1OOOpmol of 3. Based on a comparison of the ion intensities from these samples and the ion intensities from the injection of authentic 1- or 3-methyl ester, the recoveries ( fSD) were 77 f 26% for 1 and 2 and 44* 11% for 3. Recoveries of 100 pmol of 2 from blood were similar. A typical standard curve relating the amount of carprofen added to plasma to the resulting ion ratios using 2 as internal standard can be seen in Fig. 3. As expected the curve is not a straight line.'-'' The nonlinearity is due to the contribution of the I3Cisotope BIOMEDICAL MASS SPECTROMETRY, VOL. 6, NO. 8, 1979 327
B. J. HODSHON, W. A. GARLAND, C. W. PERRY AND G. J. BADER
cedure the fit was not as good. The percent difference (*SD) was 25 15,26&16,10* 3,8 f 1,4*4 and 8 f 3 for the same six concentrations. The mean percent difference in this case was 13.5f9.5. When the structural analog 3 was used as internal standard, the ion ratio of the m / z 288 corresponding to 1-methyl ester to m / z 288 corresponding to 3-methyl ester was determined using the peak area of each ion. The standard curve data was plotted as the observed ion ratios versus spiked sample concentrations. The unknown ratios of 1-methyl ester to 3-methyl ester were converted to pmol of 1using the slope and intercept values obtained from linear regression analysis. Figure 4 is a typical standard curve using 3 as internal standard. Based on the results from several repetitions, the standard curve using 1000pmol of 3 as internal standard was linear to approximately 2000 pmol. The correlation coefficient *SD for 14 standard curves of this type was 0.996 0.004. The sensitivity limit in the assay procedure using 2 as internal standard is illustrated in Fig. 5. As shown in (a), adding 100 pmol of 2 to control blood gave a virtually blank ion chromatogram at the ion corresponding to 1 as compared with the ion chromatogram from the analysis of 1 ml blood spiked with 5 pmol of 1and 100 pmoles of 2 (b). As can be seen, the ion current due to 1-methyl ester in (b) is considerably above the background noise level. To determine assay accuracy and precision, a large volume of the same control blood used to generate the standard curve was spiked to give a concentration of 200 pmol 1 m1-l. One ml aliquots were analyzed using 2 or 3 as internal standard in separate experiments. The
*
*
pmol I
Figure3. Typical standard curve relating the ion ratio of rnlz 288 to m / z 291 to the amount of 1 added to 1 ml of blood. The line is a 'least square' computer fit using program NONLIN.
peak of the 37Cl isotope peak of unlabeled carprofen. The nonlinearity could be avoided by monitoring the 37 C1 isotope (mlz 293) of the deuterated internal standard. However, interfering peaks in the m / z 293 ion chromatograms precluded the use of this mass. The nonlinear standard curve data obtained using 2 as internal standard were analyzed using computer program NONLIN12 to fit the ion ratio concentration data to the equation R = (x + A ) / ( B x+ C).13314 In this equation C is approximately equal to the amount of [2H3]carprofen added, B is approximately equal to the ratio of m / z 291 to m / z 288 in the mass spectrum of 1 and A is approximately equal to C times the ratio of m / z 288 to m / z 291 in the mass spectrum of 2. Once found, the values of A, B and C are used in the rearranged expression x = ( A- R C ) / ( R B- 1) to find an unknown x given an experimentally determined R. For three standard curves the means (*SD) of constants A, B and C using 1000 pmol of 2 as internal standard were 5.35 2.26 pmol, 0.086 f 0.013 and 984* 51 pmol, respectively. Theoretical values for A , B and C are Opmol, 0.05715 and 1000 pmol. A comparison can be made between the known concentration values and the concentration values generated from the experimentally determined ion ratios using the computer calculated standard curve parameters. The percent differences (*SD) between the expected and observed values for three standard curves generated using the NONLIN procedure were 9 * 5 , 8 * 1 3 , 3*2,2* 1 , 4 * 4 and 2* 1 for the 50,100, 250, 500, 1000 and 3000 pmol concentration points, respectively. The mean percent difference ( f SD)for the six concentration points was 4.7*3.1. If the same data were analyzed using a conventional least square pro-
*
328 BIOMEDICAL MASS SPECTROMETRY, VOL. 6, NO. 8, 1979
I
I
1
500
loo0
2ooc
pmol I
Figure 4. Typical standard curve relating the ion ratio of m l z 288 corresponding to 1 and mlz 288 cor'responding to 3 to the amount of 1 added to 1 rnl of blood.
