Volatile Organic Compounds Determined in Pharmaceutical

---

Anal. Chem. 1996, 68, 1317-1320

Volatile Organic Compounds Determined in Pharmaceutical Products by Full Evaporation Technique and Capillary Gas Chromatography/ Ion-Trap Detection Jan Schuberth

National Board of Forensic Medicine, Department of Forensic Chemistry, University Hospital, 581 85 Linko¨ ping, Sweden

Pharmaceutical products often contain volatile organic compounds (VOCs), which are made up of residual solvents from the manufacturing process and of flavoring additives. These substances may form a “signature” that perhaps could be used to reveal the product source. To study this possibility, a new method for detecting and quantitating VOCs in pharmaceutical preparations is described. It is based on extraction of the dry powder by the full evaporation technique, separation of the VOCs by gas chromatography in a capillary with an apolar stationary phase, and exposure of the compounds by ion-trap detection with the apparatus run in the full-scan mode. The search of some drug substances or pharmaceutical products for VOCs revealed ethanol, acetone, 2-propanol, methyl acetate, toluene, eucalyptol, and menthol, whose concentrations were in the range 0.008-26 mmol/kg of sample. The within-day or between-day precision studies showed, except for methyl acetate, a relative standard deviation less than 13%. The concentrations for the different compounds were at the limit of detection or of quantification in the range 0.4-4, respectively, 1-10 µmol/kg of sample. Based on the quantitative data, distinct signatures were obtained from synonymous medicines made by four diverse producers. These data indicated that the method provides a means for disclosing the origin of a drug product. A pharmaceutical preparation is often rather complex. It may, thus, contain not only the active drug substance and the vehicle but also volatile organic compounds (VOCs). Some of these stem from additives in the tablet to give it a pleasant flavor and color, while others come from residual impurities of the manufacturing process. Such chemicals are usually made up of solvents which have been used in the synthesis and crystallization of the drug substance and which can be difficult to remove completely by drying. Their presence in medicines creates some problems for the manufacturer and may also involve a health hazard for the user of the product. The quality and stability of a preparation could, thus, be impaired by solvent residues.1 A number of VOCs, i.e., methylene chloride, chloroform, trichloroethylene, benzene, and dioxane, are also classified as potential carcinogens, and the upper allowable concentrations of these in a drug substance are, therefore, officially ruled.2 The presence of VOCs in a drug product may, however, also yield some forensic possibilities, i.e., to enable the detection of 0003-2700/96/0368-1317$12.00/0

© 1996 American Chemical Society

its illicit manufacturing and distribution. It has thus been proposed3 that solvent residuals might form a “fingerprint” or “signature” of a preparation and that such a marker could be suitable for identifying the product source. This type of fingerprinting approach seems, however, not to have been tried and reported in the literature. A reason for this might be that only methods for directed analysis of VOCs in a drug product are available, but no screening techniques are known that are applicable for searching for the volatile “general unknown”. Even the current targeted tests, which are based on gas chromatography (GC) after direct liquid injection,2,4 or alternatively, after extraction by static headspace (HS)2,3,5-9 or purge and trap sampling,10 are not free from problems. The main problem is caused by the introduction of volatile impurities into the sample from the solvent used to dissolve the drug sample before analysis.4 Moreover, when the VOCs in a solubilized sample are extracted by HS, the yield is highly dependent on the matrix, and a quantitative determination is, therefore, not easily done.9 These problems could be overcome by the use of the full evaporation technique (FET), a rather new extraction method.11 It is a variant of headspace at which a small amount of the sample, without being dissolved, is equilibrated at a high temperature, allowing all the analyte to be released from the matrix into the headspace phase. This means that the standard graphs can be simply generated from measurements on small volumes of the standard solution in the headspace vials, kept at a temperature high enough to fully evaporate both the analyte and the matrix. The only main restriction of the FET is that the equilibration temperature must not exceed the temperature for the decomposition of the sample components. (1) Murthy, K. S.; Ghebre-Sellassie, I. J. Pharm. Sci. 1993, 82, 113-126. (2) Organic volatile impurities. The United States Pharmacopeia XXII, National Formulary XVII, 6th Supplement; The United States Pharmacopeial Convention: Rockville, MD, 1993; pp 2927-2929. (3) Mulligan, K. J.; McCauley, H. J. Chromatogr. Sci. 1995, 33, 49-54. (4) Clark, L.; Scypinski, S.; Smith, A.-M. Pharmacopeial Forum 1993, 19, 50675074. (5) Dennis, K. J.; Josephs, P. A.; Dokladalova, J. Pharmacopeial Forum 1992, 18, 2964-2972. (6) Kidd, W. C., III. Pharmacopeial Forum 1993, 19, 5063-5066. (7) Guimbard, J. P.; Person, M.; Vergnaud, J. P. J. Chromatogr. 1987, 403, 109-121. (8) Penton, Z. J. High Resolut. Chromatogr. 1992, 15, 329-331. (9) Kumar, N.; Gow, J. G. J. Chromatogr. A 1994, 667, 235-240. (10) Wampler, T. P.; Bowe, W. A.; Levy E. J. J. Chromatogr. Sci. 1985, 23, 6467. (11) Markelov, M.; Guzowski, J. P., Jr. Anal. Chim. Acta 1993, 276, 235-245.

