Enhancing Drug Development by Applying LC–MS–MS for Cleaning Validation in Manufacturing Equipment

February 2, 2006
Pharmaceutical Technology, Pharmaceutical Technology-02-02-2006, Volume 30, Issue 2

LC–MS–MS has the potential to become a routine technique for determining drug residues in support of cleaning verification, especially early in drug development.

Currently, liquid chromatograph–ultraviolet spectrometry (LC–UV) is typically applied to cleaning validations because of its familiarity, robustness, ease of use, and regulatory acceptability. For low-dose compounds, equipment requiring low residue limits, and compounds lacking strong chromophores, the enhanced sensitivity and selectivity of liquid chromatography–mass spectrometry–mass spectrometry (LC–MS–MS) facilitates rapid method development for the detection of low levels of residues of active pharmaceutical ingredients (APIs). LC–MS–MS is an acceptable technique for the analysis of API residues for cleaning validation. More importantly, in applications where sensitivity and selectivity are inadequate using traditional modes of detection, LC–MS–MS offers substantial advantages. LC–MS–MS will afford faster development and analysis time, potentially making it the predominant tool of choice.

To ensure safe pharmaceutical drug products, international guidance for manufacturing (e.g., the International Conference on Harmonization [ICH] or current good manufacturing practices [CGMP]) requires that levels of impurities in drug products be carefully monitored and controlled (1). In addition to impurities explicitly addressed by the ICH guidance, pharmaceutical products may be contaminated by process-related impurities, residues of cleaning agents, lubricants, and airborne matter such as dust and particulates. These products may also be tainted by cross-contamination from previously manufactured products. To minimize this manufacturing process carryover, methods for cleaning manufacturing equipment are put in place. These cleaning methods, which depend on the physical and chemical properties of the API and excipients, typically involve high-pressure spraying of a series of surfactant and solvent solutions. The cleaning methods are validated or verified to ensure cleanliness for each specific API by swabbing predefined areas of the manufacturing equipment that have product contact or collecting equipment rinses. The swabs or rinse solutions are then analyzed for API content and evaluated against an established residue limit. Residue limits are set with consideration of the therapeutic dose of the previously manufactured API, toxicity concerns, equipment surface areas, subsequent batch sizes, and dosing regimens of subsequent products (2, 3). Although each individual situation requires a thorough assessment, a carryover equivalent to 0.1% of the lowest daily therapeutic dose in the highest daily dose of subsequently prepared products is generally acceptable (2).

Analytical methods to support residue determinations for cleaning verification are an essential part of good manufacturing processes. These methods can be either nonselective (e.g., total organic carbon [4] and gravimetric analysis [5]) or compound-specific. For compound-specific assays, high-performance liquid chromatography–ultraviolet spectrometry (HPLC–UV) is most commonly used for the separation and detection of drug residues (6, 7, 8, 9). Nonetheless, determinations by gas chromatography (GC) (10), thin-layer chromatography (TLC) (11), ion-mobility spectroscopy (12) and enzyme-linked immunosorbent assay (ELISA) (13) also have been reported in the literature. For low-dosage drugs, whose development is increasingly common (14), or compounds with poor UV chromophores, the required residue limits can present significant analytical challenges. In addition, dosage-form manufacturing equipment is now available that makes use of very small batch sizes (gram scale) (15). This smaller batch size can lead to residue limits that are significantly lower (10–100-fold) than those calculated using more typical manufacturing equipment because the batch size is a multiplier in the residue limit calculation (2,3), and the available swabbing area is considerably smaller than typical manufacturing equipment. This article explores the use of LC–MS to achieve the sensitivity and selectivity necessary for the determination of low-level drug residues in performing cleaning verification analyses.