@ Heyden & Son Ltd, 1979
DETERMINATION OF CARPROFEN IN BLOOD
Table 3. Drug blood levels from one patient administered orally 100mg of carprofen. Levels were obtained using either 2 or 3 as internal standard in separate experiments Concentration Hours after dosing
0 0.5 1 2 4 6 8 12 16 24 36 48
Time
Time
Figure 5. Ion chromatograms resulting from the analysis of 1 ml of control blood to which was added 100 pmol of 2 and either 0 pmol (a) or 5 pmol of carprofen (b).The solid line is the tracing of m / z 288. The dashed line is the tracing of m l z 291.
results are shown in Table 2. Using 2 as internal standard resulted in a mean ( *SD) carprofen concentration of 193*4pmol from 10 replicates. Using 3 as internal standard, the results were quite disappointing. The mean concentration of carprofen determined from 1 0 replicates was 27 1pmol with a SD of f 11. Although the precision is reasonable (*4%), the accuracy is off by 33% from the actual amount of compound known to be present in the sample. In separate experiments carprofen concentrations in a set of samples from one patient were determined using either 2 or 3 as internal standard. The results, Table 3, show a large difference, as much as loo%, in the concentration of 1 determined
Table 2. A large volume of blood was spiked with 1 to give a concentration of 200 pmol m1-l. One ml aliquots were assayed in separate experiments using either 2 or 3 as internal standard Concentration
i
-
of 1
Using stable isotope internal standard (2)
Using structural analog internal standard (31
190 186 188 192 196 193 198 195 198 195 193.10 SD = 4.09
265 269 269 263 260 266 278 296 286 266 = 271.80 S D = 11.41
x=
@ Heyden & Son Ltd, 1979
Using stable isotope internal standard (2)
pmol ~
ml
of 1
Using structural analog internal standard (31
0
0
2200 1932 12580 4468 2979 1751 1118 815 487 222 147
2812 2626 17170 10256 5961 3554 1360 812 900 463 21 1
using either of the two internal standards, As before, the concentrations were considerably higher when the structural analog was used as the internal standard. These same samples were then analyzed using the published HPLC p r o c e d ~ r e .The ~ concentrations determined by this assay showed a good correlation (correlation coefficient = 0.998, slope = 1.28h0.03) with the concentrations determined using 2 as internal standard and a poorer correlation (correlation coefficient = 0.953, slope = 1.81k 0.18) with the concentrations determined using 3 as internal standard. Thus, the use of 3 as internal standard gives less precise and less accurate results than when 2 is used as internal standard in spite of the advantage of not requiring mass switching. Much additional work would be needed to provide an explanation of why the ion ratios from the patient or control blood were not consistent with the ion ratios from the standard curve samples. Perhaps the GC properties of 1 and 3 are so different as to preclude good quantitation. ShatkayI6 and Roberson” have both commented on the potential unreliability of GC procedures using internal standards. Based on their work errors of the magnitude observed here may not be unexpected. In addition, the intensities of the fragment ions may be sensitive to small changes in ion source temperature^.'^'^^ As mentioned previously, 1 and 2 undergo a different fragmentation process than 3. These two fragmentation routes may show different sensitivities to temperature. In this regard our procedure was to analyze the standard curve samples and then the patient or reproducibility samples. Although unexplained, our results are consistent with comparisons of the use of labeled versus nonlabeled internal standards made by other investigators. Quantitative measurement of norepinephrine using deuterated norepinephrine as internal standard demonstrated good accuracy and precision, while use of a-methylnorepinephrine for the same purpose was unsatisfactory.2” In another study, the long-term variation in peak ratio was examined using either deuterated desipramine or maprotiline as the internal standard in a desipramine assay.’l The variance of the structural aflalog internal standard was three- to ten-fold of that the stable isotope analog internal standard. BIOMEDICAL MASS SPECTROMETRY, VOL. 6, NO. 8, 1979 329
B. J. HODSHON, W. A. GARLAND, C. W. PERRY AND G. J. BADER
Since the assay using 2 as internal standard provides an acceptable quantitiative procedure for our studies with carprofen, we are not planning any further experiments with 3 as internal standard. We feel,
however, that our experience with this structural analog internal standard will be of use to other investigators quantitating biologically important substances by mass spectrometry.
