Peter R. Stout*, Nichole D. Bynum, John M. Mitchell, Michael R. Baylor, and Jeri D. Ropero-Miller

A Comparison of the Validity of Gas Chromatography Mass Spectrometry and Liquid Chromatography Tandem Mass Spectrometry Analysis of Urine Samples for Morphine, Codeine, 6-Acetylmorphine, and Benzoylecgonine Peter R. Stout*, Nichole D. Bynum, John M. Mitchell, Michael R. Baylor, and Jeri D. Ropero-Miller Research Triangle Institute International, Center for Forensic Sciences, 3040 Cornwallis Road, P.O. Box 12194, Research Triangle Park, North Carolina 27709 Abstract On November 25, 2008, the U.S. Department of Health and Human Services posted a final notice in the Federal Register authorizing the use of liquid chromatography tandem mass spectrometry (LC MS MS) and other technologies in federally regulated workplace drug testing (WPDT) programs. These rules are expected to become effective in May 2010. To support this change, it is essential to explicitly demonstrate that LC MS MS as a technology can produce results at least as valid as gas chromatography mass spectrometry (GC MS), the longaccepted standard in confirmatory analytical technologies for drugs of abuse and currently the only confirmatory method allowed for use in support of federally regulated WPDT programs. A series of manufactured control urine samples (n = 10 for each analyte) containing benzoylecgonine, morphine, codeine, and 6-acetylmorphine at concentrations ranging from 10% to 2000% of federal cutoffs were analyzed with replication by five federally regulated laboratories using GC MS (five replicate analyses per lab) and at RTI International using LC MS MS (10 replicate analyses). Interference samples as described in the National Laboratory Certification Program 2009 Manual were also analyzed by both GC MS and LC MS MS. In addition, matrix effects were assessed for LC MS MS, and both analytical technologies were used to analyze previously confirmed urine specimens of WPDT origin. Results indicated that LC MS MS analysis produced results at least as precise, accurate, and specific as GC MS for the analytes investigated in this study. Matrix effects, while evident, could be controlled by the use of matrix-matched controls and calibrators with deuterated internal standards. LC MS MS data parameters, such as retention time and product ion ratios, were highly reproducible. * Author to whom correspondence should be addressed. E-mail: pstout@rti.org. Introduction On November 25, 2008, the U.S. Department of Health and Human Services (HHS) published a final notice in the Federal Register allowing the use of liquid chromatography tandem mass spectrometry (LC MS MS) and other technologies in federally regulated workplace drug testing (WPDT) programs (1). This is anticipated to become effective in May 2010. To support this change, it is essential to explicitly demonstrate that LC MS MS as a technology can produce results at least as valid as gas chromatography mass spectrometry (GC MS). GC MS has been the accepted standard in confirmatory analytical technologies for drugs of abuse in urine and is currently the only confirmatory method allowed for use in support of federally regulated WPDT programs. A very large body of literature is available with validation studies for GC MS methods (2,3). Numerous validation studies have been published utilizing LC MS MS methods for a wide variety of target analytes in urine, oral fluid, and blood (3 11). Most of these studies provide data documenting accuracy, precision, linearity, and interferences (12,13). These criteria are required to demonstrate that a method is valid and capable of producing accurate, precise, and reliable results. Additional data are needed to support the minimal technical criteria that define acceptable instrument and batch performance for the application of LC MS MS in WPDT. This study attempted to evaluate the detection of benzoylecgonine (BZE), morphine (MOR), codeine (COD), and 6- acetylmorphine (6-AM) in urine using LC MS MS. In a previous study, Fox and colleagues (11) discussed issues of identification of some opiates using LC MS MS. One goal of this study was to evaluate and compare the detection of MOR, 398 Reproduction (photocopying) of editorial content of this journal is prohibited without publisher s permission.

COD, and 6-AM as the only target opiate analytes of federally regulated WPDT programs and not to distinguish all of the common opiates, as Fox and colleagues (11) reported. However, because issues of identification are pertinent, the potential interference from norcodeine (NCOD), hydromorphone (HYM), hydrocodone (HYC), oxymorphone (OXM), and oxycodone (OXC) was also examined. LC MS MS has some notable differences from GC MS. It is generally more sensitive than GC MS; however, there is the potential for greater variability in product ion formation in analyses using multiple reaction monitoring (MRM) or selected reaction monitoring (SRM) (3). Additionally, LC methods are more susceptible to matrix effects, which is the potential for components of the sample matrix to either suppress or enhance the ionization of target analytes (14 17). Therefore, for LC MS MS method validation, understanding the potential matrix effect, as well as interference(s), is critical. Interference studies for GC MS are required to investigate structurally similar drug analytes that may interfere with testing, whereas interference studies for LC MS MS should investigate interference from both structural analogues and matrices. Direct comparisons of results obtained from GC MS and LC MS MS using concurrent analyses of the same sample materials have not been apparent in the scientific literature. This study was designed to include such analyses to assess the ability of LC MS MS to produce results at least as valid as GC MS by directly comparing results from both analytical technologies for the same sets of manufactured urine samples and the same sets of previously confirmed urine specimens. Methods Reagents and analytes All solvents used were analytical-grade solvents and reagents purchased from Fisher Scientific (Fair Lawn, NJ), Burdick and Jackson (Muskegon, MI), or