Overcoming Matrix Effects: GC Method Development for the Determination of Triethylamine and Dimethyl Sulfoxide in a Drug Substance

Journal of Chromatographic Science 2014;52:36 41 doi:10.1093/chromsci/bms201 Advance Access publication December 21, 2012 Article Overcoming Matrix Effects: GC Method Development for the Determination of Triethylamine and Dimethyl Sulfoxide in a Drug Substance Jingzhi Tian*, Guoru Chen and Zhangfei He Merial Limited, Analytical R&D, 631 US Route 1 South, North Brunswick, NJ 08902 *Author to whom correspondence should be addressed. Email: Jingzhi.tian@merial.com Received 22 May 2012; revised 13 November 2012 A direct injection gas chromatography method was developed for the determination of triethylamine and dimethyl sulfoxide (DMSO) in a drug substance (ML-189). Matrix effects were found to result in the overestimation of DMSO when methanol was used as diluent. Multiple approaches to eliminate matrix effects were unsuccessful; these included changes in sample size, split ratio, injector temperature and injector liner (e.g., deactivated liner). Ultimately, matrix effects were eliminated after the diluent was changed from methanol to acetone. A mechanism was proposed and discussed. Introduction Triethylamine (TEA) and dimethyl sulfoxide (DMSO) are used in the manufacturing process of the active pharmaceutical ingredient ML-189. As per veterinary international conference on harmonization (VICH) guidelines (1), these residual solvents should be controlled to ensure the safety and quality of the finished product. An accurate and sensitive analytical method must be developed to determine TEA and DMSO in the drug substance. Static headspace gas chromatography (GC) (2), due to its low sample matrix interference and high sensitivity, has been the preferred methodology for detecting volatile compounds in drug substances and drug products (3 8). However, carryover generally occurs for TEA in headspace GC (9). DMSO, due to its low vapor pressure and high solubility for organic compounds, is usually used as the diluent in headspace GC. It is known to co-elute with other solvents commonly used in headspace GC, such as dimethylformamide (DMF), N, N-dimthylacetamide and benzyl alcohol (7, 10). Therefore, direct injection GC is the preferred method to determine TEA and DMSO. Matrix effects are common phenomena in environmental and food analysis when detecting pesticide residues in fat, oil, food and milk. For the first time, Erney et al. (11) studied and reported in detail the matrix-induced overestimation of organophosphorus pesticides and proposed the mechanism of matrix effects. It was stated that the sample matrix protects the analytes from thermal decomposition or blocks them from adsorption to the active sites of the GC system, primarily inside the injector liner. Thermolabile compounds or polar compounds that are capable of hydrogen bonding tend to have matrixinduced chromatographic enhancement. Sousa Freitas and other researchers (12 14) also reported matrix effects during pesticide residue analysis; Hajsˇlova and Zrostlíkova (15) summarized matrix effects in the analysis of pesticide residues in food and biotic matrices. In their review, the natures of various types of matrix effects were discussed in detail, together with the prevention, reduction and compensation of their occurrence in complex sample matrices. However, reports on drug matrix effects in the determination of residual solvents are rare. Kersten (16) reported a study on matrix effects in the GC determination of residual solvents in drug substances. In his study, 22 residual solvents were studied in three different drug substances (acidic, basic and neutral), and it was claimed that no matrix effects were observed. A preliminary direct injection GC method was developed using an Agilent DB-624 column (stationary phase: 6% cyanopropylphenyl 94% dimethylpolysiloxane; 30 m 0.53 mm i.d., 3.0 mm film thickness) with methanol as diluent: 1 ml of the solution was injected into a GC system with a split ratio of 1:5 and detected by a flame ionization detector (FID). During evaluation of the method, matrix effects were observed to result in the overestimation of DMSO in spiked sample solutions (approximately 110% of the theoretical value). In this paper, multiple approaches were employed to overcome the matrix effects, and it was found that changing the diluent can eliminate matrix effects. Thus far, this is the first report on overcoming GC matrix effects in the determination of volatile compounds in drug substances. Experimental Chemicals and reagents Methanol and acetone were purchased