A Generic Method for the Analysis of Residual Solvents in Pharmaceuticals Using Static Headspace-GC-FID/MS

Transcription

1 A Generic Method for the Analysis of Residual Solvents in Pharmaceuticals Using Static Headspace-GC-FID/MS Application Note Pharmaceuticals Authors Karine Jacq, Frank David, and Pat Sandra Research Institute for Chromatography, Pres. Kennedypark 26, B-8500 Kortrijk, Belgium Matthew S. Klee Agilent Technologies, Inc Centerville Road Wilmington, DE USA Abstract The determination of residual solvents in pharmaceuticals is one of the most important gas chromatography (GC) applications in quality assurance/quality control (QA/QC) in the pharmaceutical industry. Sample introduction is normally done using static headspace (SHS). In routine QC, GC with flame ionization detection (FID) is preferred, while mass spectrometry (MS) can be used for screening and identification. In this application note, a retention-time locked GC-MS/FID method is presented that allows the analysis of more than 50 solvents in a single run. The method was optimized to allow identification and quantification of all three International Conference on Harmonisation (ICH) classes of solutes at relevant concentration levels.

2 Introduction The determination of residual solvents (RS), formerly called Organic Volatile Impurities (OVI), in pharmaceutical products is probably the most important application of gas chromatography (GC) in pharmaceutical quality control. Recently, methods described in U.S. and European Pharmacopoeia have been reviewed, updated, and harmonized according to International Conference on Harmonisation (ICH) guideline Q3C (R3) [1]. In pharmaceutical manufacturing, approximately 60 different solvents are in typical use. This set of solvents covers a rather large range of boiling points and polarities. According to the ICH guideline, these solvents are divided into three classes. Class 1 includes benzene, carbon tetrachloride, 1,1-dichloroethane, 1,2-dichloroethylene, and 1,1,1-trichloroethane; these solvents are toxic and their use should be avoided. Class 2 solvents are less toxic, but their use should also be limited. Class 1 and Class 2 solvents are preferably being replaced by Class 3 solvents, which have low toxic potential to humans. Taking into account their relative toxicity, these solvents should be monitored in pharmaceutical products, including drug substances (or active pharmaceutical ingredients [API]) and drug products (formulations) at various levels, ranging from 2 ppm (2 µg/g drug substance) for benzene (Class 1) to 0.05 % (w/w = 5,000 ppm) for Class 3 solvents. Consequently, the analytical method(s) used to monitor these residual solvents in pharmaceutical products needs also to cover this range. For the analysis of residual solvents, gas chromatography in combination with flame ionization detection (GC-FID) is normally used. Sample preparation and introduction is done by static headspace [2]. In this way, the (mostly) volatile solvents are introduced selectively and the analytical system (inlet, column, and detector) is not contaminated by the (mostly) nonvolatile drug substance or drug product. For the separation, a thick film, medium polar column (for example, G43) is selected. Quantification is done versus an external standard. Excellent quantitative data, including low limits of detection (LODs) high repeatability and excellent linearity were obtained using the Agilent