Sigma-Aldrich (Deisenhofen, Germany). Solid-phase extraction columns were purchased from SPEware (San Pedro, CA). Drug-free human urine was collected from volunteers under Institutional Review Board-approved protocols with consent. Validation samples were manufactured at RTI International (Research Triangle Park, NC) using manufactured stock materials separate from the stock materials used for controls and calibrators. Most of the drug analytes used in the manufacture of fortified urine samples were purchased from commercial sources. MOR, OXC, HYC, and HYM were obtained from Sigma (St. Louis, MO). OXM, NCOD, and 6-AM were obtained from Alltech (Deerfield, IL). BZE and COD were provided by the National Institute on Drug Abuse (NIDA) from the NIDA drug supply repository housed at RTI. Calibrators for LC MS MS analyses were prepared by RTI using negative human urine, and all stock drug materials were obtained from Cerilliant (Round Rock, TX). Controls for GC MS and LC MS MS analyses were obtained from National Laboratory Certification Program (NLCP) Performance Testing (PT). The drug concentrations in the controls were 90 and 1500 ng/ml for BZE; 1500 and 2750 ng/ml for MOR and COD; and 8 and 80 ng/ml for 6-AM. Manufacture of validation samples In accordance with the NLCP guidance for method validation, the expanded series of samples was manufactured in drug-free human urine. Two series of fortified urine samples were manufactured: one contained MOR/COD/6-AM (OPI series) and the second contained benzoylecgonine (BZE series). Each series contained the analytes at concentrations across three orders of magnitude. The validation samples for linearity, precision, and accuracy (LPA) determinations with both GC MS and LC MS MS are shown in Table I. Validation of the opiate analyses also included interference samples constructed in drug-free human urine containing COD, MOR, and 6-AM at 40% of their respective cutoff concentrations and interfering analytes (i.e., NCOD, HYM, HYC, OXM, and OXC), as described in NLCP guidance for method validation (18) (Table II). All manufactured samples containing MOR were formulated with unconjugated drug to eliminate variations between laboratories due to differences in hydrolysis protocols. Table I. Details of the Distribution of Sample Concentrations in the Linearity, Precision, and Accuracy (LPA) Samples Concentration of Drug (ng/ml) Sample % of Opioid series BZE Number cutoff COD MOR 6-AM series 1 10% 200 200 1 15 2 20% 400 400 2 30 3 40% 800 800 4 60 4 75% 1500 1500 8 113 5 100% 2000 2000 10 150 6 125% 2500 2500 13 188 7 200% 4000 4000 20 300 8 500% 10,000 10,000 50 750 9 1000% 20,000 20,000 100 1500 10 2000% 40,000 40,000 200 3000 Table II. Detail of Drugs and Sample Concentrations (ng/ml) Included in Interference Study Sample Sample Sample Sample Sample Sample Drug 1 2 3 4 5 6 Codeine 800 800 800 Morphine 800 800 800 6-Acetylmorphine 4 4 4 Hydrocodone 5000 5000 Hydromorphone 5000 5000 Oxycodone 5000 5000 Oxymorphone 5000 5000 Norcodeine 5000 5000 399

Linearity, precision, and accuracy determination by LC MS MS and GC MS analyses The LPA sample series (Table I) was analyzed by five HHScertified drug testing laboratories using GC MS and was also analyzed by RTI using LC MS MS. The samples, along with control materials derived from past NLCP PT materials, were shipped frozen overnight from RTI to the laboratories. These control materials provided known reference points to allow deviations from expected performance to be attributed to issues within the laboratory or to the manufactured sample materials. Each laboratory was provided with sufficient sample volume for all requested analyses and was compensated for their analyses. The OPI and BZE series of manufactured samples were analyzed with each laboratory using its standard validated GC MS methods. Each laboratory was instructed to extract one set of samples and to analyze each extract five times by GC MS using five separate calibrations within a five-day period. Because the purpose of the study was to evaluate the differences in analytical method performance, this protocol limited the variation because of extraction procedures. RTI also extracted and analyzed the OPI and BZE series using LC MS MS for a total of 10 separate extractions and analyses for each series over a 5-day period (i.e., two extractions were conducted each day). Interference samples Opiates interference samples (Table II) also were shipped to the five separate reference laboratories for GC MS analysis. Laboratories were instructed to analyze the materials once using their standard opiates extraction and GC MS procedures. Five replicate LC MS MS analyses of these materials were performed at RTI. Matrix effect samples Matrix effect samples were analyzed only by LC MS MS at RTI. Matrix effects were evaluated using the methods described by Matuszewski and colleagues (17). In brief, three sets of samples were created for each target analyte. Type A samples (neat in mobile phase) were made by preparing an amount of target analyte and internal standard in mobile phase equivalent as in sample types B and C. Type B samples (post-extraction spike) were made by fortifying the eluent from the solid-phase extraction of negative urine matrix with target analytes and internal standard. Type C samples (pre-extraction spike) were made by fortifying negative urine matrix with target analytes and internal standard prior to solid-phase extraction. Target analyte concentrations were prepared at cutoff concentrations for each drug: 2000 ng/ml MOR/COD, 150 ng/ml BZE, and 10 ng/ml 6-AM. For the purposes of this study, 10 lots of urine, each from a different donor, were collected and samples were analyzed once. As described by Matuszewski and colleagues (17), comparative calculations were used to evaluate the data: ME (%) = B/A 100 RE (%) = C/B 100 PE (%) = C/A 100 where A, B, and C = the mean responses as represented by the area under the peaks for target and internal standard quantitative ions, ME = the matrix effect, RE = the recovery effect ( recovery of the extraction procedure ), and PE = the process effect ( process efficiency ). The