from Burdick & Jackson. TEA and DMSO were purchased from Sigma-Aldrich. All solvents were.99.5% purity. The ML-189 used in this study was provided by Merial Limited. Instrumentation and data acquisition The GC system was an Agilent GC 7890 equipped with an FID detector and a direct injection autosampler. Data acquisition and processing were accomplished using TotalChrom software (version 6.2.1). Table I shows the chromatographic conditions. Preparation of standard and sample solutions Working standard solutions were prepared by dissolving appropriate amounts of TEA and DMSO in diluent (methanol or acetone) to reach 100% of concentrations at their specification limits. The specification limit is 5,000 ppm in ML-189 for both TEA and DMSO. Concentrations of both TEA and DMSO in # The Author [2012]. Published by Oxford University Press. All rights reserved. For Permissions, please email: journals.permissions@oup.com

Table I Chromatographic Parameters Column Agilent J&W DB-624, 30 m 0.53 mm i.d., 3.0 mm film thickness Injector temperature 2508C Injector mode Split Split ratio 1:5 Liner Non-deactivated straight tube with glass wool (Agilent part 19251-60540) Carrier gas Helium Mode Constant flow Flow rate 3.5 ml/min Oven temperature 708C, held for 8 min 508C/min to 1308C, held for 5 min 508C/min to 2408C, held for 5 min Run time Approximately 21 min Detector FID Detector temperature 2508C Make-up gas Helium Detector mode Constant make-up Make-up gas flow 30 ml/min Injection volume 1 ml working standard solutions were 0.25, 0.125 and 0.075 mg/ml, respectively, for 500, 250 and 150 mg sample sizes. Spiked sample solutions were prepared by dissolving approximately 500, 250 or 150 mg of ML-189 in 10 ml of corresponding working standard solutions. Sample solutions were prepared by dissolving approximately 500, 250 or 150 mg of ML-189 in 10 ml of diluent. GC analysis For GC analysis, a portion of each solution was transferred into a crimp-top vial with a fixed insert (Agilent part number 9301-1388), and the vial was closed with a PTFE/rubber crimp cap (Agilent part number 5061-3370). One microliter of each solution was injected into the GC system. Data were acquired and processed by TotalChrom software (version 6.2.1). Evaluation Six spiked sample solutions were injected between two working standard solutions to evaluate recoveries. Sample solutions were injected to determine the concentrations of TEA and DMSO present in the samples. Recoveries of the analytes were calculated using the following formula: Ass Aspl Wss Wspl Cstd Vss Recovery ¼ Astd 100% Cstd Vstd where Ass ¼ peak area of analyte in spiked sample solution; Aspl ¼ peak area of analyte in sample solution; Wss ¼ weight of sample in spiked sample solution; Wspl ¼ weight of sample used in sample solution; Astd ¼ average peak area of analyte in bracket working standard solutions; Cstd ¼ concentration of analyte in working standard solution; Vss ¼ volume of spiked sample solution; Vstd ¼ volume of working standard solution used in the spiked sample solution. For each condition, six spiked sample solutions were prepared and analyzed. The relative standard deviations (RSDs) of the recoveries from the six preparations were calculated to demonstrate the precision of the results. Results and Discussion Overcoming matrix effects Chromatographic conditions are described in Table I. Initially, methanol was used as the diluent and the working standard solution was prepared based on a sample size of 250 mg. There were no interfering peaks at the retention times of TEA and DMSO in the diluent and no carryover for both TEA and DMSO in the diluent injection after working standard solution injections. Figure 1 shows representative chromatograms using methanol as diluent. Recoveries of TEA were close to 100% of theoretical value. However, 110% recoveries were found for DMSO (Table II: 250 mg sample). This phenomenon indicated the presence of matrix effects. ML-189 is a polar compound; it is an oxime derivative that contains carbonyl and hydroxyl groups. The melting point of ML-189 is more than 3008C. According to the mechanism proposed by Erney et al. (11), the recovery of DMSO can be enhanced by two pathways: (i) DMSO decomposes in working standard solutions at high temperature; in the spiked sample solutions, the presence of ML-189 prevents DMSO from thermal decomposition, which results in recoveries of.100%; (ii) ML-189 blocks the active sites within the GC inlet and results in more free DMSO molecules in the spiked sample solution than in the working standard solution. The active sites may be the free silanol groups inside