REFERENCES 1. L. 0. Randall and H. Baruth, Arch. lnt. Pharmacodyn. Ther. 220, 94 (1976). 2. J. A. F. de Silva, N. Strojny and M. A. Brooks, Anal. Chim. Acta 73,283 (1974). 3. C. V. Puglisi, J. C. Meyer and J. A. F. de Silva, J. Chromatogr. 136,391 (1977). 4. G. Palmskog and E. Hultman, J. Chrornatogr. 140,310 (1977). 5. J. S. Dutcher, J. M. Strong, S. V. Lucas, W.-K. Lee and A. J. Atkinson, Clin. Pharmacol. Ther. 22, 447 (1977). 6. H. d'A. Heck, N. W. Flynn, S. E. Buttrill, R. L. Dyer and M. Anbar, Biorned. Mass Spectrom. 5,250 (1978). 7. J. A. F. de Silva and C. V. Puglisi,AnalChem. 42,1725 (1970). 8. M. G. Lee and 6.J. Millard, Biorned. Mass Spectrom. 2, 78 (1975). 9. J. F. Pickup and K. McPherson, Anal Chem. 48, 1885 (1976). 10. J. R. Chapman and E. Bailey, J. Chromatogr. 89, 215 (1974). 11. J. F. Holland, R. E. Teets, M. A. Bieber and C. C. Sweeley, American Society for Mass Spectrometry 21st Annual Conference on Mass Spectrometry and Allied Topics, San Francisco, California (1973). Abstracts, p. 55. 12. C. M. Metzler, A Users Manual for NONLIN, Upjohn Co., Kalamazoo, Michigan 49001, USA. 13. 6. H. Min, W. A. Garland, K.-C. Khoo and G. S. Torres, Biorned. Mass Spectrom. 5,692 (1978).
330 EIOMEDICAL MASS SPECTROMETRY, V3L. 6, NO. 8, 1979
14. W. A. Garland, R. Muccino, B. H. Min, J. Cupano and W. E. Fann, Clin. Pharrnacol. Ther. 25,844 (1979). 15. J. H. Beynon, Mass a n d Abundance Tables for Use in Mass Spectrometry, p. 118. Elsevier, Amsterdam (1963). 16. A. Shatkay and S. Flavian, Anal. Chern. 49, 2222 (1977). 17. J. C. Roberson, Anal. Chern. 50, 2145 (1978). 18. F. Field, J. Am. Chern. SOC.91, 2827 (1969). 19. W. R. Sherman, W. J. Fies and E. J. Heron, American Society for Mass Spectrometry, 24th Annual Conference on Mass Spectrometry and Allied Topics, San Francisco, California (1976). Proceeding, p. 293. 20. R. C. Murphy, J. A. Zirrolli, K. Clay, S. E. Hattox and .?I Helbig, American Society for Mass Spectometry, 23rd Annual Conference on Mass Spectrometry and Allied Topics, Houston, Texas (1975). Proceedings, p. 427. 21. M. Claeys. S. P. Markey and W. Maenhaut, Biorned. Mass Spectrom. 4, 122 (1977).
Received 12 February 1979
@ Heyden & Son Ltd, 1979
@ Heyden & Son Ltd, 1979
A gas chromatographic chemical ionization mass spectrometric assay has been developed to measure carprofen in blood. The method features the addition of either a structural or stable isotope analog internal standard to plasma prior to a simple benzene extraction at pH 4.5. The residue, after removal of the benzene, is methylated with ethereal diazomethane. Following evaporation of the methylating solvents, a portion of the reconstituted residue is analyzed by gas chromatography mass spectrometry. The mass spectrometer is set to monitor in the gas chromatographic effluent the [MH]+ ions of carprofen methyl ester and the methyl ester of the internal standard generated by isobutane chemical ionization. Assay sensitivity is 5 pmol ml-'. When 200 pmol ml-' samples are analyzed using a stable isotope analog as the internal standard, the precision and accuracy are both 4%. Using a structural analog as the internal standard, the assay was neither as precise nor as accurate.
INTRODUCTION
Carprofen (1, Table 1) is currently under clinical investigation for use as an anti-inflammatory agent. The compound has been found to possess analgesic, antipyretic and anti-inflammatory properties similar to indomethacin and aspirin but with fewer adverse side effects.' Procedures have been reported for the quantitation of 1 in blood using thin-layer chromatography (TLC), high performance liquid chromatography (HPLC) with either ultraviolet or fluorescence
Table 1. Chemical structures and their designations
detection and gas chromatography (GC)with electron capture (EC) d e t e ~ t i o n . ' ~ A mass spectrometric assay for 1 was required to confirm the specificity of the HPLC procedure, to assay samples containing low levels of carprofen and to apply certain stable isotope procedures to studies of the bioavailability and metabolism of ~arprofen.'.~ Initially an assay was developed which used a structural analog of carprofen, compound 3, as internal standard. Later, compound 2, a stable isotope analog of carprofen, was used for this purpose. Data gathered from the analysis of the same sample using either 2 or 3 as internal standard permitted a comparison of the assay precision and accuracy when either a structural or stable isotope analog was used as internal standard.