G1888 static headspace sampler in combination with an Agilent 7890 GC [3]. However, since no column can guarantee a unique retention time for a given solvent, confirmation analysis by GC-FID on a capillary column coated with a different stationary phase (for example, G16) is performed. More recently, GC-MS has been successfully used for confirmation/identification purposes [4,5]. In this application note, a system configuration and operation conditions are described that allow the analysis of 56 solvents in a single run. Some solvents listed in the ICH guideline are not volatile (enough) and not amenable to SHS-GC analysis. Examples are formic acid, acetic acid, dimethyl sulfoxide (often used as a method solvent), formamide, ethylene glycol, and sulfolane. The analysis of these impurities should be performed using other methods; their analysis is not discussed here. The presented retention-time locked method, however, can be considered as generic, since it covers most solvents and can be used both for identification (by MS) and for quantification (by FID and/or MS) of residual solvents across a wide concentration range. Experimental Sample Preparation For the analysis of residual solvents in pharmaceutical products, the drug substance or drug product is typically dissolved in a low-volatility (high-boiling) solvent such as dimethyl sulfoxide (DMSO), dimethyl acetamide (DMAC), or 1,3-dimethyl- 2-imidazolidinone (DMI). For water-soluble drug substances, dissolution in water can also be used. In this work, 100 mg drug substance was dissolved in 2 ml DMSO or DMSO/water (1:1). Recently dedicated "GC headspace" -grade solvents were made available from Sigma-Aldrich (NV/SA Bornem, Belgium). DMSO, suitable for GC-HS (cat. no ) was used in this work. Solvent standard solutions were prepared in DMSO at 15 µg/ml concentration for Class 1 solvents and at 600 µg/ml concentration for Class 2 and Class 3 solvents. From these stock solutions, aliquots of 1 to 100 µl were added in a standard 20 ml HS vial filled with 2 ml DMSO/ water (1:1, v/v). The concentrations of standards are always expressed in microgram per gram of drug product or drug substance. So, the concentration of these calibration solutions ranges from 0.15 to 15 ppm (µg/g drug) for Class 1 and from 6 to 600 ppm for Class 2 and Class 3 solvents, if 100 mg product is weighed in the HS vial. (The actual concentration of standards in the vial, expressed in µg per ml solvent, is 20 times lower.) Instrumental Conditions The samples were analyzed using the SHS-GC-FID/MS configuration presented in Figure 1. Static headspace was performed using a G1888 HS autosampler. The transfer line of the headspace sampler is coupled to a standard split/splitless inlet. Separation was done on a DB-1301 column. The column effluent is split using a purged splitter Capillary Flow Technology device to FID and MS (5975 MSD). The vial pressure is regulated by an AUX EPC channel. The purge at the splitter is also regulated by the AUX EPC (second channel). A 63 cm 0.1 mm id deactivated fused silica capillary was used to connect the splitter to the MSD; a 40 cm 0.1 mm id capillary was used to connect the splitter to the FID. Flows in both capillaries are approximately 1.4 ml/min and the retention time is also similar (small offset between FID and MS retention times). The analytical and selected ion monitoring (SIM) parameters are summarized in Table 1 and Table 2, respectively. 2