Atmospheric-pressure ionization LC–MS has revolutionized many scientific disciplines in drug development and discovery (16). Tandem mass spectrometry (LC–MS–MS) is the standard technique for the quantitation of drug and metabolite levels in bioanalytical chemistry. Quantitation at the picogram-per-milliliter level is readily achievable (17,18) because of improvements in selectivity and sensitivity compared with more conventional quantitative techniques. Our laboratory has previously established that LC-MS-MS can be used with acceptable precision, accuracy, and linearity to quantitate drug-product impurities (19). Because of LC–MS–MS's selectivity, it is most successfully applied to known compounds and, as such, can be readily applied to compound-specific assays for the support of cleaning verification. Although this technique is not commonly used in a regulatory or manufacturing environment, newer requirements for low-level analysis, rapid-method development and quick sample-turnaround will make this technique advantageous for early-stage drug development. One minor drawback to using quantitative LC–MS–MS is that rugged analysis generally requires the use of an internal standard (20). In typical bioanalytical applications, the internal standard is added to each sample and standard before extraction and analysis. Because the addition of an internal standard to the manufacturing equipment would not be an acceptable practice, various addition points for the internal standard will be evaluated and techniques for using internal standards to troubleshoot recovery problems will be discussed. In this article, methods are developed and validated for the low-level residue determination of four compounds of different pharmaceutical classes.

Experimental

Materials. Compounds A–D and associated internal standards (see Figure 1 for structures) were obtained from Pfizer Global R&D, Michigan Laboratories. Polyester swabs (ITW Texwipe, Upper Saddle River, NJ) were used without any additional precleaning steps. Reagents and HPLC-grade solvents (Mallinckrodt, Phillipsburg, NJ) were used without further purification. HPLC columns were obtained from Phenomenex (Torrance, CA), Waters Corporation (Milford, MA), ThermoElectron Corporation (Waltham, MA), and Varian (Lake Forest, CA).

Figure 1: Structures of Compounds A, B, C, and D.

Sample preparation. Swabbing locations are predefined areas on pieces of the manufacturing equipment, which are representative of the most difficult regions to clean. Typical equipment swabbing procedures involve wetting a swab with swabbing solvent and passing the swab over the predefined equipment area. In general, several orthogonal passes are made with each swab, and in some cases, multiple wetted swabs may be required. A dry swab to recover additional deposited solvent immediately follows the wet swabs. The swab tips are collected in a sample vial, immersed in extraction solvent, and extracted by shaking or sonication. The extracted solution is then analyzed for compound residue. Swabbing conditions for Compounds A–D are summarized in Table I.

Table I: High-performance liquid chromatography method and swabbing conditions for Compounds A, B, C, and D.

For analyte recovery experiments, a known amount of the analyte is spiked onto surfaces representative of the manufacturing equipment (e.g., stainless steel or PTFE plates). The surface is allowed to dry completely. The swabbing procedure is executed on the plate surfaces, and the samples are analyzed to assess analyte recovery.

Standard curve preparation. Stock standard solutions were prepared from neat materials in swab solvent (see Table I) for each drug compound. Stock solutions were diluted with the respective swab solvent to prepare a standard curve with a minimum of eight points per curve. The range of the standard curves minimally covered the equivalent of 30–125% of the surface residue limits. Ranges were 100–10,000 ng/mL for Compound A, 10–500 ng/mL for Compound B, 10–600 ng/mL for Compound C, and 20–400 ng/mL for Compound D.

Internal standard preparation and addition. Internal standard stock solutions were prepared in diluent for each drug compound. Standard and sample solutions were prepared for analysis by adding 1.0 mL of each to individual HPLC vials containing 100 μL of internal standard stock solution. Internal standard stock solution was prepared to give a final concentration in the analytical sample that was midrange in the standard curve.

For internal standard recovery experiments, surfaces, swabs, or sample vials are spiked with internal standard so that the same final sample concentration of the internal standard is achieved.

Chromatography conditions. Chromatographic separations were performed on Agilent 1100 HPLC systems (Agilent Technologies, Wilmington, DE). Chromatographic conditions for Compound A using HPLC with UV detection are detailed in Table I. Because of the poor UV response of this compound, low-wavelength UV detection (210 nm) is used in addition to high sample loading (100 μL injection).

Chromatographic conditions for LC–MS–MS analysis were optimized to give adequate separation from the void to minimize ion suppression effects, while still allowing for rapid analysis times. Chromatographic details for Compounds A–D are detailed in Table I.