EXPERIMENTAL H Compound
R
Designation
d, /-2-(6-Chloro-SH-carbazol2 4 )propanoic acid
1
d, /-2-(6-Chloro-9H-carbazol2-~1)-3,3,3 trideuterio propanoic acid 3-(6-Chloro-SH-carbazol-2-yl) propanoic acid 2-(6-Chloro-9H-carbazol-2-yl) propanedioic acid dirnethyl ester 2-Trideuteriomethyl-2-(6-
2
C2H3 I -CH--COzH
3
-CH2-CH2-C02H
4
-CH-(C02CH3)2
I
-CH -C02H
5
chloro-9H-carbazol-2-yl)
propanedioic acid dirnethyl ester (6-Chloro-SH-carbazoI-2-yl)
CZH3
I
-C-(C02CH3)2 6
acetic acid methyl ester
' Department of Chemical Research.
Instrumentation
CH3
-CH2-C02-CH3
A Finnigan 3200 mass spectrometer was interfaced with the manufacturer's Model 9500 gas chromatograph and Promin@ selected ion monitor. A glass column (5 ft X 2 mm i.d.) was packed with 3% OV-17 on 100/120 mesh Gas Chrom Q purchased from Applied Science Laboratories. Before use, the column was conditioned overnight at 295 "Cwith a nitrogen flow of 10 ml min-'. GC operating temperatures were: column oven, 280285 "C; injection port, 300 "C; separator oven and transfer line, 250°C. Isobutane was used as both GC carrier gas and CI reactant gas. The ion source was operated without external heating at a pressure of 0.5 Torr. The ion source operating parameters were as follows: lens, -15 V; ion energy, +1OV; electron energy, -175 V; filament emission, 1 mA. The voltage across the continuous dynode electron multiplier was 2000 V. Each channel of the Promin@was operated at
CCC-0306-042X/79/0006-0325$03.00 BIOMEDICAL MASS SPECTROMETRY, VOL. 6, NO. 8, 1979 325
B. J. HODSHON, W. A. GARLAND, C. W. PERRY AND G. J. BADER
100 ms dwell time, 0.5 Hz frequency response and a sensitivity of lo9 V/A. Ion chromatograms were recorded on a four channel Rikadenki pen recorder.
Materials
Nanograde benzene, methanol and chloroform were obtained from Mallinckrodt Chemicals. One molar borate buffer, pH 10, and 5 M acetic acid buffer, p H 4.5 were prepared according to standard method^.^ Ethanol free ethereal diazomethane was prepared from Diazald@ (Aldrich Chemical Co.) using the manufacturer's suggested procedure. Twenty ml culture tubes (Kimble No. 45066-A) with Teflon@lined screw caps were used for sample extraction. Fifteen ml screw cap centrifuge tubes (Kimble 45 161) were used for final solvent evaporation. Extractions were performed by shaking the tubes for 20 min on an IEC No. 106234 bottle shaker. Centrifugations were performed at 10°C for 10 min at 1500Xg using a Damon CRU-500 refrigerated centrifuge. The solvents were removed at 60 "C under a gentle stream of nitrogen (N-Evap@, Organomation Assoc.). Aqueous solutions of 1, 2 and 3 were prepared separately. To the appropriate amount of compound, enough 1M, pH 10, borate buffer was added to maintain the compound in solution. The basic solution was titrated to p H 7.6 using 0.5 N HCI and the resulting solution was brought to volume with water. Using this solvation procedure, the following solutions were prepared for use as standards: 500 pmol per 100 pl of 1, 1000 pmol per 100 pl of 2 and 1000 pmol per 100 p1 of 3. Compounds 1 and 3 were obtained from Dr W. Scott, Hoffmann-La Roche. Compound 4 was prepared by methanolysis of the diethyl ester obtained as a gift from Dr W. Harry Mandeville, Hoffmann-La Roche. The synthesis of the deuterated internal standard, compound 2 from compound 4 is described below. d, 1-2-(6-Chloro-9H-carbazol-2-yl) propanoic acid (2). Sodium (253 mg, 11milli-equivalents) was dissolved in 10 ml of methanol and asolution of 4 (3.315 g, 10 mmol) in 60 ml of dioxane was added. The clear yellow solution was distilled to remove methanol until the boiling point of the distillate stabilized at 101 "C, at which point 4 was apparently in the form of its insoluble sodium salt. The suspension was cooled to 20 "C, trideuteriomethyl iodide [1.739 g, 0.763 ml, 12 mmol (Stohler Isotopes)] was added, and the mixture was stirred at room temperature for 22 h. Acetic acid (721 mg, 0.684 ml, 12 mmol) was added to neutralize the base. Additional dioxane was added and the mixture was filtered. Evaporation of the filtrate left a solid residue (3.54 g) which was crystallized from 15 ml of methanol to give a first crop, 2.37 g, m.p. 133-149 "C and a second crop, 194 mg, m.p. 125-142°C. These two crops were combined and recrystallized from toluene + hexane to afford 2.21 g of 5, m.p. 147-154 "C, which contained about 10-20% of a less polar impurity which was later identified as the N-methyl homolog of 5. Most of this material was used in the next step to prepare 2. The remainder was recrystallized three times from methanol, with attendant loss of material, to afford white crystals, 123 mg, m.p. 158-159 "C, which still included about 1-2% of the N-methyl homolog of 5. 326 BIOMEDICAL MASS SPECTROMETRY, VOL. 6, NO. 8, 1979
