Anal. Chem. 2009, 81, 4576–4580
Total Residue Analysis of Swabs by Ion Mobility Spectrometry Mark A. Strege* Lilly Research Laboratories, Eli Lilly and Company, Lilly Corporate Center, Indianapolis, Indiana 46285 Ion mobility spectrometry (IMS) is a technique attractive for use within the pharmaceutical industry for at-line determination of residues on swabs taken from the surfaces of manufacturing equipment for the purposes of cleaning validation or verification. In this study, the development of a novel IMS method to provide a measurement of total residue present on a swab is described. The technique is based upon quantitation of charged atmospheric gas reactant ion consumption (RIC) within the instrument as a direct measure of the mass of total ionizable residue. Coupled with the conventional analysis of the active pharmaceutical ingredient within a single 2 min analysis, RIC determination provided the benefit of a single measure representative of the presence of multiple residue components or unknown components. To account for differences in response between components of a model drug product (Cymbalta) and its associated cleaning agents, a strategy was proposed to determine a “worst case” total residue test result based on RIC. A limitation of the IMS method was its incompatibility with cleaners containing a high concentration of inorganic components. The methodology provided a range from 5-50 µg per 25 cm2 surface area and acceptable analyte recovery (50-100%). To decrease pharmaceutical manufacturing equipment downtime as a result of cleaning validation or verification, the use of ion mobility spectrometry (IMS) has been proposed as a substitute for traditional analytical techniques for the analysis of liquid extracts of swabs taken from equipment surfaces.1-4 However, as discussed in a previous report,5 the greatest advantage to be gained from the use of a rapid analysis technique such as IMS is its use as an at-line testing tool for the direct analysis of swabs because integration with a traditional analytical laboratory and the associated processes can be bypassed entirely. To enable the use of IMS for quantitative direct swab analysis over the range of 5-100 µg per 25 cm2 surface area (the high end of this range * E-mail:
[email protected]. Phone: (317) 276-9116. Fax: (317) 2775519. (1) Payne, K.; Fawber, W.; Faria, J.; Buaron, J.; Debono, R.; Mahmood, A. Spectroscopy 2005, 20 (Suppl.), 24–27. (2) Walia, G.; Davis, M.; Steanou, S.; Debono, R. Pharm. Technol. 2002, April, 72–78. (3) Walia, G.; Davis, M.; Stefanou, S. Pharma. Process. 2003, September, 20– 22. (4) O’Donnell, R.; Sun, X.; Harrington, P. TrAC, Trends Anal. Chem. 2008, 27 (1), 44–53. (5) Strege, M.; Kozerski, J.; Juarbe, N.; Mahoney, P. Anal. Chem. 2008, 80, 3040–3044.
4576
Analytical Chemistry, Vol. 81, No. 11, June 1, 2009
corresponding to the established “visibly clean” limit 6), modifications were made to a commercially available ion mobility spectrometer.5 Although the dynamic response range was successfully achieved in the earlier study for individual components, levels of signal suppression up to 50% were observed within analyte mixtures at high levels of total residue (>30 µg), leading to a concern that a significant amount of target analyte could be masked by ionizable matrix components. For example, under these conditions matrix effects could inhibit the determination of active pharmaceutical ingredient (API) in the presence of residual drug product excipients or cleaning agent components, resulting in a false negative IMS test result. To address this concern, an alternative approach for the use of direct swab analysis by IMS is proposed. Within the IMS instrument ionization source consisting of a chamber lined with 63 Ni foil and filled with atmospheric gas, charged gaseous “reactant ions” are formed by the high energy primary electrons emitted in the source.7 Starting with the initial collision of an electron with gaseous nitrogen, a series of charge transfer reactions has been proposed that results in the formation of stable, predominant charged species based on water or carbon dioxide. The reactant ions are visible as discrete peaks within the IMS plasmagram. In the presence of analyte, charge is transferred from the reactant ions to the analyte molecules, the peaks corresponding to the charged analyte species appear in the plasmagram, and the reactant ion signal is decreased proportionally to the extent of charge transfer. Therefore, in theory the presence of any ionizable species on a swab may be determined through measurement of the decrease in the reactant ion signal intensity relative to the blank signal background, and the suppression of an individual target component in the presence of a mixture is no longer an issue because all ionizable species in the mixture will contribute to the consumption of the reactant ion population. The determination of total ionizable residue may serve as a non-specific test, inclusive of all ionizable components in a mixture, as well as unknown ionizable contaminants, in a manner similar to traditional methodologies such as total organic carbon analysis (TOC).8,9 A significant benefit of using IMS for non-specific assessment of total residue is that this approach may be combined within one analysis with determination (6) Fourman, G.; Mullen, M. Pharm. Technol. 1993, April, 54–60. (7) Eiceman, G.; Karpas, Z. Ion Mobility Spectrometry, 2nd ed.; CRC Press: Boca Raton, FL, 2005; Chapter 3. (8) Strege, M.; Stinger, T.; Farrell, B.; Lagu, A. BioPharm 1996, 9 (4), 42–45. (9) Guazzaroni, M.; Yiin, B.; Yu, J. Am. Biotech. Lab. 1998, 16 (10), 66–67. 10.1021/ac900441k CCC: $40.75 2009 American Chemical Society Published on Web 05/06/2009
Table 1. Constitution of Sample Mixtures
sample mixture
solids [w/w %]
duloxetine [w/w %]
TOC [w/w %]
non-volatilea inorganics [w/w %]
non-volatile inorganics ) as a % of solid ) residue [w/w %]
Cymbalta drug product CIP-100 aqueous concentrate CIP-200 aqueous concentrate
100 30 50
32
85 3 2
15 25
15 83
a
At the desorber temperature of 249 °C.