Mass spectrometry conditions. The LC–MS and LC–MS–MS experiments were performed using a Micromass Quattro Micro triple quadrupole electrospray ionization mass analyzer (Waters, Beverly, MA). The capillary voltage, source temperature, and desolvation temperature were kept constant for all experiments at 3.5 kV, 120 °C, and 400 °C respectively. The entrance and exit energies for the multiple-reaction monitoring (MRM) experiments were also kept constant at –1 and 5, respectively. For all experiments, quadruple resolutions were set to 13, cone gas flow to 100 L/h, and desolvation gas flow to 500 L/h. The cone voltage and collision energies were adjusted to optimize the instrumental response for each individual compound. All LC–MS–MS experiments were performed using MRM, and transitions were chosen for each individual compound to maximize sensitivity and selectivity. Each transition was monitored for 0.2 s with a 0.05 s interchannel delay.

MS–MS analysis parameters were optimized for each analyte (see Table II). First, source conditions were selected to maximize a parent mass signal. Collision experiments were performed to identify a selective fragment for use in the MRM experiment. Finally, collision cell parameters were adjusted to optimize the signal of the daughter fragment.

Table II: Mass spectrometry–mass spectrometry parameters for Compounds A, B, C, and D.

Data treatment. All data were processed using MassLynx software with QuanLynx data quantitation package (version 3.5, Micromass). MRM chromatograms were collected with 1.0 Da windows. All data were mean smoothed twice. The data system was configured to calculate automatically and annotate the areas of the product ions of the analyte, and when applicable, the internal standard. A calibration curve was constructed using the appropriate regression analysis with 1/X weighting. All concentrations were calculated from the peak areas or peak area ratios (internal standard) against the calibration curve.

Results and discussion

Representative structures of the APIs and internal standards are given in Figure 1. Exact structures are not shown because of the proprietary nature of these compounds, but the ionizable functional groups are shown. Compound A has no significant UV chromophore. This compound, however, has significant response in positive ion electrospray LC–MS. Compounds B, C, and D have excellent UV sensitivity. Mainly because of the use of new manufacturing equipment (15), much lower residue limits were necessary for these compounds (0.1 μg/in2 compared with 10 μg/in2 residue limit for Compound A). Given the reduced residue limits, LC–MS–MS detection with its enhanced selectivity and sensitivity was chosen for determination of the residue concentrations. Stable isotopic-label internal standards of both compounds A and B were available for method development, validation, and analysis. Compounds C and D did not have stable isotopic-labeled analogues available for use. As a result, structural analogues were chosen as internal standards. The internal standard chosen for Compound C was a fluoro-substituted analogue, and an internal standard with a methyl (rather than ethyl) substituted pyrazol was used for Compound D.

The improved sensitivity of LC–MS–MS when compared with the more commonly applied HPLC–UV is shown for Compound A (poor UV chromophore) in Figure 2. Using HPLC–UV methodologies, Compound A is not observed at the required residue limit concentration of 100 ng/mL (see Figure 2a) despite high sample loading (100 μL injection) and low wavelength UV detection (210 nm). Furthermore, potential peak interferences from swab contaminants or extractables are observed in the HPLC–UV trace (see Figure 2b). In Figure 2c, the residue limit concentration is shown using the LC–MS–MS method where compound A is observed with a signal to noise ratio of 203:1. The superior selectivity of LC–MS–MS analysis is shown in Figure 2d and Figure 2e, which indicate adequate selectivity against swab extractables and cleaning solution residues.

Figure 2: Comparison of sensitivity and selectivity of LC–UV and LC–MS–MS for determining Compound A in cleaning verification samples. LC–UV traces for Compound A including (a) 100 ng/ML Compound A (b) swab blank extract. LC–MS–MS traces for Compound A including (c) limit of quantitation (100 ng/mL), (d) swab blank extract, (e) cleaning agent solution, and (f) internal standard.

Compound A uses a stable isotopic-labeled internal standard, which is traditionally added to improve method precision and accuracy for LC–MS analysis. With the stable isotopic-labeled internal standard, chromatography is typically identical to that of the parent compound (see Figure 2f) and as a result, it is subjected to the identical ionization environment as the analyte of interest. Electrospray ionization can be influenced greatly by the matrix effects from the swab, cleaning solution, or from the analytical system itself that can result in enhanced or reduced MS response (ion suppression) (21). Because response ratios are used for quantitative evaluations, the stable, isotopic-labeled internal standard mitigates the potential effects of ion suppression (22).