The diester 5 (1.744 g, 5 mmol) was dissolved in 25 ml of boiling methanol and 12.5 ml of 1N aqueous sodium hydroxide was added dropwise, causing 5 to precipitate as a fine suspension. With continued refluxing the precipitate dissolved, but within 5-10 min the boiling s o h tion became cloudy again. The mixture was refluxed overnight. A little water was added to redissolve a small amount of solid which had deposited and the clear hot soIution was treated with charcoal and filtered hot. The filtrate was carefully acidified (gas evolution) with 1 N hydrochloric acid, brought to a boil, and acetic acid added dropwise at the boiling point until all of the solids redissolved. Upon cooling the solution and seeding with 1, the crude product crystallized and was collected by filtration and washed with water. Recrystallization from toluene afforded two crops, 349 mg, m.p. 197-203 "C, and 228mg, m.p. 193-204"C, both still impure. Recrystallization from aqueous acetic acid and then from ethyl acetatethexane failed to remove all of a minor impurity (which was later identified as 6), but one small second crop (49 mg) of 2 was obtained which was >98% pure. The m.p. and IR and UV spectra of this compound were identical to that of authentic carprofen and its 'H NMR spectrum was consistent with that of the desired compound. By GCMS analysis, the compound was >99% *H3and showed chromatographic properties identical to authentic carprofen. Assay procedure
One ml of control blood, and the appropriate amount of 1 plus additional water to make l m l were added together for the standard curve samples. One ml of water was added to 1 m l of blood for the unknown biological samples. To all blood samples, standards and unknowns, were added 1000pmol of either 2 or 3 and the mixture was briefly vortexed. Three-tenths ml acetic acid buffer was added followed by 10 ml benzene; the tubes were capped and the samples were extracted. Eight ml of the benzene phase was transferred to a 15 ml conical extraction tube. Following evaporation of benzene, the residue was dissolved in 100 pl methanol. One ml ethereal diazomethane was added and the solution was allowed to stand at room temperature for 1h. This solution was evaporated and the residue was reconstituted in 50 pl chloroform. The mass spectrometer was set to monitor the [MH]+ ions corresponding to the methyl esters of 1 and 3 ( m / z 288), or 1 ( m / z 288) and 2 ( m / z 291). Two to lop1 of the final solution was injected. Approximately 60 s after injection the GC divert valve was turned off and at 75 s the ionizer was turned on. The retention time of 1-and 2-methyl esters under these conditions was approximately 140 s and the retention time of 3-methyl ester was approximately 180 s. For assays using 2 as internal standard the peak height ratios of m / z 288 to m / z 291 were calculated and converted to carprofen concentrations using a standard curve which was run with each experiment. The standard curve samples were prepared from spiked samples containing 0, 50, 100, 500, 1000, 2000 and 3000 pmol of 1 per ml blood and 1000 pmol of 2 per ml blood in each sample. The nonlinear standard curves were analyzed by a computer programmed for nonlinear @ Heyden & Son Ltd, 1979
DETERMINATION OF CARPROFEN IN BLOOD
288
"1
I
H
CH3
Mot. wt ~ 2 8 7
291
1t
+tJ Time
0 0 I
Mol.
wt =287 256 CH3 150
blood sam'ples from a subject 1 h after receiving a 100 mg oral dose of carprofen. The solid line is the tracing of rnlz 288 (carprofen and 3) and the dashed line the tracing of m l z 291 (2). Ion chromatogram (a) is the output recording using 2 as the internal standard in the assay. Ion chromatogram (b) is the output recording using 3 (smaller peak) as the internal standard in the assay. Both internal standards were added at a concentration of 1000 pmol. The carprofen level in this sample was approximately 1000 pmol.
