Comparison of Isotope Dilution Mass Spectrometry Methods for the

Jul 17, 2003 - Mary B. Satterfield,* Lorna T. Sniegoski, Michael J. Welch, and Bryant C. Nelson. National Institute of Standards and Technology, Analy...
1 downloads 0 Views 109KB Size
Anal. Chem. 2003, 75, 4631-4638

Comparison of Isotope Dilution Mass Spectrometry Methods for the Determination of Total Homocysteine in Plasma and Serum Mary B. Satterfield,* Lorna T. Sniegoski, Michael J. Welch, and Bryant C. Nelson

National Institute of Standards and Technology, Analytical Chemistry Division, Gaithersburg, Maryland 20899-0001 Christine M. Pfeiffer

Centers for Disease Control and Prevention, Division of Laboratory Sciences, Atlanta, Georgia 30341-3724

Two independent methods have been critically evaluated and applied to the measurement of total homocysteine in serum and plasma: solid-phase anion extraction (SPAE) gas chromatography/mass spectrometry (GC/MS) and protein precipitation liquid chromatography/tandem mass spectrometry (LC/MS/MS). In addition, analysis of samples prepared by SPAE was accomplished by liquid chromatography/mass spectrometry (LC/MS) and LC/MS/MS. These methods have been used to determine total homocysteine levels in several existing serum-based Standard Reference Materials (SRMs) from the National Institute of Standards and Technology and in patient plasma samples provided by the Centers for Disease Control and Prevention. The precision of the homocysteine measurements in serum and plasma was critically evaluated, and method comparisons were carried out using Bland-Altman plots and bias analysis. On the basis of the excellent precision and close agreement of the mass spectrometric (MS) methods, the MS-based methods will be used for certification of a serum-based SRM for homocysteine and folates. Increased levels of homocysteine, an amino acid produced as a catabolic byproduct of methionine demethylation, have been associated with premature arteriosclerosis1,2 and neurological defects such as spina bifida.3,4 In addition, a recent report has linked high levels of homocysteine in the blood to the later development of Alzheimer’s disease.5 The clinical importance of homocysteine as a health marker has led to the development of numerous methods of homocysteine measurement, including * Corresponding author address: NIST, 100 Bureau Drive, Stop 8392, Gaithersburg, MD 20899-8392. Tel: 301-975-5364. Fax: 301-977-0685. E-mail: [email protected]. (1) Boushey, C. J.; Beresford, S. A.; Omenn, G. S.; Motulsky, A. G. JAMA 1995, 274, 1049-57. (2) Refsum, H.; Ueland, P. M.; Nygard, O.; Vollset, S. E. Annu. Rev. Med. 1998, 49, 31-62. (3) Nelen, W. L. Clin. Chem. Lab. Med. 2001, 39, 758-63. (4) Ueland, P. M.; Nygard, O.; Vollset, S. E.; Refsum, H. Lipids 2001, 36, S33S39. (5) Seshadri, S.; Beiser, A.; Selhub, J.; Jacques, P. F.; Rosenberg, I. H.; D’Agostino, R. B.; Wilson, P. W. F.; Wolf, P. A. N. Engl. J. Med. 2002, 346, 476-83. 10.1021/ac034207x CCC: $25.00 Published on Web 07/17/2003