of individual target components, specifically the API, which traditionally has been obtained through analyte-specific methodology separate from TOC. In this study, API-specific quantitation and reactant ion consumption (RIC) measurements have been evaluated for the determination of residues corresponding to components present within a typical pharmaceutical manufacturing suite. The API duloxetine, the formulated drug product Cymbalta, and the cleaning agents associated with the drug product manufacture have been employed as a model. EXPERIMENTAL SECTION Instrumentation. A Kaye Validator Ion Trap Mobility Spectrometer was purchased from GE Sensing (Billerica, MA). Instrument desorber and detector temperatures were 249 and 205 °C, respectively, and a sample acquisition time of 120 s (5 scans acquired per second, alternating positive and negative ion mode) was utilized. Negative ion mode separation profile “plasmagrams” were created by the instrument software through summation and signal averaging across the total acquisition time. The drift gas was atmospheric air-dried through the use of a regenerative dryer assembly filled with a molecular sieve. The removal of moisture ensures consistency in the drift gas for optimal ion mobility reproducibility. Default gas flow settings were 500 cc/min in the sample flow chamber and 250 cc/min on the detector flow. Analyte migration time within the instrument was calibrated through the use of a calibration swab (purchased from GE Sensing) loaded with a mixture of ibuprofen, acetaminophen, and aspirin. A calibration check was performed daily prior to all sample analyses to ensure migration time consistency. Materials. Samples of formulated Cymbalta drug product were provided by Eli Lilly and Company (Indianapolis, IN). The drug product formulation was composed of the active pharmaceutical ingredient (API) duloxetine hydrochloride in the presence of filler, binder, lubricant, plasticizer, an enteric coating polymer, and various inorganic excipients. The cleaning concentrates CIP-100 and CIP-200 and several individual components of the cleaners (citric acid, tetrasodium ethylenediaminetetraacetic acid (EDTA), and a proprietary amphoteric surfactant) were provided by Steris, Inc. (St. Louis, MO). The major inorganic components of CIP100 (a caustic cleaner) and CIP-200 (an acid cleaner) are potassium hydroxide and phosphoric acid, respectively.10,11 The primary organic components are EDTA in CIP-100 and citric acid in CIP-200, each present in the mixtures at approximately 5% w/w. The remaining organic constituents in both cleaners consist of low levels (99% of swab test results have been 25 µg. The total IMS duloxetine response was determined via the summed response of the monomer and dimer peaks. The analytical performance parameters determined for precision and recovery from sets of six replicates at each level are displayed in Table 2. Surface recovery was determined through direct comparison of duloxetine response from swabbed surfaces versus swabs directly spiked with the API. For determination of the recovery of duloxetine from drug product, the levels in the table corresponded to the total mass of the drug product mixture spiked directly onto a swab. The precision and surface recovery performance of the IMS quantitation was comparable to that achieved through the use of the historically established methodology of liquid extraction of polyester swabs followed by quantitation by (12) Eiceman, G.; Salazar, M.; Rodriguez, M.; Limero, T.; Beck, S.; Cross, J.; Young, R.; James, J. Anal. Chem. 1993, 65, 1696–1702.
4578
Analytical Chemistry, Vol. 81, No. 11, June 1, 2009
Table 2. Analytical Performance Parameters for the Quantitation of Duloxetine API (n ) 6 at each level) total residue mass parameter
5 µg 10 µg 20 µg 25 µg
spiked swab: swab-to-swab precision (% RSD) 4.6 2.9 mean recovery from stainless steel (%) 109 92 precision of recovery from stainless steel (% RSD) 14.6 10.2 recovery of duloxetine from drug producta(%) 100 100
1.1 90 7.7 59
2.1 86 3.7 54
a 1.6, 3.2, 6.4, and 8.0 µg duloxetine were present in the drug product at the total residue mass levels of 5, 10, 20, and 25 µg, respectively.