mean responses for A, B, and C were determined across these 10 urine lots. Additionally, a comparison of the relative matrix effect was accomplished by comparing the percent coefficient of variation (%CV) in the response across the 10 matrix lots for sample types B and A. This provided a comparison of the variation due to analysis as represented by the %CV of sample type A (neat in mobile phase) with the variation due to matrix represented by the %CV of sample type B (post-extraction spike). Previously confirmed specimens Previously confirmed urine specimens were obtained from Laboratory Corporation of America Holdings (LabCorp, Research Triangle Park, NC). All specimens had been slated for destruction, and all identifiers were removed prior to shipment to RTI. Urine specimens from LabCorp were previously confirmed positive for BZE, COD, MOR, and/or 6-AM. LabCorp reanalyzed the specimens using GC MS concurrently with RTI s LC MS MS analysis. Sixty BZE and 46 opiate samples previously confirmed positive were utilized. LC MS MS analysis conducted at RTI BZE extraction. Fifty microliters of methanolic deuterated internal standard (ISTD) (final concentration: 500 ng/ml BZE-d 3 ) was added to a 1-mL sample of urine, and 3 ml of sodium phosphate buffer (100 mm, ph 6) was added to the sample. The sample was vortex mixed and centrifuged at 3000 rpm for 5 min. The sample was transferred to a solid-phase extraction column (SPEware Polychrom Clin II, 35 mg) and allowed to flow by gravity before the column was rinsed with 1 ml of deionized (DI) water followed with 1 ml of 100 mm hydrochloric acid (HCl) and then dried under nitrogen (N 2 ) at 25 psi for 2 min. The column was rinsed a second time with 1 ml each of methanol followed with ethyl acetate and then dried for 2 min at 25 psi N 2. The drugs were eluted by gravity with 2 ml of dichloromethane/isopropyl alcohol/ammonium hydroxide (80:14:2). The eluent was evaporated under a stream of N 2 to dryness at 40 C and reconstituted in 0.1 ml of 5 mm ammonium formate/acetonitrile (95:5) with 0.1% formic acid. Opiate extraction. For LPA samples in which MOR was present in non-conjugated form, 30 µl of methanolic deuterated internal standard (final concentrations: 3000 ng/ml codeine-d 3 and morphine-d 3 and 30 ng/ml 6-AM-d 3 ) was added to a 1-mL sample of urine. Two milliliters of sodium acetate buffer (100 mm, ph 4.5) and 0.5 ml of 10% hydroxylamine hydrochloride were added to the sample. The sample was vortex mixed, centrifuged at 3000 rpm for 5 min, and heated in a water bath at 60 C for 1 h. For archived samples in which the LC MS MS analysis was for total MOR, the following procedure was used: 30 µl of methanolic deuterated internal standard was added to a 1-mL urine sample. Half a milliliter of concentrated HCl and 0.5 ml 400

of 10% hydroxylamine were added to the sample. The sample was autoclaved at 121 C for 20 min (total cycle time of 50 min). Once the sample cooled, 0.5 ml of 45% potassium hydroxide (KOH) was added and the sample mixed. Then 0.5 ml of saturated sodium bicarbonate (NaHCO 3 ) was added, and the sample was vortex mixed. Previously used PT samples were included in these batches as controls. These samples contained morphine glucuronide to control for the hydrolysis of morphine in these samples. After either pretreatment step, the sample was poured into Table III. LC Method Parameters Column Flow Stop Post Injection Temp Rate Time Time Volume Analyte Column ( C) (ml/min) Gradient (min) (min) (µl) BZE Zorbax XDB- 50 0.8 5% B at 1 min 7 2 5 C 18 (1.8 µm, 40% B at 6 min 4.6 50 mm) 95% B at 7 min Opiates Zorbax XDB- 35 0.6 5% B at 0.1 min 5.5 5 5 C 18 (3.5 µm, 13% B at 3 min 4.6 50 mm) 75% B at 3.5 min 95% B at 5.5 min solid-phase extraction columns (SPEware Polychrom Clin II, 35 mg) that had been preconditioned with 2 ml each of methanol, DI water, and 0.1 M acetic acid. The column was washed with 2 ml each of acetic acid, DI water, hexane, and methanol. After the addition of the hexane and methanol, the column was dried at 25 psi for 5 min. The drugs were eluted with 2 ml of dichloromethane/isopropyl alcohol/ammonium hydroxide (80:14:2) by gravity and evaporated under N 2 to dryness at 40 C and reconstituted in 0.1 ml of 5 mm ammonium formate/acetonitrile (ACN) (95:5) with 0.1% formic acid. LC MS MS method. The LC MS MS equipment consisted of an Agilent 1200 series LC coupled to an Agilent 6410 triple-quadrupole MS with an electrospray source (Santa Clara, CA). Table III lists the LC method conditions used. The mobile phase components were (A) 5 mm formate (with 0.1% formic acid) and (B) acetonitrile (with 0.1% formic acid). All analyses were conducted in MRM mode using the MS MS conditions and ions listed in Table IV. Collision energies and capillary voltages were optimized for each ion. Calibration. The calibration method used for the quantification of each drug analyte is given Table IV. MS MS Method Parameters and Ions Used Precursor Fragmentation Collision Product Collision Product Product Capillary Dry Gas Dry Gas Ion Voltage Energy Ion 1 Energy Ion 2 Ion 3 Voltage Flow Temp. Drug (m/z) (V) Voltage (V) (m/z) Voltage (V) (m/z) (m/z) (V) (L/min) ( C) BZE 290.1 130 15 168 25 105 NA 3500 13 350 BZE-d 3 293.1 110 15 171 25 105 NA 3500 13 350 MOR 286.1 150 30 155 35 201 165 3500 11 350 MOR-d 3 289.0 150 30 201 45 157 NA 3500 11 350 COD 300.1 150 35 165 30 183 215 3500 11 350 COD-d 3 303.1 150 30 215 30 199 NA 3500 11 350 6-AM 328.3 150 25 268 25 210 165 3500 11 350 6-AM-d 3 331.2 150 30 271 30 191 NA 3500 11 350 Table V. Summary of Five-Point Calibration, Internal Standard (ISTD), and QC Sample Concentrations (ng/ml) and Calibration Types Used for LC MS MS Analyses Drug* Calibration Type Calibration Points ISTD Conc. QC Conc. Avg. r 2 Max r 2 Min r 2 BZE Linear 10, 150, 500, 500 90 and 1500 0.9997 1.0000 0.9988 (y = mx + b) 1000, 4000 MOR Quadratic 150, 500, 2000, 3000 1500 (gluc) 2750 (free) 0.9997 1.0000 0.9987 (y = ax 2 + bx +c) 20,000, 50,000 COD Quadratic (y = ax 2 + bx +c) 150, 500, 2000, 3000 1500 and 2750 0.9994 1.0000 0.9979 20,000, 50,000 6-AM Quadratic 1, 10, 50, 150, 300 (y = ax 2 + bx +c) 30 8 and 80 0.9986 0.9999 0.9909 * The n for each compound is 10. For morphine, one control contained morphine-3-glucuronide, and one contained free morphine. 401