the liner, trace metals possibly present in the GC injector, or even the residue from the sample matrix. To overcome matrix effects, the most commonly used approach is to prepare sample and standard solutions using the same matrix, which, in this case, means preparing standard solutions with the presence of the same amount of drug substance as free TEA and DMSO. However, this procedure is inconvenient and uncommon in pharmaceutical analyses. Another approach is to eliminate active sites in the liner or decrease sample concentration. It is only theoretically viable to make the liner free of active sites, because of the unavailability of commercial inert materials that can sustain long exposures to high temperatures. Two types of liners, non-deactivated straight tubes with glass wool and deactivated low pressure liners, were evaluated for liner effects. To investigate the effect of sample size, working standard solutions were prepared based on different sample sizes, and different split ratios were also studied. To investigate whether DMSO decomposition was the reason, recoveries of DMSO and TEA under different injector temperatures were studied. The results of these investigations are summarized in Table II. During the study, one parameter was changed under each condition while the other parameters were kept constant, as listed in Table I. Six replicate determinations were made under each condition. For DMSO, the mean recovery under each condition ranged from 109 to 116%. No significant differences were observed among the recoveries for different sample sizes (500, 250 and 150 mg). Considering the sensitivity of the method, smaller sample sizes were not investigated. Increasing Overcoming Matrix Effects 37

Figure 1. Overlain chromatograms of methanol, sample solution, spiked sample solution and working standard solution. Table II Recoveries of DMSO and TEA under Different Conditions (Methanol as Diluent)* DMSO Preparation 500 mg sample 250 mg sample 150 mg sample split ratio 1:10 Change liner Injector temperature (2208C ) 1 114 114 116 115 111 109 117 2 113 112 114 115 111 110 116 3 114 110 113 114 110 108 115 4 112 110 114 114 112 109 115 5 113 110 112 114 110 111 115 6 113 111 114 114 110 109 116 Mean 113 111 114 114 111 109 116 RSD (%) 0.7 1.5 1.2 0.4 0.8 1.0 0.7 TEA 1 101 99 101 99 102 100 100 2 103 100 101 102 102 102 101 3 102 99 101 101 102 101 99 4 101 99 102 101 103 103 100 5 102 99 100 102 102 103 100 6 102 100 101 101 102 103 99 Mean 102 99 101 101 102 102 100 RSD (%) 0.7 0.5 0.5 1.1 0.2 1.2 0.7 *Note: Method conditions are the same as those listed in Table I. Deactivated low pressure liner, Agilent part 5183-4647. Sample size of 250 mg. Injector temperature (2808C ) the split ratio from 1:5 to 1:10 did not reduce the recoveries of DMSO to normal range, which is as predicted because the injected amount into the GC inlet was the same. A change of split ratio only affected the signal intensity of the analytes, and the amount of injected sample matrix remained the same. The matrix effects did not get weaker when a deactivated liner was used. The mean recovery was 111%, which was the same as the mean recovery using the initial non-deactivated liner. This result showed that the cause of the matrix effects could not be reduced by using a deactivated liner, and further proved that even deactivated liners are not free of active sites (15). For TEA, the mean recovery in each condition varied from 99 102% and was within normal range (80 103%) (11, 17). At three different injector temperatures, the recoveries of DMSO were reduced from 116 to 109% as the injector temperature was changed from 280 to 2208C. However, even at the injector temperature of 2208C, recoveries of DMSO were still approximately 110%. These results indicated that even DMSO decomposes at high temperatures; thermal decomposition is not the primary reason for the overestimation of DMSO. The boiling point of DMSO is 1898C. If the injector temperature is too low, DMSO in the sample matrix may not vaporize completely before it is transferred into the column. Therefore, lower injector temperatures were not investigated. Of all of the preceding approaches, only decreasing the injector temperature reduced matrix effects to some extent, and none of the approaches could completely eliminate matrix effects and reduce the recoveries of DMSO to normal range. According to the literature (11, 15), polar compounds that have carbonyl, organophosphate or amino groups and are capable of hydrogen bonding tend to have matrix effects in GC analysis. They are easily adsorbed inside the GC inlet, which 38 Tian et al.