UV absorbance. For the recovery of duloxetine in the presence of drug product, the suppression of the API response observed at the higher levels of residue mass (reflected by recoveries of 50-60%) is representative of the effects of ionization competition.5 Determination of Total Ionizable Residue. For determination of total ionizable residue via RIC, the integrated peak areas of both reactant ion peaks (migrating at 3.4 and 5.0 ms) were included in the calculation of the total reactant ion signal as a summed value. The reactant ion signal generated from the analysis of blank swabs was determined to be acceptably stable over time through the testing of 42 blank swab replicates performed within a set of 120 samples total, representing approximately 10 h total instrument time. The precision of this data set was 5.6% RSD, and the baseline response was observed to have drifted by approximately 20% from start to finish. To account for the drift in the reactant ion signal response, the analysis of a blank swab after every 9 samples was established within the methodology. The reproducibility of the reactant ion response was also evaluated over the course of 2 months during which approximately 600 samples (including 200 swab blanks interspersed across the sample sets) were analyzed. The reactant ion response from the blank analyses was observed to remain very stable throughout this time, as indicated by a precision of 8.8% RSD and no observed overall trend in drift of response. It is important to note that no
Table 3. Stainless Steel Surface Recovery at Levels of 5, 20, and 50 µg based on RIC Determination (n ) 6 at each level) mean recovery from stainless steel [%]
Figure 2. Plots of RIC vs total residue obtained for (() duloxetine, (b) surfactant, (9) drug product, (2) citric acid, (∆) CIP-200, (0) CIP100, and (O) EDTA obtained across the range of 5-50 µg.
preventative maintenance (such as replacement of the PTFE membrane between the desorber and the ionization chamber) was performed during this time period, indicating the robustness of the instrument for this application. The precision calculated from the analysis of six replicates of 5 µg citric acid enabled the estimation of limits of detection (3 × SD) and quantitation (10 × SD) equal to 0.9% RIC and 3.0% RIC, respectively. Because the conventional TOC method in place for the support of Cymbalta manufacturing employs a range of 5-50 µg, a dynamic range of 5-50 µg was evaluated for indirect quantitation of total residue via IMS measurement of RIC for the drug product and cleaning agent mixtures and several key components of these mixtures (duloxetine, citric acid, EDTA, and surfactant) (see Figure 2). For the CIP-100 cleaner, potassium hydroxide is the primary component by weight at approximately 80% of the solids and is known to be highly water-soluble and easily removed from surfaces subjected to aqueous rinsing. Therefore, there was no interest or requirement to test for residues of potassium hydroxide on equipment surfaces, and the residue mass levels for CIP-100 represented in Figure 2 were prepared at levels corresponding to the amount of volatile organic constituents in the cleaner. The RIC responses were observed to vary significantly, most likely because of differences in volatility and in the abilities of the different chemical species to acquire a negative charge or abstract a proton.7 The curves for the components generating high response (most notably duloxetine and surfactant) were logarithmic because of the ionization capacity threshold effects within the instrument. The analyte mixture CIP200 generated a curve of lower response approaching linearity. It is important to note that phosphoric acid was volatile at the 249 °C temperature provided by the desorption chamber, and its contribution to the RIC response of CIP-200 mixture was confirmed experimentally. However, the quantitation of phosphoric acid was not of interest for this application because it is an inorganic species that is easily removed by aqueous rinsing. Both the CIP-100 mixture and its component EDTA demonstrated little or no response, most likely because tetrasodium EDTA and potassium hydroxide are non-volatile at 249 °C. However, even with the adjustment of the total mass on the swabs described above, the recovery of response for CIP-100 in Figure 2 was much lower than expected. An investigation revealed that the presence of potassium hydroxide in a mixture with surfactant loaded on a swab at w/w ratios of 1:1 to 4:1 resulted in the
sample
5 µg
20 µg
50 µg
Drug Product CIP-200 Citric Acid Surfactant
126 100 76 104
78 103 98 54
78 92 98 64
Table 4. Stainless Steel Surface Recovery Precision at Levels of 5, 20, and 50 µg based on RIC Determination (n ) 6 at each level) precision of recovery from stainless steel [% RSD] sample
5 µg
20 µg
50 µg
Drug Product CIP-200 Citric Acid Surfactant
1.6 2.1 2.1 1.4
7.1 2.8 4.4 4.7
4.2 8.4 6.2 13.5
quenching of the surfactant response to levels of approximately 60 and 30% recovery, respectively, as directly compared to the response of the surfactant alone. It therefore appeared that the presence of the non-volatile inorganic base (and to a lesser degree the EDTA) had resulted in the formation of a physical barrier or “residue coating” that had inhibited the desorption of the volatile analytes. Through the determination of RIC, mean stainless steel surface recoveries were calculated for the drug product, CIP-200, citric acid, and surfactant by quantitation of samples taken from surfaces versus response calibration curves created from corresponding samples loaded directly on swabs. The mean (n ) 6) recovery data, together with the precision associated with the recoveries, are displayed in Tables 3 and 4. The results were comparable to the performance of TOC methodology established for cleaning validation support.13 Strategy for Practical Application of RIC-Based Residue Quantitation. For the practical application of the RIC approach for quantitation of total residue collected by swab sampling of manufacturing equipment, the following strategy is proposed to take into account the significant difference in response observed between the individual species present in the drug product and cleaning agents. The first step following swab sampling and generation of a plasmagram is the identification of the specific components present on the swab via direct match of unknown peak migration time versus the migration time of standards. This “forensics identification” is made possible by the fact that IMS peak migration time is extremely stable (data representing 15 injections of duloxetine obtained over the course of 10 h demonstrated a migration time precision for both monomer and dimer of e1% RSD), and the components of interest (reactant ion peaks at 3.4 and 5.0 ms, duloxetine at 5.4 and 8.0 ms, citric acid at 4.5 ms, and surfactant at 7.1 ms) are adequately resolved with no interference from the sample matrixes. To provide a “worst case” estimate of total residue, a response curve is employed to (13) Holmes, A.; Venderwielen, A. PDA J. Pharm. Sci. Technol. 1997, 51 (4), 149–152.