Analytical Chemistry, Vol. 68, No. 8, April 15, 1996 1317

Even if mass spectrometry (MS) is used, the identification in a drug product of a VOC detected in trace amounts at the screening procedure may be difficult. A reason for this is that low-molecular-mass compounds often yield very nonspecific mass spectra. The identification of an unknown analyte of this type, therefore, usually has to be based upon both its mass spectrum and its retention index.12 To help the screening procedure by facilitating interlaboratory comparison of retention indices, the separation of the VOCs is preferably done in a capillary with an apolar stationary phase12,13 rather than, as earlier reported for the analysis of VOCs in drug formulations,2-10 in a capillary or packed column with various types of polar stationary phases. In the work to be presented here, I will describe a method suitable both for screening a pharmaceutical product for VOCs and for quantifying these. The extraction is done by the FET, the separation by GC in a capillary with an apolar stationary phase, and the detection by ion-trap detection (ITD). Data on the limits of detection (LOD) and of quantitation (LOQ), as well as on the precision of within-day and between-day measurements on pharmaceutical preparations are given. Finally, based on the quantitative data, the signatures of the same type of drug product delivered by four different manufacturers, or the signatures of the same drug product delivered at a 4-year interval by a single manufacturer, are presented as three-dimensional plots. EXPERIMENTAL SECTION Materials. The drug substances, risperidone and piroxicam, were obtained from two of the drug companies marketing pharmaceutical products with these compounds. Tablets with ciprofloxacin or spironolactone were purchased from the official pharmacy. Two different batches of ciprofloxacin tablets (250 mg) were used, both manufactured by the same company but at different times. The expiration dates of the products were September 1995 (batch A in Figure 3) and October 1999 (batch B in Figure 3). The spironolactone tablets (50 mg) were made by four different companies (manufacturers A-D in Figures 1 and 3). The solutions used for generating the standard graphs contained, per liter of water/1% methanol, 10 mmol of ethanol, 2 mmol each of acetone and 2-propanol, and 1 mmol each of methyl acetate, eucalyptol, and l-menthol, or, per liter of methanol, 1.2 mmol of toluene. Instrumentation. The analyses were done on the dry powder of a drug substance or of a tablet that had been crushed in a mortar. Sample aliquots were weighed in a closed headspace vial with a balance (Sartorius, MC1 Research RC 210 P) capable of monitoring weights down to 10 pg. The headspace gas was embraced in a 21.4-mL headspace vial (Apodan, Copenhagen, Denmark, V. No. 092357), sealed with a Teflon-lined septum in a crimp cap. The headspace vial was put into an 18906B accessory kit for constant heating time, hooked up to a Hewlett-Packard Model 19395A autosampler. The GC was done with a HewlettPackard Model 5890 and with the separation in a DB-1 capillary (30-m × 0.25-mm-i.d., coated with 1 µm of methylsiloxane) from J&W Scientific (Folsom, CA). The capillary was inserted without a flow restrictor directly into the ion source of a Finnigan MAT Model ITD800 ion-trap detector. The tuning of the device was done manually to resolve the m/z 69 and 70 peaks and the m/z (12) Schuberth, J. J. Chromatogr. A 1994, 674, 63-71. (13) Evans, M. B.; Haken, J. K. J. Chromatogr. 1989, 472, 93-127.