results in matrix effects. The matrix effects in this study could be attributable to the high polarity and hydrogen bonding capability of DMSO; therefore, DMSO is easily adsorbed inside the GC inlet. When the spiked sample solution was injected, the active sites of the GC inlet were blocked by the drug substance and resulted in more free DMSO molecules than the working standard solution, and thus a recovery of higher than 100%. TEA had no matrix effects under each condition in this study. In this study, the possible reason is that TEA has much less hydrogen bond forming ability and is less polar than DMSO; therefore, TEA is not as easily adsorbed to the active sites inside the GC inlet as DMSO. Another approach evaluated in this study was to change the sample diluent. If a solvent has adsorption ability that is similar to or stronger than DMSO on the active sites of the GC inlet, the vast majority of the diluents molecules may block the active sites, and DMSO molecules will not be significantly adsorbed. Acetone has a similar structure to DMSO; the sample has good solubility in acetone, and the retention time of acetone is not close to any of the analytes when using the same chromatographic conditions as described in Table I. Therefore, acetone is a good candidate to replace methanol. Figure 2 shows representative chromatograms using acetone as diluent. Solutions were prepared based on 250 and 150 mg sample sizes, and recoveries of TEA and DMSO are summarized in Table III. According to the results listed in Table III, recoveries of TEAand DMSO at both sample concentrations ranged from 94 102% and were within an acceptable range. Matrix effects were eliminated by using acetone as the diluent. A possible mechanism of this matrix effect is attributable to the similarity in structure between acetone and DMSO; the active sites of the GC inlet were saturated with acetone molecules because of its excess compared to DMSO. The adsorption of DMSO toward the active sites was minimized in both working standard solutions and spiked sample solutions, which provided better recovery and accuracy of DMSO. Method validation Recoveries of TEA and DMSO were slightly better when the 150 mg sample was used. However, considering the sensitivity of the method, a sample size of 250 mg was selected in the final method. The final method was validated per VICH requirements (18) and was proven to be accurate, precise, specific, linear, robust and sensitive. Specificity The specificity of the method was evaluated by injecting one blank (acetone) solution, one working standard solution and one sample spiked with working standard solution. No interference peaks were observed at the retention times of TEA and DMSO in the blank solution injection. In the spiked sample solution, the resolution between DMSO and its adjacent peak was 6.2 and no adjacent peak was detected for TEA. Figure 2. Overlain chromatograms of acetone, sample solution, spiked sample solution and working standard solution. Overcoming Matrix Effects 39

Table III Recoveries of TEA and DMSO at Different Sample Sizes (Acetone as Diluent) DMSO TEA Levels 250 mg sample 150 mg sample 250 mg sample 150 mg sample 1 95 98 98 102 2 96 96 98 101 3 95 95 97 101 4 94 95 97 99 5 94 97 96 102 6 95 96 97 101 Mean 95 96 97 101 RSD (%) 0.6 1.5 0.8 1.0 Linearity The linearity of the method was assessed by using standard solutions at 50, 80, 100, 150 and 200% levels of the specified concentrations for both TEA and DMSO. Linear regression was performed between concentration and peak area. The correlation coefficient (R) was 1.00 for both TEA and DMSO. Accuracy The accuracy of the method was assessed by spiking TEA and DMSO into ML-189 drug substance so that their concentrations reached levels of the specified concentrations of 50% (three preparations), 100% (six preparations) and 200% (three preparations). Due to the possible presence of TEA and DMSO in the drug substance used in the validation, the amounts of TEA and DMSO in ML-189 were determined in triplicate. The averages were used as the concentrations of TEA and DMSO in the drug substance. The accuracy of the method was demonstrated with the recovery. At 50, 100 and 200% of working standard concentration levels, TEA recoveries ranged from 97 to 98% and DMSO recoveries ranged from 93 to 96%. Precision The precision was evaluated with repeatability and intermediate precision. The repeatability was assessed by evaluating the recoveries and RSD of recoveries for TEA and DMSO at a level of 100%. Six preparations at 100% level of the specification concentrations in accuracy were used to demonstrate repeatability. For the intermediate precision study, a second analyst performed a repeatability study independently on a different day by using a different instrument and a different GC column. The mean value and RSD of the recoveries were reported. The mean recoveries for TEA and DMSO from the second analyst were 99 and 98%, respectively. Sensitivity and robustness The quantitation limit (QL) of the method was determined by injecting diluted standard solution; the signal-to-noise (S/N) ratio of the QL solution should be at least 10. The