Whereas single-point calibrations commonly are used with UV detection, quantitative LC–MS generally requires a calibration curve (19). In these methods, calibration is achieved by regression analysis of standard curve data covering roughly two orders of magnitude. Sample peak areas (or peak area ratios with internal standard) are used to quantitate samples with respect to the calibration curve. Results from the standard curve of Compound B (see Table III) outline the process of generating an appropriate calibration. The calibration range is indicated in injection solution concentration (10 to 1500 ng/mL) as well as residue limits on a 4-in2 swab area. The average peak areas of the analyte and internal standard are shown in columns 4 and 5, and the peak area ratio of the analyte to the internal standard is listed in column 6. Linearity can be assessed by simply noting correlation coefficients (r2 ). Over large concentration ranges, however, the use of response factors is generally superior (23). Response factors are calculated by dividing peak response (peak area, or peak area ratio for internal standard calibration) by standard concentration. For a truly linear response, the response factor should be independent of concentration over the range of the method.

Table III: Tabular results for Compound B curve fitting.

Linearity can be problematic with LC–MS analysis because of the ionization efficiency of the electrospray source. As concentration increases, the ionization efficiency can decrease and may require a quadratic fit for the calibration curve. This effect is clearly demonstrated by observing the peak areas of the analyte and internal standard with increasing analyte concentrations (see Table III). Theoretically, the peak area of the internal standard should be consistent throughout the curve. In this electrospray source, however, the peak intensity decreases significantly as the analyte concentration increases. This effect also can be seen by monitoring response factors for the external calibration (see Table III, column 7), which decrease by roughly 30% at the high end of the curve. In the case of an internal standard calibration with a stable isotopic-label, the analyte and the internal standard are subjected to the same reduced ionization efficiency. Because the analyte and internal standard are affected equally, the ratio of their peak areas is unaffected by the reduced efficiency, resulting in a significantly better linear curve fit. Response factors for internal standard calibration (see Table III) illustrate this effect, showing only a 10% change over the range of the method. Another sensitive measure of goodness of fit is the curve residuals (see Table III, columns 9 and 10), which were used to assess fit quality during method validation (24).

In an effort to improve linearity of the analyte response, an alternate ionization source was explored. When atmospheric pressure chemical ionization (APCI) is used in place of the electrospray source, signal linearity is significantly improved, even without the use of an internal standard (data not shown). Unfortunately, the sensitivity obtained was insufficient for this application, and APCI ionization was not further pursued.

An acceptable curve fit is achieved when a regression analysis has a correlation coefficient of not less than 0.98 over the range tested (covering at least 30–120% of the API residue limit). The residuals measure the relative error of the regression; acceptable limits are ±10% over the entire range of the calibration curve. Appropriate fittings and weightings for calibration curves are determined by minimizing the residuals of the correlation (24).

Although not formally governed by the ICH (25), validation elements and required criteria consistent with this guidance are presented in Table IV. An acceptable analytical method must be selective (absence of interference) against surface extractables, swab extractables, cleaning solvents, and interferences from the analytical system itself. The system precision, a measurement of a single standard repeatedly injected, should give not more than 5.0% relative standard deviation (RSD). The signal-to-noise ratio at the quantitation limit should be at least 10:1, with precision of not more than 10% RSD. Method accuracy is determined by aliquoting (i.e., spiking) known amounts of the API onto a surface representative of the manufacturing equipment (e.g., stainless steel or PTFE). The surfaces are then swabbed and analyzed per the method. Acceptable recoveries are not less than 85%. Lower recovery values, however, may be acceptable if a recovery correction factor is included in the residue calculation. The closeness of repeated recovery results is a measure of method precision, which should be not more than 10% RSD.

Table IV: Validation results with and without internal standard (IS) for Compound A.