in Table V. Calibration and data reduction were accomplished using Mass Hunter software (San Jose, CA). For all compounds, LC MS MS analysis was conducted using both a five-point calibration and a one-point calibration with a single calibrator at the federally mandated cutoff concentration (1). The calibration type (i.e., linear or quadratic) for a five-point calibration was optimized for each compound and was used throughout the study. For the acceptance of data analyzed, control quantitation for batches had to be within 20% of the target concentration for the control. At least 10 data points across the peak were needed. Even for the MOR/COD/6-AM analyses, there were typically more than 20 data points across peaks. The ratio of the product ions had to be within 20% of the average of the calibrator ratios, and the retention time had to be within 2% of the average of the calibrators. Calibrators were back-calculated against the model curve, and the back calculated value had to be within 20% of their theoretical concentration. Statistical analysis. Statistical tests conducted using Microsoft Excel 2003 (Seattle, WA) included linear regression analyses with subsequent analysis of variance (ANOVA) of the regression fit, calculation of average, standard deviation, and %CV. Student t-tests were used for the comparison of analyses of the previously confirmed specimens. For these tests, significance was assigned at the p < 0.05 level. For the comparison of GC MS and LC MS MS analyses of the LPA samples, a two-way imbalanced ANOVA was conducted (SAS PROC GLM using SAS, Raleigh, NC). If the crossed term of analytical method and concentration was significant, the two methods were compared after adjusting for the effect of the expected concentration. This allowed for the determination of significant differences between the two analytical methods at each concentration in the analyzed series. Again, significance was assigned at the p < 0.05 level. Results and Discussion Materials analyzed by both GC MS and LC MS MS The precision and accuracy of results from the five HHScertified laboratories conducting GC MS analyses are presented in Table VI, and the analogous LC MS MS results are presented in Table VII. For GC MS results, within-laboratory precision represented by %CV was very tight (< 3% CV). Between-laboratory %CV was less than 10% for all compounds except 6-AM, which Table VI. Evaluation of Precision and Accuracy of GC MS Results Regression Comparison to Target Concentrations Precision and ANOVA Analysis of Line Fit* Average Average % Accuracy Evaluation 95% 95% between within Average Accuracy Confidence Confidence p Drug n lab %CV lab %CV accuracy %CV r 2 Slope interval Intercept interval value BZE 244 5.81 1.80 100.9 3.44 0.998 1.00 0.99 to 1.00 0.2 5.3 to 5.6 0.616 MOR 237 5.59 1.51 101.9 4.60 0.992 0.98 0.97 to 0.99 18.5 71 to 111 0.723 COD 237 8.73 1.39 95.2 6.43 0.985 0.93 0.91 to 0.94 53.0 60 to 166 0.357 6-AM 229 13.15 2.98 98.9 5.46 0.990 1.03 1.02 to 1.04 0.2 1.27 to 0.77 0.631 * Regression analyses for MOR and COD excluded the highest concentration material (40,000 ng/ml) because the value exceeded the ULOL for some laboratories. Results are from five HHS-certified laboratories, each analyzing five replicates of each sample type. Samples that did not meet acceptance criteria were not included in calculations. The number of samples for each target compound varies because some laboratories did not report all results. All slopes were significantly different from zero at the p < 0.00001 level. None of the compounds had intercepts significantly different than zero. Table VII. Evaluation of Precision and Accuracy of LC MS MS Results* Precision Regression Comparison to Target Concentrations Average Average % Accuracy Evaluation and ANOVA Analysis of Line Fit overall overall Average Average 95% 95% % CV % CV accuracy Accuracy accuracy Accuracy Confidence Confidence p Drug n (5-point cal) (1-point cal) (5-point cal) % CV (1-point cal) % CV r 2 Slope interval Intercept interval value BZE 100 4.74 4.81 99.8 5.3 97.0 10.8 0.996 1.0 0.98 to 1.01 3.0 16.3 to 10.2 0.649 MOR 100 7.43 15.96 102.0 10.7 102.3 21.1 0.998 1.0 0.99 to 1.01 31.1 76 to 138 0.567 COD 100 6.23 6.26 101.7 7.5 105.5 7.0 0.996 0.96 0.95 to 0.97 144.0 10.6 to 299 0.067 6-AM 100 8.14 8.76 100.6 11.5 105.9 10.6 0.991 0.96 0.94 to 0.97 0.6 0.68 to 1.89 0.352 * Results are from 10 analyses conducted by RTI over 5 days of the 10 concentration series. The %CV in this table represents a between-day %CV for both five-point calibration and one-point calibration. %CV calculated across all concentrations analyzed. The average %CV and the average accuracies are across all concentrations tested. All slopes were significantly different from zero at the p < 0.00001 level. None of the intercepts were significantly different from zero (p < 0.05). 402

demonstrated greater between-laboratory variability with a 13.15% CV. The number of samples (n in Table VI) used for this analysis varied for each compound because samples not meeting acceptance criteria for GC MS (e.g., ion ratios outside the acceptable range) were excluded. The excluded samples (i.e., 6 for BZE, 13 for MOR, 13 for COD, and 21 for 6-AM) were at either the lowest concentration or the highest concentration of the series (i.e., samples at 10% of cutoff or 2000% BZE/BZE-d 3 (500/500 ng/ml) Codeine/Codeine-d 3 (500/3000 ng/ml) Morphine/Morphine-d 3 (500/3000 ng/ml) 6AM/6AM-d 3 (10/100 ng/ml) Figure 1. Typical chromatography and monitored MRM for each of the target compounds and their respective internal standards from the LC MS MS analysis. of cutoff). The values for controls were not included in this analysis. In Table VI, accuracy for GC MS analyses was assessed by determining % accuracy and regression analysis comparing target concentration with measured concentration. First, % accuracy (measured concentration/target concentration 100) was determined. This allowed the calculation of the average % accuracy for each