3 Figure 1. SHS-GC-FID/MS system configuration. Table 1. Analytical Parameters SHS (G1888 Agilent HS autosampler) Loop size: 1 ml Vial pressure: 14 psig (96.5 kpa) Headspace oven: 80 C Loop temperature: 120 C Transfer line temperature: 120 C Equilibration time: 10 min, high shake Pressurization: 0.15 min Vent (loop fill): 0.5 min Equilibration time: 0.1 min Inject: 0.5 min GC (7890A) Inlet: Results and Discussion Split/splitless, split 1/10, headspace liner (P/N ) Inlet temperature: 250 C Split ratio: 1:10 Carrier gas: Helium Inlet pressure: 160 kpa * AUX pressure (at splitter): 60 kpa Column: DB m 0.18 mm 2 µm (J & W) Oven: 40 C (5 min) 5 C/min 80 C 10 C/min 200 C (2 min) FID: 300 C, 40 ml/min H 2, 400 ml/min air MS (5975 C Inert XL MSD) Transfer line: 300 C Mode: Simultaneous scan/sim SCAN range: m/z SIM: See Table 2 * Retention time locking was applied. The column head pressure was adjusted to elute toluene at min. Column Selection Many different columns can be used for the analysis of residual solvents after static headspace extraction. Typically, a column coated with a thick film of a medium polar stationary phase is selected. A classical column for residual solvent analysis is a 30 m 0.53 mm id coated with 1 to 3 µm DB-624 (β = 44). On this column, good separation is obtained in about 30 minutes of analysis time [3]. Table 2. SIM Parameters Group Start time Ions , , 43, 45, , 43, 45, 58, 59, 61, 74, , 43, , 61, 73, , 57, , 42, , 43, 61, 72, , 45, 47, 59, 72, , 56, 61, 84, 97, , 45, 49, 61, 62, 74, 76, 78, , , 56, 95, , , 58, 61, , 59, , 43, 52, 55, 58, 73, 79, , 55, , 58, , 56, , 91, 106, , 78, 87, 105, 108, , 98, , 104, 132 In this work, a narrow-bore, thick-film column with equivalent stationary phase, DB-1301 (equivalent to G43), was selected. The lower phase ratio (β = 22) results in better resolution for the first eluting (but frequently detected) solvents, such as methanol, ethanol, diethylether, acetone, isopropanol, and acetonitrile. It is not possible to find a stationary phase that is able to separate all 60 solvents in approximately 30 minutes. In most practical cases, however, only three to five solvents need to be determined. The described method allows the analysis of 56 solvents and eventually the same column can also be used for the analysis of ethylene oxide. For this reason, column selection and separation conditions can be considered as generic. 3