Analytical Chemistry, Vol. 81, No. 11, June 1, 2009
4579
link RIC to residue mass representing the component of lowest response observed to be present in the sample plasmagram. For example, if a mixture of duloxetine, citric acid, and surfactant was observed to be present in a swab sample, the total residue determined for the analysis would be based on the RIC response curve for citric acid. The presence of IMS peaks of unknown identity based on IMS migration time correlation (i.e., chemical species not originating from drug product or cleaners) would also be processed to provide a “worst case” quantitative estimate based on citric acid response. CONCLUSIONS Unlike TOC, IMS is a separation technique that enables specific determination of the individual components (such as the API) which may be present on the surface of pharmaceutical manufacturing equipment. At ambient pressure and temperature in the Validator instrument where the length of the drift tube and the potential applied across the drift tube are fixed constants, analyte migration time-of-flight (ToF) is inversely proportional to the traditional measure of reduced mobility (K0, corresponding to D2/VS where D ) length of the drift tube, V is the applied potential, and S is the analyte migration time).14 As has been routinely established for security or military applications of IMS, the identification of components appearing in a swab sample plasmagram may be made through direct comparison with the ToF values established for standards of known compounds. Through the use of RIC, the capability of IMS to provide an estimate of total ionizable residue has also been evaluated. The potential advantage of IMS is its ability to provide quantitation of both the API and the total residue within a single 2 min analysis. A significant advantage of utilizing RIC analysis is the fact that the presence of multiple components, together with unknown ionizable contaminants present in a surface residue, will result in a cumulative response. ToF measurements enable the identification of components contributing to the RIC response, and a “worst (14) Eiceman, G.; Karpas, Z. Ion Mobility Spectrometry, 2nd ed.; CRC Press: Boca Raton, FL, 2005; Chapter 1.
4580
Analytical Chemistry, Vol. 81, No. 11, June 1, 2009
case” estimate of total residue may be based upon the least responsive component identified to be present in the residue. However, a disadvantage lies in the fact that although the IMS analysis and data generation processes are rapid, the estimation of total ionizable residue as described above is a somewhat complex procedure (because of differential analyte response) which will benefit from the use of instrument software or spreadsheets specifically designed for the application. A minor limitation of the use of direct swab analysis by IMS is its inability to detect inorganic species (such as potassium hydroxide) or organic species (such as metal-bound EDTA) that are not volatile at the temperature provided by the instrument sample desorber. In general, this issue is not a major concern because inorganic species have not traditionally been determined by conventional techniques (such as TOC, HPLC, UV) for equipment cleaning verification. However, it was also demonstrated in this study that the cleaning agent CIP-100 is not compatible with direct swab IMS because it contains such high levels of non-volatile constituents that the response of the volatile organics within the residue is significantly inhibited. As is the case for conventional IMS analyses, RIC determination is also limited by ionization capacity. The development of an instrument with an enhanced capacity ionization source or a gaseous sample flow splitter to provide analysis of samples at levels >50 µg will not only facilitate a quantitative range useful for cleaning verification up to the “visibly clean” limit of 100 µg, but also would eliminate the diminished recovery of analytes at high levels that occurs because of ionization competition within mixtures. ACKNOWLEDGMENT The author wishes to acknowledge Douglas Classen and David Linson for their technical assistance.
Received for review February 27, 2009. Accepted April 22, 2009. AC900441K