drug analyte and the associated %CV. All drugs were 95.2 101.9% accurate for GC MS with accuracy %CVs less than 6.5%. The second means of assessing accuracy was a regression analysis comparing target concentration with measured concentration. If results were completely accurate and precise over the entire range tested, the regression results would theoretically be an r 2 = 1.000, a slope of 1.000, and an intercept of 0.000. A theoretically perfect analytical precision would yield an ANOVA analysis indicating that a model-fit with a slope significantly different than 0.000 and an intercept not significantly different than 0.000. As indicated in Table VI, all GC MS results were very close to ideal, indicating the results were highly precise and accurate. Because the highest concentration point (i.e., 40,000 ng/ml) for MOR and COD was above some of the laboratories upper limits of linearity (ULOL), this point was not used in the regression and accuracy analyses for GC MS results. Qualitatively, LC MS MS results demonstrated good chromatographic performance, exhibiting symmetrical peaks and at least 90% resolution with adjacent peaks. An example of typical chromatography observed in the LC MS MS is presented in Figure 1. The LC MS MS sample results as presented in Table VII were quantitated using both a five-point calibration, as described in the methods for each drug, and a one-point calibration using a cutoff calibrator (i.e., the calibration method most commonly used in workplace testing). Similar to GC MS analysis, calculations for precision and accuracy for LC MS MS results are presented in Table VII. For most analytes, the average %CV was highly comparable for five-point and one-point calibrations for LC MS MS results. MOR results were notably less precise when the quantitation was calculated by a one-point calibration at 2000 ng/ml because of results from the highest concentration points (20,000 and 40,000 ng/ml). The average overall %CV was higher for all analytes using the one-point calibration for LC MS MS analyses than the average within-laboratory %CV for GC MS results. This is due to the exclusion of points above the reported ULOL and limits of detec- 403

tion (LOD) for GC MS analyses but not for LC MS MS. Accuracy was again assessed by calculating average accuracy and performing the same regression analysis as for GC MS results. LC MS MS analyses were comparably accurate, ranging between 99.8% and 102.0% accurate when using a five-point calibration and 97.0% and 105.9% accurate when using a onepoint calibration. For the purposes of this study, the LC MS MS results obtained with a five-point calibration and best fit to a quadratic or linear equation were used for comparison with GC MS results. This comparison produced very good calibration results, as indicated by control materials and back-calculating calibrators against the model curve. When GC MS and LC MS MS results at each concentration were compared using a two-way ANOVA, no significant difference (p < 0.05) was observed at any concentration for BZE and 6-AM. For MOR, a significant difference was observed for the 40,000 ng/ml concentration (p = 0.019, LC MS MS mean = 40,200 ng/ml, GC MS mean = 36,800 ng/ml, or a 9% difference). This concentration was above the limit of linearity reported by most of the laboratories for GC MS analysis. For COD, significant differences were observed for all of the concentrations listed in Table VIII. Although statistically significant, the differences were small (< 10%) for all but the 40,000 ng/ml sample, which was above most laboratories reported limit of linearity. In the evaluation of performance testing samples in the NLCP, results are evaluated for error of quantitation deviating more than 20% or 50% from the target for any given sample (18). Table IX summarizes the comparison of GC MS and LC MS MS results evaluated with this 20% and 50% criteria. Table VIII. Summary of Significant Differences Between LC MS MS and GC MS Analysis of COD in LPA Samples as Determined by Two-Way ANOVA with Significance Assigned at p < 0.05 COD Target GC MS LC MS MS Concentration Mean Mean % (ng/ml) (ng/ml) (ng/ml) p Difference 800 761 809 0.0464 5.9 1500 1417 1517 0.0019 6.6 2000 1962 2051 0.0168 4.3 2500 2366 2514 0.0011 5.9 4000 3776 4064 0.0004 7.1 10,000 9436 10,061 0.0060 6.2 40,000 32,249 38,430 0.0015 16.1 Table IX. Summary of the Number of Samples by Each Method (GC MS and LC MS MS)* 10% 20% 40% 75% 100% 125% 200% 500% 1000% 2000% Drug LC GC LC GC LC GC LC GC LC GC LC GC LC GC LC GC LC GC LC GC BZE 1(20) 1(50) MOR 1(50) 5(20) 1(20) 5(20) 5(20) COD 1(20) 5(20) 5(20) 14(20) 6-AM 4(20) 4(20) 6(50) 5(20) 11(20) 4(20) 1(20) 2(20) * Values reflected would have been outside of either a 20% window or 50% window away from the target by concentration % of the cutoff (NLCP grading criteria for urine PT samples). Number in parentheses indicates either a 20% error or a 50% error. Table X. Summary of Comparison of Interference Samples Analyzed by GC MS and LC MS MS* GC MS Analysis LC MS MS Analysis Mean %CV Mean %CV HYC & OXC & HYC & OXC & HYC & OXC & HYC & OXC & NCOD HYM OXM NCOD HYM OXM NCOD HYM OCM NCOD HYM OXM 5000 5000 5000 5000 5000 5000 5000 5000 5000 5000 5000 5000 Drug n ng/ml ng/ml ng/ml ng/ml ng/ml ng/ml n ng/ml ng/ml ng/ml ng/ml ng/ml ng/ml MOR 5 836.4 806.4 803.3 7.4 8.3 5.9 5 844.1 781.9 792.6 1.8 4.9 2.5 (800 ng/ml) COD 5 866.2 784.6 808.6 6.8 8.0 6.8 5 871.2 877.3 873.0 3.1 2.4 3.1 (800 ng/ml) 6-AM 5 4.2 4.3 4.5 8.1 9.0 6.4 5 4.6 4.4 3.6 11.9 10.0 6.7 (4 ng/ml) * Statistical comparison by t-test with significance assigned at p < 0.05. p = 0.002. 404