4 The separation obtained by SHS-GC-FID/MS for a solvent test mixture in DMSO/water is shown in Figure 2. The concentration of the solutes was 16 ppm for the five Class 1 solvents and 560 ppm for the Class 2 and Class 3 solvents. With one injection, three data files are obtained. The upper trace shows the total ion chromatogram (TIC) obtained by MSD in scan mode, the middle trace shows the corresponding SIM data, and the lower chromatogram is the FID trace. Toluene elutes at minutes in all three chromatograms and the offset in retention time between FID and MS is less than 0.03 minute or 2 seconds for all solutes. A detailed view of the separation is shown in Figure 3, showing three elution windows of the FID trace. All solutes can be detected and are labeled on the chromatograms except for tetrachloromethane (trace at minutes), 2-methoxy ethanol (co-elutes with benzene, 1,2-dimethoxyethane, and 1,2-dichloroethane), 2-ethoxyethanol (trace at minutes), DMAC (co-elutes with cumene), and 1-methyl-2-pyrrolidone (trace at minutes). Validation In a previous application note [3] it was demonstrated that the quantitative data obtained using a G1888 SHS 7890A GC combination resulted in equal or better quantitative results than a G GC combination. Electronic pressure control of carrier gas and vial pressure results in excellent repeatability and linearity. An optional pneumatic control module can be used to control the vent pressure by back pressure regulation (BPR), resulting in increased sensitivity (by a factor of 2 to 4) and smaller relative standard deviations (RSDs) (reduced by a factor of 2). In this study, good quantitative data were obtained even though the BPR approach was not used. Using BPR would provide higher performance for those seeking it. Limits of detection, linearity, and repeatability were evaluated with the different modes of detection. Repeatability was checked at 3 ppm level (µg/g, equivalent to 0.15 µg/ml in vial) for Class 1 solvents and at 100 ppm for the others (n = 6). Five-point calibration curves (plus a blank) were measured between 0.15 and 15 ppm for Class 1 solvents, and between 6 and 600 ppm for most others. For some solutes giving low response, linearity was tested in the 500 to 5,000 ppm range. For coeluting compounds, the experiments were performed with single compound solutions in order to allow peak integration with FID. The results are summarized in Table 3. Column 4 in Table 3 indicates possible coelution of target compounds. Most solutes are chromatographically resolved and can thus be quantified by either FID or MS when present in the same sample. In some cases, coelution is observed (labeled with C). In these cases, quantification is still possible by MS after ion extraction or by using selected ion monitoring (SIM) mode. In columns 5 through 8, the limit of detection (LOD) for the three detection modes, determined from the lowest calibration level at S/N = 3, are compared to the ICH limits. In most cases (43 of the 56 analytes), the LODs were well below the ICH limits for all three detection modes. Those instances wherein the LOD was above the ICH limit are highlighted in Table 3 in boldface. For the Class 1 solvents, it is clear that MS in SIM mode is preferred. The benefit of using SIM is illustrated in Figure 4a for some Class 1 solvents. At 10.6 minutes, 1,1,1-trichloroethane coelutes with cyclohexane, as seen in the FID trace. However, both solutes can be accurately measured with extracted ion chromatograms using m/e = 97 for 1,1,1-trichloroethane and m/e = 56 for cyclohexane. The same can be observed in Figure 4b for the overlapped peaks of benzene, 1,2-dimethoxyethane and 1,2-dichloroethane at 11.5 minutes in the FID trace and the extracted ion peaks of m/e = 78 for benzene, m/e 45 for 1,2-dimethoxyethane, and m/e = 62 for 1,2-dichloroethane (2-methoxyethanol is not detected at this level). Using MS, all compounds, except 2-methoxyethanol can be detected at LOD < ICH limit. Some of the other more polar solutes (2-ethoxyethanol, DMF, DMAC, and 1-methyl pyrrolidone) also give a low response in FID, whereas MS detection limits are satisfactory. As reported before [3], back-pressure regulation of the vent pressure is expected to help decrease the LOD by a factor of two, so this an important consideration if you are using FID for quantitation. In addition, increasing sample size and/or injection volume (headspace sample loop size) would also help if you are doing quantitation with FID. The repeatability of the SHS-GC-FID/MS method is excellent. RSDs were mostly below 5 percent, both for GC-FID and GC- MS (SIM and SCAN). The average RSDs were 4.4 percent for SCAN, 3.8 percent for SIM, and 3.0 percent for FID. Again, the values were higher for some more polar solutes. Finally, good linearity was obtained for most compounds. Except for 2-methoxyethanol, 2-ethoxyethanol, DMF, DMAC, and 1-methyl-2-pyrrolidone (compounds with higher LODs), linearity was excellent with the three types of detection (R² > 0.99). 4