© 2003 American Chemical Society

liquid chromatography with fluorescence detection (LC/FD)6-8 or electrochemical detection (LC/ED),9,10 fluorescence polarization immunoassay (FPIA),11,12 enzyme immunoassay (EIA),13,14 gas chromatography/mass spectrometry (GC/MS),15-17 liquid chromatography/tandem mass spectrometry (LC/MS/MS),18,19 and liquid chromatography/mass spectrometry (LC/MS).20 Ideally, each of these methods could be used interchangeably, with results determined by one method directly comparable to results determined by another method. To determine if such is the case, at least 10 method comparison studies have been carried out in which patient samples were tested by different laboratories using a variety of methods.21-34 These comparison studies enable investigation of interlaboratory and intermethod differences, with (6) Jacobsen, D. W.; Gatautis, V. J.; Green, R.; Robinson, K.; Savon, S. R.; Secic, M.; Ji, J.; Otto, J. M.; Taylor, L. M. Clin. Chem. 1994, 40, 873-81. (7) Pfeiffer, C. M.; Huff, D. L.; Gunter, E. W. Clin. Chem. 1999, 45, 290-93. (8) Ubbink, J. B.; Hayward Vermaak, W. J.; Bissbort, S. J. Chromatogr., A 1991, 565, 441-46. (9) D’Eramo, J. L.; Finkelstein, A. E.; Boccazzi, F. O.; Fridman, O. J. Chromatogr., B 1998, 720 (1-2), 205-10. (10) Houze, P.; Gamra, S.; Madelaine, I.; Bousquet, B.; Gourmel, B. J. Clin. Lab. Anal. 2001, 15, 144-53. (11) Marangon, K.; O’Byrne, D.; Devaraj, S.; Jialal, I. Am. J. Clin. Pathol. 1999, 112, 757-62. (12) Shipchandler, M. T.; Moore, E. G. Clin. Chem. 1995, 41, 991-94. (13) Frantzen, F.; Faaren, A. L.; Alfheim, I.; Nordhei, A. K. Clin. Chem. 1998, 44, 311-16. (14) O’Brion, S. D.; Kelleher, B. P.; McPartlin, J.; O’Gorman, P.; Browne, M.; White, B.; Smith, O. P. Blood Coagulation Fibrinolysis 2000, 11, 367-69. (15) Stabler, S. P.; Marcell, P. D.; Podell, E. R.; Allen, R. H. Anal. Biochem. 1987, 162, 185-96. (16) Stabler, S. P.; Lindenbaum, J.; Savage, D. G.; Allen, R. H. Blood 1993, 81, 3404-13. (17) MacCoss, M. J.; Fukagawa, N. K.; Matthews, D. E. Anal. Chem. 1999, 71, 4527-33. (18) Gempel, K.; Gerbitz, K.-D.; Casetta, B.; Bauer, M. F. Clin. Chem. 2000, 46, 122-23. (19) Magera, M. J.; Lacey, J. M.; Casetta, B.; Rinaldo, P. Clin. Chem. 1999, 45, 1517-22. (20) Nelson, B.; Pfeiffer, C. M.; Sniegoski, L. T.; Satterfield, M. B. Anal. Chem. 2003, 75, 775-84. (21) Barbe, F.; Abdelmouttaleb, I.; Chango, A.; Gerard, P.; Quilliot, D.; Klein, M.; Lambert, D.; Nicolas, J.-P.; Gueant, J.-L. Amino Acids 2001, 20, 43540. (22) Ceppa, F.; Drouillard, I.; Chianea, D.; Burnat, P.; Perrier, F.; Vaillant, C.; El Jahiri, Y. Ann. Biol. Clin (Paris) 1999, 57, 474-80. (23) Ducros, V.; Candito, M.; Causse, E.; Couderc, R.; Demuth, K.; Diop, M. E.; Drai, J.; Gerhardt, M. F.; Quillard, M.; Read, M. H.; Sauvant, M. P. Ann. Biol. Clin (Paris) 2002, 60, 421-28.

Analytical Chemistry, Vol. 75, No. 17, September 1, 2003 4631

Table 1. Levels of Total Homocysteine in NIST SRM Serum SRM serum

SPAE GC/MS ( U95a (µmol/L)

SPAE LC/MS ( U95 (µmol/L)

SPAE LC/MS/MS ( U95 (µmol/L)

Prot Pcptb LC/MS/MS ( U95 (µmol/L)

mean of four methods ( U95 (µmol/L)

909b Level Ic 909b Level II 1951a Level I 1951a Level II

13.27 ( 0.16 21.49 ( 0.12 9.71 ( 0.08 9.90 ( 0.08

13.44 ( 0.54 21.75 ( 0.24 10.02 ( 0.20 10.05 ( 0.20

13.44 ( 0.12 21.60 ( 0.22 9.72 ( 0.24 9.78 ( 0.24

13.53 ( 0.22 21.15 ( 0.28 9.84 ( 0.20 10.11 ( 0.20

13.42 ( 0.16 21.49 ( 0.46 9.82 ( 0.28 9.96 ( 0.30

a The U has a level of confidence of ∼95%. b Prot pcpt ) protein precipitation sample preparation. c Individual values are the average of eight 95 sample preparations of each type of SRM serum from two separate days (four sample preparations/day) with two injections per sample preparation (n ) 16). Two different vials of the same SRM serum level were used for each sample preparation, and two individual aliquots were tested from each vial.