For concentrations between 40% and 1000% of the cutoff, neither GC MS nor LC MS MS had any 50% errors, and there were only four 20% errors (by GC MS) for low-concentration 6-AM. This indicated that LC MS MS is at least as capable as GC MS at producing acceptable results as would be required on performance testing samples in the NLCP. Table X compares the results of the interference samples detailed in Table II when analyzed by both GC MS and LC MS MS. These samples contained the interferants (NCOD, HYC, HYM, OXC, and OXM) at 5000 ng/ml in the presence of analytes of interest at 40% of cutoff concentrations and in the absence of the analytes of interest altogether. The mean quantitative result from GC MS and LC MS MS analyses were compared using the Student t-test with significance assigned at p < Figure 2. Example chromatography for COD, MOR, and 6-AM with the presence of 5000 ng/ml of each structural analogue to evaluate as an interfering compound. The top row of chromatograms for each target is the target compound; the lower row is the ISTD. All compounds were unaffected by the presence of high concentrations of the interfering compounds. The ISTD for 6-AM had a close shouldering peak that was resolved (as determined by the valley between the peaks being 10% of the 6-AM-d 3 peak). 0.05. Only the 6-AM results from LC MS MS with OXC and OXM present were significantly different than the GC MS result (p = 0.002). Figure 2 presents typical chromatography observed for the LC MS MS analysis of samples containing structural analogue compounds as potential interferants in the presence of target analytes at 40% of cutoff concentrations. There was a slight shoulder on one 6-AM peak in the sample containing NCOD; however, this met chromatography acceptance criteria (i.e., 90% resolution). No MOR, COD, or 6-AM was detected in the interference samples without the addition of target compound and high concentrations of structural analogue compounds analyzed by GC MS. For LC MS MS, 72 ng/ml COD was quantitated in samples containing 5000 ng/ml NCOD and no COD. It was unclear whether COD was present as a contaminant in the NCOD stock material used to manufacture the samples, or if a slight carryover occurred during the manufacture of the samples. The presence of this low concentration of COD did not appear to affect the quantitation of COD in the samples containing COD at 40% of the cutoff concentration. As can be seen in Table X, the average COD result was within 10% of the target, and there was no significant difference between the GC MS and LC MS MS results by Student t-test. The only compound with significantly different GC MS and LC MS MS results was 6-AM with OXC/OXM present with an average GC MS result of 4.5 ng/ml and an average LC MS MS result of 3.6 ng/ml. Although statistically significant, this difference was small and not indicative of interference due to the presence of the structural analogues. GC MS and LC MS MS results were both unaffected by the presence of structural analogue compounds that might possibly interfere with analyses. Figure 3 presents a comparison of the average accuracy of GC MS and LC MS MS results across the concentration range tested for each target compound. The error bars represent one standard deviation (SD) for each average. From this, it is apparent that the average accuracies across the concentration range are very similar for GC MS and LC MS MS. The notable exceptions are for 6-AM GC MS results at the 10% of cutoff concentration and for MOR and COD results at 2000% of the cutoff concentration. The 10% of cutoff 6-AM and the 2000% of cutoff MOR and COD concentrations were outside the limits of quantitation for GC MS analyses for many laboratories; thus, low accuracy at these concentrations would be expected for the GC MS analysis. Evaluation of LC MS MS data components and matrix effects For LC MS MS results, Table XI summa- 405

rizes the averages and distributions of other qualifier data for the results, such as retention time (RT), ion ratios, and ISTD responses. The distributions of RTs for both target analyte and ISTD were very tight with %CVs less than 2% for all compounds. The ratio of quantitation product ion to qualifier product ion for both target compounds and ISTD demonstrated little variation with %CVs less than 5% for all target analytes except 6-AM. As would be expected, internal standard responses were more variable. The NLCP requires that internal standard ion response for the quantifying ion of a specimen be within 0.5 and 2 times of that obtained with the calibrator, the average of calibrator and controls, or the average of all calibrators, controls, and specimens in the analytical batch (18). Although the %CVs are large, few samples fell outside of current NLCP criteria. Only three BZE and two COD samples (3% and 2% of samples, respectively) exhibited low internal standard response outside the acceptable range. For MOR, the 40,000 ng/ml sample consistently exhibited low internal standard responses (8 of the 10 replicate analyses of this concentration), and this is consistent with ion suppression of the internal standard by the elevated concentration of MOR. Matrix effect was assessed only for LC MS MS results. Table XII summarizes the results of the matrix effect analysis. For COD, MOR, and 6-AM, there was a slight ion enhancement indicated by the matrix effect % greater than 100. BZE did not exhibit a matrix effect (ME % = 100). The same calculations are presented for the internal standard. Although there was an apparent ion enhancement effect, there was no relative matrix effect indicated by the similar (if not lower) variability (%CV) in responses of target compounds spiked into 10 different lots of urine (sample type B) compared with the analysis of target compounds in mobile phase only (sample type A) or compared with the variability of target compounds spiked into a single lot of urine. Figure 3. Comparison of average accuracies for GC MS and LC MS MS over all concentrations. Error bars represent 1 SD. Table XI. Comparison of LC MS MS Data Components and Variability in These Components for all Series Samples Run Average Target Average ISTD Average Average Average ISTD Drug n Qualifier Ratio %CV Qualifier Ratio %CV Target RT %CV ISTD RT %CV Response %CV BZE 100 25.4 1.09 24.8 1.06 4.12 0.21 4.12 0.21 540,599 80.07 MOR 100 49.6 4.81 53.2 2.05 0.54 0.90 0.54 1.02 61,899 32.49 COD 100 75.3 4.52 64.1 2.16 1.58 1.78 1.56 1.99 100,679 29.44 6-AM 100 38.8 11.6 47.8 7.51 2.36 0.24 2.36 0.24 5133 16.18 Table XII. LC MS MS Evaluation of Matrix Effect Based on Matuszewski and Colleagues (17) Target Ion Response ISTD Ion Response Relative Matrix Effect Variability