Both linear and quadratic curve fit results are assessed using 1/X weighting. Columns three and four of Table IV show the results for Compound A without an internal standard. With a linear fit, the validation does not meet the requirements for accuracy of the curve fit (residuals ≤10%). When a quadratic fit is applied for the regression analysis (see Column 4), the correlation coefficient is improved, and the residuals are significantly decreased at all points on the curve when compared with the linear regression. Columns five and six of Table IV show the performance with the inclusion of an internal standard. When internal standard calibration is applied, improved correlation coefficients and residuals are observed, even with a linear regression analysis.

Validation data for Compound B using both linear and quadratic calibration curves and with and without the use of stable isotopic-labeled internal standard are presented in Table V. Without the use of the internal standard, sufficient system precision is not achieved. In addition, the method does not meet the residual requirements when a linear regression is applied without the internal standard. To achieve acceptable residuals without an internal standard, a quadratic curve fit is required. When the internal standard is incorporated into the regression analysis, residuals improve greatly for each fit, and both linear and quadratic fits give acceptable regression results.

Table V: Validation results with and without internal standard (IS) for Compound B.

Table VI shows the required validation criteria and results for Compound C. The internal standard for Compound C is not a stable isotopic-labeled compound, but one with a fluorine substitution. As a result, the internal standard does not exactly co-elute with the analyte, but has a relative retention time of 1.1. Although inclusion of a structural analogue internal standard improves results, a quadratic fit is required to give acceptable curve residuals and system precision. Recovery results for Compound C are close to the lower acceptable limits.

Table VI: Validation results with and without internal standard (IS) for Compound C.standard (IS) for Compound B.

Validation results for Compound D are compiled in Table VII. As with Compound C, the internal standard is not a stable isotopic label and has significantly different retention characteristics than the analyte. Again, the inclusion of an internal standard improves the residuals of the curve fit, correlation coefficients, and precision at the limit of quantitation. The recovery of Compound D did not meet the validation requirements set forth. Nonetheless, in cases where lower recoveries are obtained, a recovery correction factor may be applied so long as adequate method precision is achieved.

Table VII: Validation results with and without internal standard (IS) for Compound D.

As Table VI and Table VII show, lower recoveries were obtained for Compounds C and D. During the course of method development, even lower recoveries were obtained. The recovery issues suggest the compounds were difficult to extract from the swab solution, the swab tip, or the surface.

The internal standard used for quantitative LC–MS–MS experiments also can be used to investigate the extraction methods. To evaluate low-recovery problems, the internal standard compound can be spiked into the sample at progressively earlier points in the analysis. For typical sample analysis, the internal standard is added to the HPLC vial with a measured aliquot of sample solution. The internal standard can be added to the sample vial, spiked onto the swab material, or even spiked onto the surface to be analyzed as part of the recovery experiment. The internal standard compounds should have similar properties to the analyte of interest and should be extracted with similar efficiencies. As a result, the analyte and internal standard collectively spiked onto a surface should be recovered in equal ratios. Because the response ratios are used for quantitation by the internal standard method, the recoveries for this experiment should theoretically be close to 100%.

Analyte recoveries with internal standard spiking at different points in the extraction process for both Compound C and Compound D are shown in Table VIII. Recoveries for Compound C generally increase as the internal standard addition point is moved earlier in the extraction process. Recovery does not increase significantly going from one step to the next, suggesting that the contributions to the recovery problems for Compound C are spread out roughly equally among surface, swab, and solution recoveries.

Table VIII: Recovery results and precision (N=6) for Compound C and Compound D with the addition of internal standards (IS) in various stages of the analytical method. Reported recoveries are from stainless steel surfaces (4 in2).

Recovery results from the internal standard spiking experiment for Compound D clearly demonstrate that recovery from the surface is the major contributor to poor recovery for this analyte. Because increased recovery did not result from the addition of the internal standard to the swab, additional development around the swab extraction would not yield improved recoveries. Knowing that the source of the recovery problems was extraction from the surface, method development activities could be focused in this area. Subsequent experiments (see Table VII) showed acceptable recovery results with the larger Texwipe TX714A swab in place of the TX761 swab typically used in our laboratories.

The internal standard spiking method described above can be a useful tool in identifying the source of reduced analyte recovery. In cases for which extraction from the swab tip or sample solution are identified as the source of recovery problems, the analytical method can be adapted readily to include spiking of internal standard onto the swab tips or into the sample solution. If surface recovery is the major contribution to poor recovery, however, method development activities must continue because spiking the internal standard onto the manufacturing equipment before swabbing is not an appropriate solution.