the goals of improving methods that exhibit bias and improving interlaboratory and intermethod validity. Of the homocysteine method comparison studies published since 1997, all but two found that most methods for homocysteine determination were not interchangeable, with results from one method or lab producing values unacceptably different from results from another method or lab. In some cases, variation among labs using the same method was as high as the variation among labs using different methods.29 Most of the study authors indicated that the discrepancies between the results were due at least in part to the lack of certified reference materials. This is supported by the two studies in which close correlation was observed between different labs and different methods.28,31 In both of these cases, plasma calibrants were supplied to each lab. Although the plasma calibrants were not certified reference materials per se, the samples acted as quality control materials, since they were used by each lab and for each different method, allowing traceability to the same material. Use of these plasma calibrants for quantitation of the supplied patient samples resulted in homocysteine values with improved comparability among laboratories and among different methods. The clinical importance of homocysteine as a health marker, the need for measurement verification among multiple laboratories and methods, and the necessity of traceability to internationally recognized certified reference materials, as described by European Directive 98/79/EC on in vitro diagnostic medical devices, have been factors in the National Institute of Standards and Technology’s (NIST) decision to develop a reference material for homocys(24) Eliason, S. C.; Ritter, D.; Chung, H. D.; Creer, M. Clin. Chem. 1999, 45, 315-16. (25) Hanson, N. Q.; Eckfeldt, J. H.; Schwichtenberg, K.; Aras, O.; Tsai, M. Y. Clin. Chem. 2002, 48, 1539-45. (26) Moller, J.; Christensen, L.; Rasmussen, K. Scand. J. Clin. Lab. Invest. 1997, 57, 613-19. (27) Moller, J.; Rasmussen, K.; Christensen, L. Clin. Chem. 1999, 45, 1536-42. (28) Nexo, E.; Engbaek, F.; Ueland, P. M.; Westby, C.; O’Gorman, P.; Johnston, C.; Kase, B. F.; Guttormsen, A. B.; Alfheim, I.; McPartlin, J.; Smith, D.; Moller, J.; Rasmussen, K.; Clarke, R.; Scott, J. M.; Refsum, H. Clin. Chem. 2000, 46, 1150-56. (29) Pfeiffer, C. M.; Huff, D. L.; Smith, S. J.; Miller, D. T.; Gunter, E. W. Clin. Chem. 1999, 45, 1261-68. (30) Sigit, J. I.; Hages, M.; Brensing, K. A.; Frotscher, U.; Pietrzik, K.; von Bergmann, K.; Lutjohann, D. Clin. Chem. Lab. Med. 2001, 39, 681-90. (31) Tripodi, A.; Chantarangkul, V.; Lombardi, R.; Lecchi, A.; Mannucci, P. M.; Cattaneo, M. Thromb. Haemostasis 2001, 85, 291-95. (32) Ubbink, J. B.; Delport, R.; Riezler, R.; Vermaak, W. J. H. Clin. Chem. 1999, 45, 670-75. (33) Ubbink, J. B. Semin. Thromb. Hemostasis 2000, 26, 233-41. (34) Ueland, P. M.; Refsum, H.; Stabler, S. P.; Malinow, M. R.; Andersson, A.; Allen, R. H. Clin. Chem. 1993, 39, 1764-79.