in Single from 10 Lots Lot of Matrix %CV for %CV for Average %CV of analyte in analyte in Cutoff Calibrator Drug* ME (%) RE (%) PE (%) ME (%) RE (%) PE (%) type B # type A** Analyzed Over Five Days BZE 100 74 74 103 74 76 1.75 2.44 3.87 MOR 133 88 118 150 89 134 1.79 6.45 4.00 COD 135 86 117 154 87 134 8.75 8.45 6.88 6-AM 118 84 100 138 83 115 11 9.3 7.83 * Concentrations of the samples were targeted at 150 ng/ml BZE, 2000 ng/ml MOR/COD, and 10 ng/ml 6-AM. ME (%) = B/A 100, where A, B, and C are expressed as the average peak areas from the analysis of 10 different lots of urine matrix. RE (%) = C/B 100, where A, B, and C are expressed as the average peak areas from the analysis of 10 different lots of urine matrix. PE (%) = C/A 100, where A, B, and C are expressed as the average peak areas from the analysis of 10 different lots of urine matrix. # Type B samples are spiked with ISTD and target analyte after extraction of the matrix. ** Type A samples are prepared with an equivalent amount of target and ISTD in mobile phase. 406

Table XII also gives the variability in a single lot of urine matrix for all target analytes. There is not a relative matrix effect (% CV < 10%) in the comparison of the %CV for five separate cutoff calibrators extracted and analyzed over five days. This material was prepared for all analyses from the same lot of urine and provides an indication of the variability due to a single lot of matrix. This variability is comparable to the variability observed across 10 different lots of urine matrix. Though the matrix effect was small, there was an apparent difference in the quantitative MOR, COD, and 6-AM results of samples spiked before and after extraction. This difference is of importance when comparing the quantitative results of each sample type (A, B, and C). Table VIII summarizes quantitative results for each matrix effect evaluation sample type. Note that quantitation was based on a calibration curve constructed in urine matrix prior to extraction. As can be seen in Table XIII, there is a slight (~ 12%) but significant difference in the quantitative values calculated for MOR, COD, and 6-AM in sample type A (neat in mobile phase) from sample types B (post-extraction spike) and C (pre-extraction spike). There was no significant difference between B and C. The neat in mobile phase sample quantitations are slightly above the target concentration, and the matrix containing samples are slightly below the target concentration. This result, coupled with the apparent slight difference in matrix effect for the target compounds and ISTD evident in Table XII, highlights the importance of the use of matrix-matched calibrators and controls that are subjected to the same treatment as samples. As there was no apparent relative matrix effect or differing matrix effect between urine lots, the matrix effect observed could be controlled for by the use of matrix-matched controls and calibrators. Further work is needed to determine whether synthetic urine has a comparable matrix effect to human urine used in this study and would be appropriate to use in general. Validation of any method using a synthetic matrix should include evaluation of the potential matrix effect. Previously confirmed specimens A comparison of the results for the previously confirmed specimens presented in Table XIV summarizes the regression analysis for the results of GC MS and LC MS MS analyses. Previously confirmed samples were compared by a paired t-test and regression analysis of both analytical technologies. ISTD responses from GC MS analysis and LC MS MS analysis are presented for comparison. Also, average RT and average product ion ratios are presented for the LC MS MS data. The RT and ion ratios were very tightly distributed, and the variability of ISTD responses were lower than GC MS ISTD responses for these archived samples. There was no significant difference (p < 0.05) between the GC MS and LC MS MS analyses of BZE and 6-AM. There were significant (p < 0.05) differences between the LC MS MS analyses of MOR and COD. MOR results were on average 5% greater by LC MS MS analysis than by GC MS; so although statistically significant, the difference was very small. The COD results were on average 32% greater by LC MS MS. Likely, the difference in COD values was due to differences in the hydrolysis of samples. For LC MS MS analysis, the acid hydrolysis was conducted at 120 C (high pressure) for 20 min; Table XIII. Summary of Quantitative Results Associated with Each Matrix Effect Evaluation Sample Type with Differences Between Each Group Evaluated by a One-Way ANOVA BZE (Target 150 ng/ml) MOR (Target 2000 ng/ml) COD (Target 2000 ng/ml) 6-AM (Target 10 ng/ml) Sample Mean % Mean % Mean % Mean % Type* (ng/ml) % CV Accuracy (ng/ml) % CV Accuracy (ng/ml) % CV Accuracy (ng/ml) % CV Accuracy A 154 2.6 102.8 2231 6.5 111.6 2117 4.1 105.8 10.6 5.6 105.6 B 160 4.2 106.4 1929 5.1 96.4 1846 1.6 92.3 9.2 4.3 91.8 C 156 1.9 104.1 1978 1.8 98.9 1863 3.0 93.2 9.0 3.2 90.3 * Significantly different than sample types B and C for MOR, COD, and 6-AM (p < 0.001). Sample type A = neat in mobile phase. Sample type B = post-extraction spike. Sample type C = pre-extraction spike. Table XIV. Comparison Summary of GC MS and LC MS MS Analysis of Previously Confirmed Samples Regression t-test Ion GC MS ISTD LC MS MS Drug n r 2 Slope Result (p) Ratio %CV RT %CV Response %CV ISTD Response %CV BZE 60 0.980 1.05 0.870 23.60 1.41 4.12 0.14 60,622 117.22 123,014 50.18 MOR* 40 0.994 1.02 0.003* 40.04 2.65 0.53 1.07 57,930 91.54 6903 36.72 COD 21 0.993 0.679 0.003* 71.21 1.81 1.57 1.65 96,240 104.80 18,285 39.49 6-AM 5 0.979 1.01 0.497 34.09 7.66 2.35 0.17 1878 28.04 * Significantly different GC MS and LC MS MS results for MOR and COD (p < 0.05). Differences in n value are due to differing presence of COD and MOR in the samples. For the regression analyses none of the compounds had intercepts significantly different than zero. 407