Conclusions

Liquid chromatography–mass spectrometry can be applied readily to support cleaning validation in the manufacture of pharmaceutical products. The sensitivity of this analytical method enables the detection of extremely low levels of the drug, showing an advantage over the more traditional liquid chromatography–utlraviolet spectrometry. More importantly, the selectivity advantages of liquid chromatography–mass spectrometry–mass spectrometry (LC–MS–MS) as compared with liquid chromatography–ultaviolet spectrometry render extensive method development for liquid chromatography unnecessary, allowing for rapid development of fast chromatographic runs. Short run times for cleaning verification are essential because the manufacturing equipment is placed in quarantine until it has been verified to be sufficiently clean.

In this article, the authors demonstrated the development of LC–MS–MS methods for several compounds that meet generally accepted validation criteria. The use of an internal standard enhanced the linear response when compared with an external calibration and helps to improve system precision and fit residuals. Both stable isotopic-label and structural analogue internal standards are acceptable. Stable isotopic-label internal standards should be used, however, when available. Analysis can be achieved without an internal standard, but a quadratic fit was typically required, and the system precision and fit residuals generally suffer.

Internal standard spiking techniques were demonstrated to be valuable tools for the evaluation and potential improvement of recovery issues of drug residues from the cleaning verification method. Evaluation of data from the internal standard spiking experiments described herein allow for subsequent method development efforts to be focused on the major sources of low-recovery problems.

LC–MS–MS has outstanding potential to become a routine technique for the determination of drug residues in support of cleaning verification, especially early in drug development. Certainly, in applications where sensitivity and selectivity are inadequate using traditional modes of detection, LC–MS–MS should be given serious consideration as an alternate mode of detection. The faster development and analysis time will position LC–MS–MS to become the predominant technique of choice for detecting drug residues on manufacturing equipment.

Kevin J. Kolodsick is a senior scientist, Holly Phillips is an associate scientist, Jennifer Feng is a principal scientist, and Carol A. Kingsmill* is a senior principal scientist at Analytical Research and Development, Michigan Pharmaceutical Sciences, Pfizer Global Research and Development, 2800 Plymouth Road, Ann Arbor, MI 48105, tel. 734.622.1779, carol.Kingsmill@pfizer.comMatthew Molski was a student intern at the College of New Jersey (Ewing, NJ) at the time of research.

*To whom all correspondence should be addressed.

Submitted: Sept. 21, 2005. Accepted: Sept. 30, 2005.

References

1. International Conference on Harmonization, Q3B: Impurities in New Drug Products (ICH Geneva, Switzerland, Feb. 5, 2003).

2. D.A. LeBlanc, "Establishing Scientifically Justified Acceptance Criteria for Cleaning Validation of APIs," Pharm. Technol. 24 (10), 160–168 (2000).

3. D.A. LeBlanc, "Establishing Scientifically Justified Acceptance Criteria for Cleaning Validation of Finished Drug Products," Pharm. Technol. 22 (10), 136–148 (1998).

4. K.M. Jenkins et al., "Application of Total Organic Carbon Analysis to Cleaning Validation," PDA J. Pharm. Sci. Technol. 50 (1), 6–15 (1996).

5. "Points to Consider for Cleaning Validation–Technical Report No. 29," PDA J. Pharm. Sci. Technol. 52 (6), 1–23 (1998).

6. M.B. Boca, Z. Apostolides, and E. Pretorius, "A Validated HPLC Method for Determining Residues of a Dual Active Ingredient Anti-Malarial Drug on Manufacturing Equipment Surfaces," J. Pharm. Biomed. Anal. 37 (3), 461–468 (2005).

7. M.J. Nozal et al., "Development and Validation of an LC Assay for Sumatriptan Succinate Residues on Surfaces in the Manufacture of Pharmaceuticals," J. Pharm. Biomed. Anal. 30 (2), 285–291 (2002).