I
..H
250
200
a 300
350
4 M
m/z Figure 1. lsobutane (0.5Torr) CI mass spectra of (a) carprofen ( l ) ,(b) the stable isotope internal standard (2) and (c) the structural analog internal standard (3).
regression analysis as described in the discussion. For assays using 3 as internal standard, the data from the standard curve samples were analyzed by linear regression analysis.
RESULTS AND DISCUSSION The isobutane CI mass spectra of the methyl esters of 1, 2 and 3 are shown in Fig. 1. As expected the spectra are quite simple consisting principally of [MH]+ molecular ions. The spectra of the methyl esters of 1 and 2 show small ions corresponding t o the loss of the elements of methyl formate. The spectrum of 3-methyl ester shows an ion corresponding to the loss of the elements of methanol. Loss of methyl formate from 3-methyl ester would generate an unstabilized primary carbonium ion, while the same loss from the methyl esters of 1 or 2 would yield a stabilized benzylic carbonium ion. Typical ion chromatograms from the analysis of the same blood sample using either 2 or 3 as the internal standard can be seen in Fig. 2. If 2 is used as the internal standard both the internal standard and carprofen elute @ Heyden & Son Ltd, 1979
Time
Figure 2. Ion chromatograms from the duplicate analysis of 1 ml
C ~ ~ C " , - C H * - IIc
50t
-
at the same time. This procedure has the theoretical disadvantage that two different ions must be monitored. It has been shown that such ion switching can introduce considerable error.' Using 3 as the internal standard the same ion can be monitored for both carprofen and the internal standard. This procedure has the disadvantage of requiring very exacting conditions in order to chromatographically separate carprofen and its internal standard. In this regard a large number of injections of an ethyl acetate extract of control blood were required at the beginning of each set of injections in order to decrease peak tailing and increase sensitivity. Since compounds 1 and 3 are not structurally identical, peak areas, calculated as the peak widths at half-height times peak heights, were used to determine ion intensities in assays using 3 as internal standard. Assay recoveries were determined from six, 1 ml, blood samples each spiked with 1000 pmol of 1 and six, 1 ml, blood samples each spiked with 1OOOpmol of 3. Based on a comparison of the ion intensities from these samples and the ion intensities from the injection of authentic 1- or 3-methyl ester, the recoveries ( fSD) were 77 f 26% for 1 and 2 and 44* 11% for 3. Recoveries of 100 pmol of 2 from blood were similar. A typical standard curve relating the amount of carprofen added to plasma to the resulting ion ratios using 2 as internal standard can be seen in Fig. 3. As expected the curve is not a straight line.'-'' The nonlinearity is due to the contribution of the I3Cisotope BIOMEDICAL MASS SPECTROMETRY, VOL. 6, NO. 8, 1979 327
B. J. HODSHON, W. A. GARLAND, C. W. PERRY AND G. J. BADER
cedure the fit was not as good. The percent difference (*SD) was 25 15,26&16,10* 3,8 f 1,4*4 and 8 f 3 for the same six concentrations. The mean percent difference in this case was 13.5f9.5. When the structural analog 3 was used as internal standard, the ion ratio of the m / z 288 corresponding to 1-methyl ester to m / z 288 corresponding to 3-methyl ester was determined using the peak area of each ion. The standard curve data was plotted as the observed ion ratios versus spiked sample concentrations. The unknown ratios of 1-methyl ester to 3-methyl ester were converted to pmol of 1using the slope and intercept values obtained from linear regression analysis. Figure 4 is a typical standard curve using 3 as internal standard. Based on the results from several repetitions, the standard curve using 1000pmol of 3 as internal standard was linear to approximately 2000 pmol. The correlation coefficient *SD for 14 standard curves of this type was 0.996 0.004. The sensitivity limit in the assay procedure using 2 as internal standard is illustrated in Fig. 5. As shown in (a), adding 100 pmol of 2 to control blood gave a virtually blank ion chromatogram at the ion corresponding to 1 as compared with the ion chromatogram from the analysis of 1 ml blood spiked with 5 pmol of 1and 100 pmoles of 2 (b). As can be seen, the ion current due to 1-methyl ester in (b) is considerably above the background noise level. To determine assay accuracy and precision, a large volume of the same control blood used to generate the standard curve was spiked to give a concentration of 200 pmol 1 m1-l. One ml aliquots were analyzed using 2 or 3 as internal standard in separate experiments. The
*
*
pmol I
Figure3. Typical standard curve relating the ion ratio of rnlz 288 to m / z 291 to the amount of 1 added to 1 ml of blood. The line is a 'least square' computer fit using program NONLIN.