1318

Analytical Chemistry, Vol. 68, No. 8, April 15, 1996

Table 1. Limit of Detection and Quantitation of Substances Determined by FET/GC/ITD total analyte amount (nmol) in the test sample substance

m/z

LODa

LOQa

ethanol acetone 2-propanol methyl acetate eucalyptol l-menthol toluene dichloromethane chloroform trichloroethylene benzene dioxane

31 + 45 43 45 43 71 + 81 + 93 71 + 81 + 95 91 + 92 49 + 84 + 86 83 + 85 60 + 95 + 97 78 58 + 88

0.208 0.068 0.135 0.086 0.027 0.157 0.019 0.039 0.044 0.049 0.025 0.114

0.600 0.196 0.390 0.249 0.078 0.453 0.054 0.111 0.127 0.142 0.072 0.327

a The LOD, equal to 3 times, and LOQ, equal to 10 times the standard deviation of the background noise, were calculated according to Knoll.14

131 and 132 peaks. Evaluation of the raw data was carried out with Datamaster II (program version 1.3, Finnigan MAT). Analytical Procedures. The same instrumental operations for the HS, GC, and ITD were used as earlier described12 with the following exceptions: 130 °C (HS equilibration temperature), 28 min (HS equilibration time), 210 kPa (HS auxiliary gas pressure), and 150 °C (GC injector temperature). Criterion of Full Evaporation. The theoretical background of the FET, which has been described in detail,11 rests on the equation, Cg ) C0Vc(KVr + Vg)-1, where C0 is the analyte concentration in the sample, Vc the sample volume before the equilibration, Cg the analyte concentration in the gas phase, Vr the condensed sample volume after the equilibration (Vr ≈ Vc), Vg the volume of the gas (headspace vial volume), and K the partition coefficient. When the condition of full evaporation is met, KVr should be insignificant as compared with Vg; Cg will then become independent of the temperature and linearly related to the sample size. Estimation of the LOD and LOQ. To assay the LOD and LOQ for the different VOCs, the height of the largest noise peak was measured at the appropriate mass number or combination of mass numbers (Table 1) in a preselected retention time interval of chromatograms from empty headspace vials. This was equal to 100 multiples of the width of the calibration peak at one-half peak height. From these data, the peak height (hblank), equal to 3 times (LOD) or 10 times (LOQ) the standard deviation of the gross blank signal, was calculated.14 To account for variations of the hblank, the mean hblank ( SD was measured in seven experiments, and the minimum peak height needed for the identification or quantitation of a VOC (Table 1) was set at hblank + 3SD. The analyte concentration that gave rise to this signal, i.e., the LOD or LOQ, was finally calculated from the peak height of the calibration sample. Precision Studies. To evaluate the within-day and betweenday precision, double determinations were done on six different days on samples prepared each day from tablets of ciprofloxacin and of spironolactone. For each day, the variance and the mean value of the results were calculated. The within-day standard deviation was determined from the average of the six daily (14) Knoll, J. E. J. Chromatogr. Sci. 1985, 23, 422-425.

Figure 1. Reconstituted mass chromatograms showing some of the VOCs in spironolactone tablets from four different manufacturers (A-D). The numbers on the x-axis indicate the retention time, and those on the y-axis the mass number or combination of mass numbers selected. The chromatogram of each pharmaceutical product is separated into two parts, at which the different mass numbers in the left or right trace have been chosen to focus on compounds with short and long retention times, respectively. Peaks: 1, ethanol; 2, 2-propanol; 3, eucalyptol; 4, l-menthone (tentative identification); 5, l-menthol; and 6, menthyl acetate (tentative identification).

Figure 2. Peak areas of VOCs versus the amount of tested drug substance or pharmaceutical product.

variances, and the between-day standard deviation was determined from the variance of the six daily mean values.15 RESULTS Figure 1 shows the total ion current and the reconstituted mass chromatograms of spironolactone tablets from four different manufacturers. The left-hand traces, monitoring peaks at scan numbers between 180 and 800, show some volatile impurities, while the right-hand traces, examining peaks in the scan range 800-2400, reveal the supplementary volatile compounds. (15) Bookbinder, M. J.; Panosian, K. J. Clin. Chem. 1986, 32, 1734-1737.