4632 Analytical Chemistry, Vol. 75, No. 17, September 1, 2003

teine in human serum. A series of isotope dilution mass spectrometric methods for determination of homocysteine in plasma and serum with the goal of using these methods for certification of a homocysteine in human serum SRM have been developed, critically evaluated, and applied, including two independent methods, solid-phase anion extraction (SPAE)-GC/MS and protein precipitation LC/MS/MS. The use of two independent methods and the application of a homocysteine reference compound of known purity for calibration producing statistically similar results provide evidence for the accuracy of the methods. Until new SRMs for homocysteine in serum are available, the presently available SRMs tested for this study (SRM 909b, Human Serum, Levels I and II, and SRM 1951a, Lipids in Frozen Human Serum, Levels I and II) could be used for validation of homocysteine levels based on the values given in this paper (see Table 1). EXPERIMENTAL SECTION Reagents and Materials. DL-Homocystine was purchased from Fluka (Milwaukee, WI). Dithiothreitol (DTT) and an additional supply of DL-homocystine were purchased from Sigma (St. Louis, MO). The two sources of homocystine were produced by different manufacturers and asserted to be >99% (Sigma) and 99.2% (Fluka) pure by the suppliers. The internal standard, homocystine-d8 (DL-homocystine-3,3,3′,3′,4,4,4′,4′-d8), was purchased from CDN Isotopes (Quebec, Canada) and asserted to be 97.9 atom % deuterium at the labeled sites. Both sources of homocystine and the homocystine-d8 were assessed for purity by LC/MS and GC/MS with no impurities observed. N-Methyl-N(tert-butyldimethylsilyl)-trifluoroacetamide (MTBSTFA) was purchased from Pierce Chemical Company (Rockford, IL). All solvents used were HPLC grade from commercial suppliers. Chemicals and solvents were used without further purification. Patient plasma samples from the Centers for Disease Control and Prevention (Atlanta, GA) were evaluated, as were serum samples SRM 909b, Human Serum, Levels I and II, and SRM 1951a, Lipids in Frozen Human Serum, Levels I and II from NIST (Gaithersburg, MD). The serum samples were prepared from pooled samples from anonymous donors. Sample Preparation. Note: All steps were performed gravimetrically. Percentages refer to fractions by mass (g/g) unless mentioned otherwise. Standard solutions were prepared as described previously20 and involved reduction of the homocystine and homocystine-d8 to their respective monomers, homocysteine and homocysteine-d4 through the use of dithiothreitol. Calibration curves were prepared from aqueous standards, each of which was spiked with the working solution of homocysteine (Hcy) and

homocysteine-d4 (Hcy-d4) as required to produce ratios of Hcy/ Hcy-d4 bracketing the levels in the serum and plasma samples. For GC/MS analysis, the aqueous samples were taken through the SPAE process for cleanup and to provide a more suitable matrix for drying down under nitrogen gas prior to derivatization. The use of blanks for determination of homocysteine in serum and plasma was not possible because of the endogenous amounts of homocysteine present. To determine the linear dynamic range, two solutions were prepared gravimetrically, one containing ∼10 µmol/L Hcy-d4 in 1.5% DTT and the second containing ∼500 µmol/L Hcy and 10 µmol/L Hcy-d4 in 1.5% DTT. These two solutions were used to volumetrically prepare a series of serial dilutions with a range of 0.01-500 µmol/L Hcy, each with ∼10 µmol/L Hcy-d4. Serum and plasma samples for analysis by SPAE LC/MS and SPAE LC/MS/MS were prepared by measuring 200 µL of sample and adding enough of the working solution of Hcy-d4 ( ∼70 µmol/ L) to produce a final approximately 1:1 ratio of unlabeled to labeled homocysteine. The standards and samples were mixed with 400 µL of 1.5% DTT in 0.15 mol/L NaOH and then heated at 40 °C for 15 min to fully reduce the multiple forms of homocysteine. Samples prepared for analysis by SPAE GC/MS were prepared by measuring 125 µL of plasma or serum and spiking it with enough labeled internal standard to produce a final approximately 1:1 ratio of unlabeled to labeled homocysteine. The standards and samples were mixed with 100 µL of 1% DTT in 1 mol/L NaOH and diluted with distilled water to 1.1 mL, then heated at 40 °C for 1 h. To extract the homocysteine for analysis by GC/MS, LC/MS, and LC/MS/MS, solid-phase anionic exchange was used with BioRad Poly-prep prefilled chromatography columns (AG 1-X8 Resin, chloride form) based on the method by Stabler et al.16 The columns were prepared by single washes of 3 mL of methanol followed by 3 mL of water. After addition of the plasma or serum, the columns were washed again with 9 mL of water, then 3 mL of methanol, followed by elution of the homocysteine with 3 mL of 0.4 mol/L acetic acid in methanol. Samples prepared for analysis by LC/MS or LC/MS/MS were dried by vacuum centrifugation and reconstituted by the addition of 200 µL of 1.5% DTT. For measurement of the homocysteine by GC/MS, the eluate was dried under nitrogen and then converted to tert-butyldimethylsilyl esters of homocysteine by the addition of 25 µL of MTBSTFA and 50 µL of acetonitrile. Determination of homocysteine by protein precipitation with analysis by LC/MS/MS was based on the method developed by Magera et al.19 Aliquots of 100 µL of serum were spiked with the standard solution of homocysteine-d4 to a final concentration of ∼10 µmol/L. After 30 µL of 10% DTT was added, the solution was thoroughly mixed on a vortex mixer and allowed to sit at room temperature for 15 min. To precipitate the protein, 200 µL of 0.1% formic acid plus 0.05% trifluoroacetic acid in acetonitrile was added to the sample, and the solution was vortex-mixed again. The sample was centrifuged at 13000g for 1 min, and the supernatant was removed for analysis. Instrumentation. GC/MS. An HP 5971A GC/MS with an electron ionization (EI) source and a 30-m 5% mole fraction phenyl methyl polysiloxane (DB-5MS, J&W Scientific, Folsom, CA) column was used for analysis of the samples. The injector