peak of the 37Cl isotope peak of unlabeled carprofen. The nonlinearity could be avoided by monitoring the 37 C1 isotope (mlz 293) of the deuterated internal standard. However, interfering peaks in the m / z 293 ion chromatograms precluded the use of this mass. The nonlinear standard curve data obtained using 2 as internal standard were analyzed using computer program NONLIN12 to fit the ion ratio concentration data to the equation R = (x + A ) / ( B x+ C).13314 In this equation C is approximately equal to the amount of [2H3]carprofen added, B is approximately equal to the ratio of m / z 291 to m / z 288 in the mass spectrum of 1 and A is approximately equal to C times the ratio of m / z 288 to m / z 291 in the mass spectrum of 2. Once found, the values of A, B and C are used in the rearranged expression x = ( A- R C ) / ( R B- 1) to find an unknown x given an experimentally determined R. For three standard curves the means (*SD) of constants A, B and C using 1000 pmol of 2 as internal standard were 5.35 2.26 pmol, 0.086 f 0.013 and 984* 51 pmol, respectively. Theoretical values for A , B and C are Opmol, 0.05715 and 1000 pmol. A comparison can be made between the known concentration values and the concentration values generated from the experimentally determined ion ratios using the computer calculated standard curve parameters. The percent differences (*SD) between the expected and observed values for three standard curves generated using the NONLIN procedure were 9 * 5 , 8 * 1 3 , 3*2,2* 1 , 4 * 4 and 2* 1 for the 50,100, 250, 500, 1000 and 3000 pmol concentration points, respectively. The mean percent difference ( f SD)for the six concentration points was 4.7*3.1. If the same data were analyzed using a conventional least square pro-
*
328 BIOMEDICAL MASS SPECTROMETRY, VOL. 6, NO. 8, 1979
I
I
1
500
loo0
2ooc
pmol I
Figure 4. Typical standard curve relating the ion ratio of m l z 288 corresponding to 1 and mlz 288 cor'responding to 3 to the amount of 1 added to 1 rnl of blood.
@ Heyden & Son Ltd, 1979
DETERMINATION OF CARPROFEN IN BLOOD
Table 3. Drug blood levels from one patient administered orally 100mg of carprofen. Levels were obtained using either 2 or 3 as internal standard in separate experiments Concentration Hours after dosing
0 0.5 1 2 4 6 8 12 16 24 36 48
Time
Time
Figure 5. Ion chromatograms resulting from the analysis of 1 ml of control blood to which was added 100 pmol of 2 and either 0 pmol (a) or 5 pmol of carprofen (b).The solid line is the tracing of m / z 288. The dashed line is the tracing of m l z 291.
results are shown in Table 2. Using 2 as internal standard resulted in a mean ( *SD) carprofen concentration of 193*4pmol from 10 replicates. Using 3 as internal standard, the results were quite disappointing. The mean concentration of carprofen determined from 1 0 replicates was 27 1pmol with a SD of f 11. Although the precision is reasonable (*4%), the accuracy is off by 33% from the actual amount of compound known to be present in the sample. In separate experiments carprofen concentrations in a set of samples from one patient were determined using either 2 or 3 as internal standard. The results, Table 3, show a large difference, as much as loo%, in the concentration of 1 determined
Table 2. A large volume of blood was spiked with 1 to give a concentration of 200 pmol m1-l. One ml aliquots were assayed in separate experiments using either 2 or 3 as internal standard Concentration
i
-
of 1
Using stable isotope internal standard (2)
Using structural analog internal standard (31
190 186 188 192 196 193 198 195 198 195 193.10 SD = 4.09
265 269 269 263 260 266 278 296 286 266 = 271.80 S D = 11.41
x=
@ Heyden & Son Ltd, 1979
Using stable isotope internal standard (2)
pmol ~
ml
of 1
Using structural analog internal standard (31
0
0
2200 1932 12580 4468 2979 1751 1118 815 487 222 147
2812 2626 17170 10256 5961 3554 1360 812 900 463 21 1