8. J. Lambropoulos, G.A. Spanos, and N.V. Lazaridis, "Development and Validation of an HPLC Assay for Fentanyl, Alfentanil, and Sufentanil in Swab Samples," J. Pharm. Biomed. Anal. 23 (2-3), 421–428 (2000).

9. T. Mirza et. al, "Cleaning Level Acceptance Criteria and a High Pressure Liquid Chromatography Procedure for the Assay of Meclizine Hydrochloride Residue in Swabs Collected from Pharmaceutical Manufacturing Equipment Surfaces," J. Pharm. Biomed Anal. 19 (5), 747–756 (1999).

10. T. Mirza et al., "Capillary Gas Chromatographic Assay of Residual Methenamine Hippurate in Equipment Cleaning Validation Swabs," J. Pharm. Biomed. Anal. 16 (6), 939–950 (1998).

11. Z. Katona et al., "Cleaning Validation Procedure Eased by Using Overpressured Layer Chromatography," J. Pharm. Biomed. Anal. 22 (2), 349–353 (2000).

12. R. Debono et al., "Using Ion Mobility Spectrometry for Cleaning Verification in Pharmaceutical Manufacturing." Pharm. Technol. North America 26 (4), 72–78 (2002).

13. F. J. Rowell, Z.F. Miao, and R.N. Neeve, "Pharmaceutical Analysis Nearer the Sampling Point: Use of Simple, Rapid On-Site Immunoassays for Cleaning Validation, Health and Safety, and Environmental Release Applications," J. Pharmacy and Pharmacology 50, 47 (1998).

14. J. Uetrecht, "Prediction of a New Drug's Potential to Cause Idiosyncratic Reactions," Curr. Opin. Drug Discov. Devel. 4 (1), 55–59 (2001).

15. Capsugel, (www.xcelodose.com, Morris Plains, NJ), Accessed Jan. 12, 2006.

16. D.T. Rossi, "The Impact of Atmospheric Pressure Ionization," in Mass Spectrometry in Drug Discovery, D. T. Rossi and M. W. Sinz, Eds. (Marcel Dekker, New York, 2002), pp. 15–24.

17. B.K. Matuszewski, M.L. Constanzer, and C.M. Chavez-Eng, "Matrix Effect in Quantitative LC–MS–MS Analyses of Biological Fluids: A Method for Determination of Finasteride in Human Plasma at Picogram Per Milliliter Concentrations," Anal. Chem. 70 (5), 882–889 (1998).

18. T.K. Majumdar et al., "Trace Level Quantitation of Iralukast in Human Plasma by Microbore Liquid Chromatography–Tandem Mass Spectrometry," Rapid Commun.Mass Spectrom. 14 (6), 476–481 (2000).

19. K.J. Kolodsick, D.T. Rossi, and C.A. Kingsmill, "Breaking Down Barriers: Can LC–MS Revolutionize the Quantitation of Drug Product Impurities?" LCGC North America 21 (5), 468–479 (2003).

20. J.D. Gilbert, T.V. Olah, and D.A. McLoughlin, "High Performance Liquid Chromatography with Atmospheric Pressure Ionization Tandem Mass Spectrometry as a Tool in Quantitative Bioanalytical Chemistry," in Biochemical and Biotechnological Applications of Electrospray Ionization Mass Spectrometry, P. Snyder Ed. (ACS Symposium Series #619, Washington DC, 1996), pp. 330–350.

21. T.M. Annesley, "Ion Suppression in Mass Spectrometry."Clin. Chem. 49 (7), 1041–1044 (2003).

22. L.E. Sojo, G. Lum, and P. Chee, "Internal Standard Signal Suppression by Co-Eluting Analyte in Isotope Dilution LC-ESI-MS," The Analyst 128 (1), 51–54 (2003).

23. L.R. Snyder, J.J. Kirkland, and J.L. Glajch, Practical HPLC Method Development (Wiley-Interscience, New York, 2nd ed., 1997), pp. 692–693.

24. M.M. Kiser and J. W. Dolan, "Selecting the Best Curve Fit," LCGC Europe 17 (3), 138–143 (2004).

25. International Conference on Harmonization, Q2B: Validation of Analytical Procedures: Methodology (ICH Geneva, Switzerland, Nov. 6, 1996).