temperature was 250 °C, the detector temperature was 280 °C, and the electron multiplier voltage was tune voltage plus 518. After injection (splitless) at 150 °C, the temperature was held for 0.5 min before ramping to 200 °C at 30 °C/min and was then ramped to 290 °C at 15 °C/min before being held at 290 °C for 2 min, for a total time of 10.2 min. The fragments of [M - 57]+ were monitored at m/z 420 and 424, corresponding to the loss of a butyl group from the unlabeled and labeled derivatized homocysteine. For confirmation, fragments of [M - 159]+, derivatized homocysteine minus a tert-butyldimethylsilyl group and a carboxylic acid group at m/z 318 and 322 were monitored. LC/MS and LC/MS/MS. Using a Waters 2790 LC, injections of the homocysteine samples were made onto a Supelcosil CN column, 5 µm (4.6 mm × 250 mm) (Supelco, Bellefonte, PA). Isocratic elution at 0.5 mL/min from 0 to 8 min in 96:4 0.1% formic acid in water/0.1% formic acid in methanol (volume fractions) was used. To clear the column, the mobile phase was shifted to 100% 0.1% formic acid in methanol between 8 and 9 min and then maintained at that concentration for an additional minute. From 10 to 11 min, the mobile phase was switched back to 96:4 0.1% formic acid in water/0.1% formic acid in methanol and then held at initial conditions from 11 to 15 min. The column temperature was 30 °C, and the samples were kept at 10 °C. The mass spectrometer used was a Micromass Quattro Ultima in the positive electrospray ionization mode. For detection by LC/ MS, instrument parameters were optimized for single-ion recording (SIR) of protonated homocysteine and protonated homocysteined4 at 136+ and 140+. When detection by LC/MS/MS was required, instrumental conditions were optimized for production of the dominant fragments of the protonated analytes. The multiple reaction monitoring mode (MRM) was used in which the transitions 136+ f 90+ and 140+ f 94+ were monitored, representing loss of a carboxylic acid group from protonated homocysteine and protonated homocysteine-d4, respectively. After completion of the experiments, linear regression analyses of aqueous calibration curves were used to calculate the amount of homocysteine in the samples on the basis of the integrated peak area ratios of homocysteine to homocysteine-d4. Statistical Evaluation of Data. To assess within-run and total precision, values were averaged, the standard deviations were determined, and the coefficients of variation (% CV) were calculated. For method comparison, Bland-Altman plots35,36 were created, with the mean values of the two methods under comparison plotted vs the difference of the two methods: method 1 - method 2. As a measure of systematic bias, the 95% confidence interval of the mean difference was calculated as the mean difference ( two times the standard error (standard deviation divided by the square root of the number of samples). A range of the mean difference ( two standard deviations was plotted for comparison to the actual values. Proportional bias was calculated as the ratio of the value determined by the first method to the value determined by the second method, the answer subtracted from 1 and then converted to a percent. Individual bias values were averaged to produce the mean proportional bias. (35) Bland, J. M.; Altman, D. G. Lancet 1986, 1, 307-10. (36) Dewitte, K.; Fierens, C.; Stockl, D.; Thienpont, L. M. Clin. Chem. 2002, 48, 799-802.