for GC MS, the acid hydrolysis was conducted at 110 C for 2 h at ambient pressure. Thus, there was likely a difference in the efficiency of the conjugate hydrolysis and differential hydrolysis of various methods that have been reported as well as different efficacy depending on which glucuronide is evaluated (3,6,19). All other parameters were tightly distributed and comparable between the two methods (Table XIV). All correlations between the two analyses were very good (> 0.979). Conclusions Based upon the data collected, GC MS and LC MS MS were comparable for the analyses of BZE, MOR, COD, and 6-AM. Both technologies provided accurate, precise, and specific results without interference from structural analogues. In previously confirmed specimens, results from both technologies were directly comparable. In manufactured materials in human urine, results were highly reproducible both within and between the two technologies. Evaluation of specific parameters of the LC MS MS data demonstrated that LC MS MS retention times were highly reproducible over several months and across multiple different samples. Product ion ratios were also very reproducible, and internal standard responses were consistently within 0.5 to 2 times the mean ISTD response from calibrators. Matrix effect was observed in the LC MS MS analyses of MOR, COD, and 6-AM but did not impact the overall results. These results did emphasize the need for matrix-matched controls and calibrators that are subjected to the same treatment as samples, as well as the use of deuterated internal standards as a means of controlling for matrix effects. This study demonstrated that LC MS MS technology can produce results at least as valid as GC MS for confirmatory testing of BZE, MOR, COD, and 6-AM at urine concentrations and conditions appropriate for federally regulated WPDT. It was demonstrated that LC MS MS can meet the NLCP s chromatographic criteria for retention time stability, peak shape, ion ratio, and calibration in both manufactured and archived samples. Similar studies are under way to evaluate and report findings for other routinely analyzed analytes for federally regulated WPDT, including amphetamines, phencyclidine, and 11- nor-9-carboxy- 9 -tetrahydrocannabinol in urine. References 1. 73 CFR 71858. November 25, 2008. Mandatory guidelines for Federal Workplace Drug Testing Programs. http://edocket. access.gpo.gov/2008/e8-26726.htm (accessed April 2009). 2. M.R. Moeller and T. Kraemer. Drug abuse monitoring in blood for control of driving under the influence of drugs. Ther. Drug Monit. 24(2): 210 221 (2002). 3. S.P. Vorce and D.J. Kuntz. Confirmation: traditional and newer analytical techniques. In Handbook of Workplace Drug Testing, 2nd ed., J.D. Ropero-Miller and B.A. Goldberger, Eds. AACC Press, Washington, D.C., 2009, pp 61 98. 4. X. Xu, M.M. Iba, and C.P. Weisel. Simultaneous and sensitive measurement of anabasine, nicotine and nicotine metabolites in human urine by liquid chromatography tandem mass spectrometry. Clin. Chem. 50: 2323 2330 (2004). 5. S.S. Johansen and H.M. Bhatia. Quantitative analysis of cocaine and its metabolites in whole blood and urine by high performance liquid chromatography coupled with tandem mass spectrometry. J. Chromatogr. B Analyt. Technol. Biomed. Life Sci. 852: 338 344 (2007). 6. S.V. Kala, S.E. Harris, T.D. Freijo, and S. Gerlich. Validation of analysis of amphetamines, opiates, PCP, cocaine and benzoylecgonine in oral fluid by liquid chromatography tandem mass spectrometry. J. Anal. Toxicol. 32: 605 611 (2008). 7. M. Wood, M. Laloup, F.M.M. Ramirez, and K.M. Jenkins, M.S. Young, J.G. Ramaekers, G. DeBoeck, and N. Samyn. Quantitative analysis of multiple illicit drugs in preserved oral fluid by solid phase extraction and liquid chromatography tandem mass spectrometry. Forensic Sci. Int. 150: 227 238 (2005). 8. C.J. Briscoe, M.R. Stiles, and D.S. Hage. System suitability in bioanalytical LC MS/MS. J. Pharm. Biomed. Anal. 44: 484 491 (2007). 9. S. Ma and M. Zhu. Recent advances in applications of liquid chromatography tandem mass spectrometry to the analysis of reactive drug metabolites. Chem. Biol. Interact. 179: 25 37 (2009). 10. M. Careri and A. Mangia. Validation and qualification: the fitness for purpose of mass spectrometry-based analytical methods and analytical systems. Anal. Bioanal. Chem. 386: 38 45 (2006). 11. E.J. Fox, S. Twigger, and K.R. Allen. Criteria for opiate identification using liquid chromatography linked to tandem mass spectrometry: problems in routine practice. Ann. Clin. Biochem. 46: 50 57 (2009). 12. B. Maralikova and W. Weinmann. Confirmatory analysis for drugs of abuse in plasma and urine by high-performance liquid chromatography tandem mass spectrometry with respect to criteria for compound identification. J. Chromatogr. B Analyt. Technol. Biomed. Life Sci. 811: 21 30 (2004). 13. M. Jemal, Z. Ouyang, and M.L. Powell. A strategy for a postmethod-validation use of incurred biological samples for establishing the acceptability of a liquid chromatography/tandem-mass spectrometric method for quantitation of drugs in biological samples. Rapid Commun. Mass Spectrom. 16: 1538 1547 (2002). 14. P.J. Taylor. Matrix effects: the Achilles heel of quantitative highperformance liquid chromatography electrospray-tandem mass spectrometry. Clin. Biochem. 38: 328 334 (2005). 15. A. Van Eeckhaut, K. Lanckmans, S. Sarre, I. Smolders, and Y. Michotte. Validation of bioanalytical LC MS/MS assays: evaluation of matrix effects. J. Chromatogr. B Analyt. Technol. Biomed. Life Sci. 877(23): 2198 2207 (2009). 16. E. Chambers, D.M. Wagrowski-Diehl, Z. Lu, and J.R. Mazzeo. Systematic and comprehensive strategy for reducing matrix effects in LC MS/MS analyses. J. Chromatogr. B 852: 22 34 (2007). 17. B.K. Matuszewski, M.L. Constanzer, and C.M. Chavez-Eng. Strategies for the assessment of matrix effect in quantitative bioanalytical methods based on HPLC MS/MS. Anal. Chem. 75: 3019 3030 (2003). 18. National Laboratory Certification Program, Manual for Urine Laboratories. Effective date: 1 January 2009. RTI International, Research Triangle Park, NC. 19. P. Wang, J.A. Stone, K.H. Chen, S.F. Gross, C.A. Haller, and A.H.B. Wu. Incomplete recovery of prescription opioids in urine using enzymatic hydrolysis of glucuronide metabolites. J. Anal. Toxicol. 30: 570 575 (2006). 408