Table 1 lists the mass numbers or combination of mass numbers that were used for detecting or quantitating the VOCs in some pharmaceutical products. The table also shows the minimum amount of the compounds needed in the headspace vial for these tests. Even though dichloromethane, chloroform, trichloroethylene, benzene, and dioxane have not been found in the drug substances or tablets investigated, these VOCs have been included in the table since they are regulated with official tolerance levels.2 To see if the criterion of full evaporation was met in accord with the equation given in the Experimental Section, the signal response at different sample masses of some drug substances or Analytical Chemistry, Vol. 68, No. 8, April 15, 1996

1319

Table 2. Precision of Repetitive Determinations of VOCs in Tablets by FET/GC/ITD relative standard deviation (%) substance

mean value (mmol/kg)

within-day (n ) 6)

between-day (n ) 6)

ethanol acetone 2-propanol methyl acetate eucalyptol menthol toluene

3.2 0.008 0.016 0.046 0.009 25.8 0.026

5.1 8.5 9.7 23.1 9.1 3.9 4.3

12.8 5.8 7.5 34.5 8.8 4.6 12.9

Figure 3. Three-dimensional plots of the concentrations of VOCs in pharmaceutical products. The upper plot shows the VOCs in different batches of the same medicine produced at a 4-year interval. The lower plot shows the VOCs in the same type of medicine made by four different manufacturers.

tablets was studied. As shown in Figure 2, the data indicated that the peak area of the VOCs studied related linearly with the sample mass. Data on the within-day and between-day precision for the determinations of VOCs in ciprofloxacin (batch B) and spironolactone (manufacturers A and D) tablets are shown in Table 2. Figure 3 presents three-dimensional plots of the concentrations for the different VOCs in ciprofloxacin and spironolactone tablets. The upper plot shows the results from the analysis of ciprofloxacin tablets purchased in 1992 (batch A) and in 1995 (batch B), and the lower one shows those from the testings of spironolactone 1320

Analytical Chemistry, Vol. 68, No. 8, April 15, 1996

tablets supplied by four different manufacturers (A-D). As can be seen, the concentrations of the VOCs in the ciprofloxacin elicited rather small differences in their concentrations, whereas the spironolactone tablets showed marked dissimilarities in their contents of VOCs. DISCUSSION The sensitivity and precision of the method described here for analyzing VOCs in drug products have been studied. When the data on the total analyte amount in the test sample, as shown in Table 1, were recalculated on a sample mass of 50 mg, the concentrations for the listed VOCs at the LOD amounted to 0.4-4 µmol/kg and at the LOQ to 1-10 µmol/kg. These LOD values were about 10 times lower than those obtained with HS/GC/MS3 (quadrupole mass spectrometer in scanning mode), and the LOQ data were well below the official tolerance concentrations (4001300 µmol/kg) for the controlled VOCs. The within-day or between-day precision for measuring VOCs in pharmaceutical products seemed also to be sufficient, except for methyl acetate. No explanation for the imprecision of the testings for this VOC can be given. The present data compared well with the corresponding within-day precision for the assay by HS/GC of 13 644 ppm (297 mmol/kg) of ethanol and 1177 ppm (20 mmol/kg) of acetone in cephalosporin raw material7 or that by HS/GC/MS (quadrupole mass spectrometer in selected ion monitoring mode) of 1.4 mg/kg (0.02 mmol/kg) of acetone in sulfamethazine (spiked sample).3 In the present study, the analytes have been extracted from the dry pharmaceutical sample by the FET, separated from one another by GC in a capillary with an apolar stationary phase, and monitored by ITD with the mass spectrometer run in the full scan mode. This combination of methods created a system suited both for the unbiased search for the volatile “general unknown” and for the identification and quantitation of low concentrations of the detected compound. Owing to its versatility, the procedure should be ideal for revealing the “signature” of a drug preparation in terms of its content of volatile flavoring agents and impurities from the manufacturing process. Figure 3 also shows distinct signatures of four medicines having the same pharmacologically active component and the same amount of it, but being delivered by diverse producers. That these markers differed with regard both to the volatile impurities and the components of the flavoring agent indicated that the manufacturing process, as well as the peppermint additive, varied substantially between the suppliers. On the other hand, different batches of the same medicine made by one manufacturer at a 4-year interval yielded rather similar signature VOCs. The method may, thus, offer an efficient forensic means for disclosing the production and distribution of an illict copy of a pharmaceutical product. This implies, though, that the between-batch variations of the VOC concentrations in a given legal drug formulation are sufficiently small. Considering the versatility and good precision of the method, it may also be useful for the authorities supervising the quality of drug preparations.

Received for review October 31, 1995. Accepted January 23, 1996.X AC951084A X

Abstract published in Advance ACS Abstracts, March 1, 1996.