Analytical Chemistry, Vol. 75, No. 17, September 1, 2003

4633

Measurement uncertainty was calculated using a 4-level nested ANOVA model37 that includes variance among injections, among aliquots from the same vial, among aliquots of different vials of the same SRM, and among measurements on different days. RESULTS AND DISCUSSION Analysis of Currently Available SRMs by GC/MS, LC/MS, and LC/MS/MS. Analyses of total homocysteine in SRM 909b, Human Serum, Levels I and II, and SRM 1951a, Lipids in Fresh Frozen Human Serum, Levels I and II, were performed. SRM 909b, a lyophilized material, is certified by NIST for levels of clinical analytes, including cholesterol, glycerides, urea, and others, as shown in the SRM Certificate of Analysis.38 SRM 1951a, a frozen human serum, contains certified amounts of cholesterol and glycerides.39 These natural-matrix materials, in addition to purecompound SRMs, such as cholesterol, allow clinical laboratories to test their instruments, techniques, and calibrants against precisely and accurately determined analytes in the pure form or in complex matrixes. Calibration and testing of instrumental and extraction capabilities using SRM materials are designed to improve the reliability and accuracy of clinical tests. Preparation of a new SRM serum with certified levels of low, moderate, and high levels of homocysteine and folates is underway at NIST. To choose an appropriate reference method or methods, two sample preparation procedures and three mass spectrometrybased methods of determination were investigated. Most methods, as do these, measure total homocysteine, a composite of the forms of homocysteine found in circulation: protein-bound homocysteine, free homocysteine, homocysteine oxidized to the dimer homocystine, and other mixed disulfides, such as cysteinehomocysteine. For all three methods, isotope dilution with deuterated homocysteine, homocysteine-d4, was employed. Sample preparation by SPAE based on the procedures developed by Stabler et al.16 was chosen because of the reproducibility of results and its versatility for detection by GC/MS, LC/MS, and LC/MS/ MS. As an additional test of the validity of the SPAE technique, sample cleanup by protein precipitation was carried out in the manner of Magera et al.19 with determination of homocysteine possible only by LC/MS/MS. Although measurement of homocysteine after protein precipitation produces samples with too many additional analytes for accurate determination by LC/MS and GC/MS, the excellent selectivity of LC/MS/MS permits accurate homocysteine determination. Comparison of the results by two independent methods provides further support of their accuracy. The homocysteine determinations after SPAE are dependent only on the instrument used and the ions monitored for analysis because the same sample preparation was used for all measurements. Derivatization of the homocysteine with MTBSTFA yields volatile products that are amenable to separation and detection by GC/MS: the tert-butyldimethylsilyl esters of homocysteine and (37) MRC Biostatistics Unit. WinBUGS Version 1.3. Bayesian Inference Using Gibbs Sampling. http://www.mrc-bsu.cam.ac.uk/bugs (accessed December 2002). (38) National Institute of Standards and Technology Certificate of Analysis. http:// patapsco.nist.gov/srmcatalog/certificates/view_cert2gif.cfm?certificate) 909b (accessed December 2002). (39) National Institute of Standards and Technology Certificate of Analysis. http:// patapsco.nist.gov/srmcatalog/common/view_cert.cfm?srm)1951a (accessed December 2002)