QL of the method was 51 ppm for TEA and 49 ppm for DMSO. A robustness study was conducted by deliberately changing some of the chromatographic parameters, such as initial oven temperature and flow rate. One standard solution was analyzed under each condition and compared to the result under normal method conditions. No critical parameters were identified and the method was proven to be robust. Conclusion A direct injection GC method was developed for the determination of TEA and DMSO in the drug substance ML-189. During development, the recoveries of DMSO were higher than normal range because of matrix effects when methanol was used as the diluent. The matrix effects were eliminated by changing the diluent from methanol to acetone, and the mechanism was discussed. The final method was validated per VICH guidelines and the method is able to accurately determinate TEA and DMSO in the drug substance ML-189. This study also demonstrated that drug matrix effects do exist, although they are not as severe as in environmental and food analyses for pesticide residues. Acknowledgments Authors would like to thank Drs Daniel Cohen and Ian Chung for their support in this study. References 1. International Conference on Harmonization; Harmonized tripartite guideline on impurities: Residual solvents (VICH GL 18). International Cooperation on Harmonization of Technical Requirements for Registration of Veterinary Medicinal Products (VICH), Bruxelles, (2000). 2. Hachenberg, H., Schmidt, A.P.; Gas chromatographic headspace analysis. Heyden Press, Rheine, Germany, (1977). 3. The United States Pharmacopoeia. Residual solvents,467., USP35/NF30, The United States Pharmacopeial Convention, Rockville, MD, (2012). 4. Identification and Control of Residual Solvents (2.4.24). European Pharmacopoeia, 7th edition. teuropean Directorate for the Quality of Medicines & HealthCare, Strasbourg, France, (2010). 5. Witschi, C., Doelker, E.; Residual solvents in pharmaceutical products: Acceptable limits, influences on physicochemical properties, analytical methods and documented values; European Journal of Pharmaceutics and Biopharmaceutics, (1997); 43: 215 242. 6. B Hymer, C.; Residual solvent testing: A review of gaschromatographic and alternative techniques; Pharmaceutical Research, (2003); 20: 337 344. 7. Somuramasami, J., Wei, Y.-C., Soliman, E. F., Rustum, A. M.; Static headspace gas chromatographic method for the determination of low and high boiling residual solvents in Betamethasone valerate; Journal of Pharmaceutical and Biomedical Analysis, (2011); 54: 242 247. 8. Cheng, C., Liu, S., Mueller, B.J., Yan, Z.; A generic static headspace gas chromatography method for determination of residual solvents in drug substance; Journal of Chromatography A, (2010); 1217: 6413 6421. 9. Wald, G.V., Albers, D., Nicholson, L., Langhorst, M., Bell, B.; Development of a headspace gas chromatography method to determine residual aliphatic amines in oligonucleotides; Journal of Chromatography A, (2005); 1076: 179 182. 10. D Autry, W., Zheng, C., Bugalama, J., Wolfs, K., Hoogmartens, J., Adams, E.; Liquid paraffin as new dilution medium for the analysis of high boiling point residual solvents with static headspace-gas chromatography; Journal of Pharmaceutical and Biomedical Analysis, (2011); 55: 1017 1023. 11. Erney, D.R., Gillespie, A.M., Gilvydis, D.M., Poole, C.F.; Explanation of the matrix-induced chromatographic response enhancement of organophosphorus pesticides during open tubular column gas chromatography with splitless or hot on-column injection and flame photometric detection; Journal of Chromatography A, (1993); 638: 57 63. 40 Tian et al.

12. Sousa Freitas, S., Lanc as, F.M.; Matrix effects observed during pesticides residue analysis in fruits by GC; Journal of Separation Science, (2009); 32: 3698 3705. 13. Chu, X.-G., Hu, X.-Z., Yao, H.-H.; Determination of 266 pesticides in apple juice by matrix solid-phase dispersion and gas chromatography mass selective detection; Journal of Chromatography A, (2005); 1063: 201 210. 14. Mastovska, K., Lehotay, S.J., Anastassiades, M.; Combination of analyte protectants to overcome matrix effects in routine GC analysis of pesticide residues in food matrixes; Analytical Chemistry, (2005); 77: 8129 8137. 15. Hajšlova, J., Zrostlíkova, J.; Marix effects in (ultra)trace analysis of pesticide residues in food and biotic matrices; Journal of Chromatography A, (2003); 1000: 181 197. 16. Kersten, B.S.; Drug matrix effect on the detection of residual solvents in bulk pharmaceuticals by wide-bore capillary gas chromatography; Journal of Chromatographic Science, (1992); 30: 115 119. 17. Gillespie, A.M., Walters, S. M.; Rapid clean-up of fat extracts for organophosphorus pesticide residue determination using C18 solid-phase extraction cartridges; Analytica Chimica Acta, (1991); 245: 259 265. 18. International Conference on Harmonization. Harmonized tripartite guideline on validation of analytical procedures: Methodology (VICH GL 12). International Cooperation on Harmonisation of Technical Requirements for Registration of Veterinary Medicinal Products (VICH), Bruxelles, (1998). Overcoming Matrix Effects 41