using either of the two internal standards, As before, the concentrations were considerably higher when the structural analog was used as the internal standard. These same samples were then analyzed using the published HPLC p r o c e d ~ r e .The ~ concentrations determined by this assay showed a good correlation (correlation coefficient = 0.998, slope = 1.28h0.03) with the concentrations determined using 2 as internal standard and a poorer correlation (correlation coefficient = 0.953, slope = 1.81k 0.18) with the concentrations determined using 3 as internal standard. Thus, the use of 3 as internal standard gives less precise and less accurate results than when 2 is used as internal standard in spite of the advantage of not requiring mass switching. Much additional work would be needed to provide an explanation of why the ion ratios from the patient or control blood were not consistent with the ion ratios from the standard curve samples. Perhaps the GC properties of 1 and 3 are so different as to preclude good quantitation. ShatkayI6 and Roberson” have both commented on the potential unreliability of GC procedures using internal standards. Based on their work errors of the magnitude observed here may not be unexpected. In addition, the intensities of the fragment ions may be sensitive to small changes in ion source temperature^.'^'^^ As mentioned previously, 1 and 2 undergo a different fragmentation process than 3. These two fragmentation routes may show different sensitivities to temperature. In this regard our procedure was to analyze the standard curve samples and then the patient or reproducibility samples. Although unexplained, our results are consistent with comparisons of the use of labeled versus nonlabeled internal standards made by other investigators. Quantitative measurement of norepinephrine using deuterated norepinephrine as internal standard demonstrated good accuracy and precision, while use of a-methylnorepinephrine for the same purpose was unsatisfactory.2” In another study, the long-term variation in peak ratio was examined using either deuterated desipramine or maprotiline as the internal standard in a desipramine assay.’l The variance of the structural aflalog internal standard was three- to ten-fold of that the stable isotope analog internal standard. BIOMEDICAL MASS SPECTROMETRY, VOL. 6, NO. 8, 1979 329
B. J. HODSHON, W. A. GARLAND, C. W. PERRY AND G. J. BADER
Since the assay using 2 as internal standard provides an acceptable quantitiative procedure for our studies with carprofen, we are not planning any further experiments with 3 as internal standard. We feel,
however, that our experience with this structural analog internal standard will be of use to other investigators quantitating biologically important substances by mass spectrometry.
REFERENCES 1. L. 0. Randall and H. Baruth, Arch. lnt. Pharmacodyn. Ther. 220, 94 (1976). 2. J. A. F. de Silva, N. Strojny and M. A. Brooks, Anal. Chim. Acta 73,283 (1974). 3. C. V. Puglisi, J. C. Meyer and J. A. F. de Silva, J. Chromatogr. 136,391 (1977). 4. G. Palmskog and E. Hultman, J. Chrornatogr. 140,310 (1977). 5. J. S. Dutcher, J. M. Strong, S. V. Lucas, W.-K. Lee and A. J. Atkinson, Clin. Pharmacol. Ther. 22, 447 (1977). 6. H. d'A. Heck, N. W. Flynn, S. E. Buttrill, R. L. Dyer and M. Anbar, Biorned. Mass Spectrom. 5,250 (1978). 7. J. A. F. de Silva and C. V. Puglisi,AnalChem. 42,1725 (1970). 8. M. G. Lee and 6.J. Millard, Biorned. Mass Spectrom. 2, 78 (1975). 9. J. F. Pickup and K. McPherson, Anal Chem. 48, 1885 (1976). 10. J. R. Chapman and E. Bailey, J. Chromatogr. 89, 215 (1974). 11. J. F. Holland, R. E. Teets, M. A. Bieber and C. C. Sweeley, American Society for Mass Spectrometry 21st Annual Conference on Mass Spectrometry and Allied Topics, San Francisco, California (1973). Abstracts, p. 55. 12. C. M. Metzler, A Users Manual for NONLIN, Upjohn Co., Kalamazoo, Michigan 49001, USA. 13. 6. H. Min, W. A. Garland, K.-C. Khoo and G. S. Torres, Biorned. Mass Spectrom. 5,692 (1978).
330 EIOMEDICAL MASS SPECTROMETRY, V3L. 6, NO. 8, 1979
14. W. A. Garland, R. Muccino, B. H. Min, J. Cupano and W. E. Fann, Clin. Pharrnacol. Ther. 25,844 (1979). 15. J. H. Beynon, Mass a n d Abundance Tables for Use in Mass Spectrometry, p. 118. Elsevier, Amsterdam (1963). 16. A. Shatkay and S. Flavian, Anal. Chern. 49, 2222 (1977). 17. J. C. Roberson, Anal. Chern. 50, 2145 (1978). 18. F. Field, J. Am. Chern. SOC.91, 2827 (1969). 19. W. R. Sherman, W. J. Fies and E. J. Heron, American Society for Mass Spectrometry, 24th Annual Conference on Mass Spectrometry and Allied Topics, San Francisco, California (1976). Proceeding, p. 293. 20. R. C. Murphy, J. A. Zirrolli, K. Clay, S. E. Hattox and .?I Helbig, American Society for Mass Spectometry, 23rd Annual Conference on Mass Spectrometry and Allied Topics, Houston, Texas (1975). Proceedings, p. 427. 21. M. Claeys. S. P. Markey and W. Maenhaut, Biorned. Mass Spectrom. 4, 122 (1977).
Received 12 February 1979
@ Heyden & Son Ltd, 1979
@ Heyden & Son Ltd, 1979