5 Abundance Abundance Abundance Figure 2. A 56-component solvent mixture analyzed by SHS-GC-FID/MS. 5

6 Response Methanol Pentane Ethanol Ethyl ether 1,1-dichloroethene Acetone 2-propanol + ethyl formate Acetonitrile Methyl acetate Dichloromethane Z-1,2-dichloroethene + tert-butyl methyl ether Hexane 1-propanol Nitromethane E-1,2-dichloroethene 2-butanone Ethyl acetate 2-butanol - THF Chloroform 1,1,1-trichloroethane Cyclohexane Response Isobutyl alcohol Benzene +1,2-dimethoxyethane + 1,2-dichloroethane Isopropyl acetate Heptane 1-butanol Trichloroethene Methylcyclohexane 1,4-dioxane Propyl acetate 4-methyl-2-pentanone Isoamyl alcohol + pyridine Toluene (Octane) Isobutyl acetate 1-pentanol 2-hexanone Butyl acetate Response Chlorobenzene M-xylene P-xylene O-xylene DMSO Anisole Cumene Tetralin Figure 3. Detailed FID chromatogram (16 ppm Class 1 and 560 ppm Class 2 and Class 3 solvents). 6

7 Solvent Influence The influence of the solvent used to dissolve the drug substance or drug product was also evaluated. In general, equally good quantitative data in terms of linearity and repeatability are obtained when using water, DMSO, DMAC, DMI, or mixtures thereof. The LOD (and slope of the calibration curve), however, depends on the solvent used. In general, for apolar solvents in the most polar matrix (water), the lowest LODs will be obtained. For the (apolar) Class 1 solvents, the LOD can be up to 10 lower (more sensitive) if static headspace is performed in DMSO/water versus pure DMSO. Matrix Influence The presence of a relatively high concentration of drug substance in the sample solution (100 mg/2 ml) might influence the absolute response of the solutes. To evaluate that poten- tial with this method, some SHS-GC-FID/MS analyses were performed on solutions of 100 mg drug substance (promethazine) in the solvent (2 ml DMSO/H 2 O). A blank (not spiked) and two spiking levels (for example, 3 and 8 ppm for Class 1 solvents, and 110 and 280 ppm for Class 2 and 3 solvents) were analyzed in triplicate. Both repeatability (RSDs) and linearity (R²) were comparable with the data obtained without active pharmaceutical ingredient. In general, the responses for the spiked samples were between 80 and 105 percent of the response obtained without active pharmaceutical compound (API). This recovery is well within the limits generally accepted for trace impurity analysis. Table 3. Validation Results for 56 Solutes RSD r 2 Class 1: _ 3 ppm Class 1: ppm ICH RT Co- LOD (ppm, rel. 100 mg API) Class 2 and 3: _ 100 ppm Class 2 and 3: ppm Compound class (min) elution 1 ICH SCAN SIM FID SCAN SIM FID SCAN SIM FID Methanol No n-pentane No Ethanol No Ethyl ether No ,1-dichloroethene No Acetone No propanol C Ethyl formate C Acetonitrile No Methyl acetate No Dichloromethane No Z-1,2-dichloroethene C t-butyl methyl ether C n-hexane No propanol No Nitromethane No E -1,2-dichloroethene Partial butanone Partial Ethyl acetate No butanol No

8 Table 3. Validation Results for 56 Solutes (continued) RSD r 2 Class 1: _ 3 ppm Class 1: ppm ICH RT Co- LOD (ppm, rel. 100 mg API) Class 2 and 3: _ 100 ppm Class 2 and 3: ppm Compound class (min) elution 1 ICH SCAN SIM FID SCAN SIM FID SCAN SIM FID THF No chloroform Partial ,1,1-trichloroethane C Cyclohexane C Tetrachloromethane No ² ³ Isobutyl alcohol No Benzene C methoxyethanol C ² 9.5 ² 16 ² 1 point 0.988³ 0.868³ 1,2-dimethoxyethane C < ,2-dichloroethane C ² ² 0.999³ ³ Isopropyl acetate No n-heptane No butanol C Trichloroethene C Methylcyclohexane No ,4-dioxane No Propyl acetate No ethoxyethanol No ² ² 0.977³ ³ 4-methyl-2-pentanone No Isoamyl alcohol No Pyridine No Toluene No Isobutyl acetate No pentanol No hexanone No n-butyl acetate No DMF No ² 10.9 ² 17.8 ² 0.970³ 0.977³ 0.937³ Chlorobenzene No Xylene No Xylene No Xylene No DMAC C ² 17.3 ² 19.5 ² 0.914³ 0.921³ 0.935³ Cumene C < Anisole No < methyl-2-pyrrolidone No ND² 12.1 ² ND² 1 point 0.972³ 1 point Tetralin No ND = Not detected 1 C = Coelution, but resolved by MS; Partial = partial overlap 2 Calculated at 2,500 ppm 3 Calculated in the 500 to 5,000 ppm range 8

9 Application The method was applied to a number of available drug substances. In samples of penicillin V and levamisol (tetramisole), traces of residual solvent were detected. As an illustration, the FID chromatograms for the two samples (for each analysis, 100 mg sample was dissolved in 2 ml DMSO/water, 1:1) are shown in Figures 5 and 6. In the penicillin sample (Figure 5), a trace amount of n-butyl acetate was detected. The concentration, measured by external calibration, was 52 ppm well below the ICH limit (5,000 ppm). In the levamisol sample (Figure 6), a trace amount of toluene was detected. The concentration was 66 ppm, well below the 890 ppm ICH limit. It is interesting to note that a trace amount of dimethyl sulfide was detected in both chromatograms. DMS is an impurity in the DMSO solvent used to dissolve the samples. In addition, an "unknown" was detected in levamisol. Through MS spectral library searching, the peak was identified as 2-chloropropane. This solvent is not included in the ICH solvent list. However, the ability to unequivocally identify this unknown in the sample clearly demonstrates the advantage of using parallel MS detection. Figure 4a. Comparison of overlapped peaks in FID and individual SIM ion chromatograms for 1,1,1-trichloroethane (m/e = 97) and cyclohexane (m/e = 56). Solvent Backflush Capillary column backflushing is starting to be routinely implemented as a means of improving analysis cycle times and data quality. The simplicity and benefits of implementing backflush for residual solvents analysis were evaluated. To reduce the analysis time, late-eluting solvents (DMSO and DMAC) can be backflushed by increasing the pressure at the column outlet (AUX pressure controlling the purged splitter) and lowering the inlet pressure. This is demonstrated in Figures 7A through 7C. In a normal run of a solvent mixture, DMSO elutes at 20.8 minutes, and the standard run continues to 200 C to ensure that there is no carry-over of sample components into the next run. If no analytes of interest elute after DMSO, backflushing can be used to reduce cycle times. Figure 7B shows the same analysis as in Figure 7A, only with a backflush initiated at 20 minutes (just before elution of DMSO). The outlet pressure was increased from 60 to 200 kpa and the GC was held at 150 C for 10 minutes. No peaks are observed after the backflush was initiated. Next, a blank run (without backflush, original method) was performed (Figure 7C). It clearly shows that DMSO solvent was totally removed. The ability to implement backflush was very simple because of the purged-split configuration of FID and MSD. Thermal stress on the column was certainly reduced and the cooldown time from 150 C instead of the 200 C original ending temperature was faster. Further improvement in cycle time can be achieved by reducing backflush time to the minimum time required to fully backflush DMSO. Figure 4b. Comparison of overlapped peaks in FID and individual SIM ion chromatograms for benzene (m/e = 78), dimethoxyethane (m/e = 45), and 1,2-dichloroethane (m/e = 62). Conclusions More than 50 residual solvents can be determined in pharmaceutical products in a single run using a static headspace GC-FID/MS configuration. Quantification can be performed routinely by FID for most target compounds, while MS is especially suited for trace level determination of Class 1 solvents and for identification of unknowns. Mass spectrometry also excels for determination of coeluting peaks through use of extracted ion or SIM ion chromatograms, thereby eliminating the need for additional analyses on dissimilar columns. The quantitative data, including repeatability, linearity, and LOD are excellent, meeting or exceeding ICH guidelines. 9

10 Response dimethyl sulfide n-butyl acetate Figure 5. Analysis of commercial penicillin sample. n-butyl acetate was determined to be present at 52 ppm well below the ICH limit of 5,000 ppm toluene Response chloropropane dimethyl sulfide Figure 6. Analysis of levamisol sample. Toluene was determined to be present at 66 ppm well below the ICH limit of 890 ppm. 10

11 Response Figure 7A. Standards analysis without backflush Response Figure 7B. Standards analysis with backflush at 20 minutes (just before DMSO elution). 11

12 Response 1 References Figure 7C. Blank run after the backflush run demonstrates that the backflush completely removed DMSO. 1. ICH Harmonised Tripartite Guideline, Q3C(R3), 2. R. L. Firor, The Determination of Residual Solvents in Pharmaceuticals Using the Agilent G1888 Network Headspace Sampler, Agilent Technologies publication EN, A. E. Gudat, R. L. Firor, and U. Bober, Better Precision, Sensitivity, and Higher Sample Throughput for the Analysis of Residual Solvents in Pharmaceuticals Using the Agilent 7890A GC System with G1888 Headspace Sampler in Drug Quality Control, Agilent Technologies publication EN, R. L. Firor and A. E. Gudat, The Determination of Residual Solvents in Pharmaceuticals Using the Agilent G1888 HS/6890GC/5975 inert MSD System, Agilent Technologies publication EN, A. E. Gudat and R. L. Firor, The Determination of Extractables and Leachables in Pharmaceutical Packaging Materials Using Headspace/GC/MS, Agilent Technologies publication EN, For More Information For more information on our products and services, visit our Web site at Agilent shall not be liable for errors contained herein or for incidental or consequential damages in connection with the furnishing, performance, or use of this material. Information, descriptions, and specifications in this publication are subject to change without notice. Agilent Technologies, Inc., 2008 Printed in the USA September 18, EN