4634

Analytical Chemistry, Vol. 75, No. 17, September 1, 2003

homocysteine-d4. Using electron ionization, the derivatized homocysteine fragments, (M - butyl)+ were monitored at 420+ and 424+ (See Figure 1A). As further validation of the method, fragments of (M - tert-butyldimethylsilyl - carboxylic acid)+ were monitored. Close correlation between the ratios of 420+:424+ and 318+:322+ was noted, indicative of the lack of chemical interferences. For determination of total homocysteine by LC/MS, the protonated analytes at 136+ and 140+ were monitored, as shown in Figure 1B. Because LC/MS can be subject to interference by analytes with similar molecular masses and retention times, the use of the SPAE method for sample cleanup greatly reduced the potentially interfering molecules. Furthermore, precise and accurate determination of total homocysteine levels was possible via selected ion monitoring LC/MS. The third mass spectrometric method employed, LC/MS/MS, was able to quantify homocysteine levels after sample cleanup by SPAE (see Figure 1C) or protein precipitation (see Figure 1D). Detection of analytes by LC/MS/MS with a triple quadrupole mass spectrometer is an extremely selective method based on passing only ions of the masses of the target protonated molecular ions through the first quadrupole, collisionally dissociating these ions in the middle quadrupole, the “collision cell”, and monitoring fragments specific to the precursor in the third quadrupole. The transition monitored for detection by LC/MS/MS was loss of the carboxylic acid group from the protonated homocysteine or protonated homocysteine-d4, a loss of 46 u. The possibility of interferences is low since the interfering compound must have the same retention time and same molecular mass as well as dissociation to the same fragments under the same conditions as homocysteine or the internal standard. The linear dynamic ranges for total homocysteine as determined by GC/MS, LC/MS, and LC/MS/MS in this study were 0.15-56 µmol/L, 0.06-30 µmol/L, and 0.014-467 µmol/L, respectively. Since the normal range of total homocysteine is 5-15 µmol/L, the ranges of all three methods cover expected levels. Higher levels of homocysteine, such as those greater than the 30 µmol/L upper limit of the LC/MS method, can be measured by diluting the sample appropriately. SPAE was employed for sample cleanup of the SRM serum for all three MS techniques, and sample cleanup by protein precipitation was used for LC/MS/MS analysis of the SRM serum. Results of the determination of total homocysteine in SRM serum by SPAE GC/MS, SPAE LC/MS, and SPAE LC/MS/MS, and by protein precipitation LC/MS/MS are shown in Table 1. Individual values are the average of eight sample preparations of each type of SRM serum from two separate days (four sample preparations/ day) with two injections per sample preparation. Two different vials of the same SRM serum level were used for each sample preparation, and two individual aliquots were tested from each vial. The mean value and approximate 95% uncertainty, U95, for each SRM using each method were calculated,37 as were the mean and U95 of the SRM as a composite of the four methods. The U95 takes into account variances due to dual injections of the same preparation, preparations from the same vial, heterogeneity between different vials of the same SRM, and systematic differences between days. The greatest uncertainty in all methods was

Figure 1. (A) Selected ion chromatograms of derivatized homocysteine and homocysteine-d4 exhibiting loss of a butyl group. The homocysteine was extracted from SRM 909b, Human Serum, Level 1 by SPAE and analyzed by GC/MS. (B) Selected ion chromatograms of protonated homocysteine and protonated homocysteine-d4 extracted from SRM 909b, Human Serum, Level 1 by SPAE and analyzed by LC/MS. (C) Multiple reaction monitoring chromatograms of the transitions of protonated homocysteine and protonated homocysteine-d4 exhibiting loss of a carboxylic acid group. The homocysteine was extracted from SRM 909b, Human Serum, Level 1 by SPAE and analyzed by LC/MS/MS. (D) Multiple reaction monitoring chromatograms of the transitions of protonated homocysteine and protonated homocysteine-d4 exhibiting loss of a carboxylic acid group. The homocysteine was extracted from SRM 909b, Human Serum, Level 1 by protein precipitation and analyzed by LC/MS/MS. Table 2. Within-Run and Total Variation of SRM Serum SRM serum 909b Level 1 909b Level 2 1951a Level 1 1951a Level 2

within-run variationb total variationc within-run variation total variation within-run variation total variation within-run variation total variation

SPAE GC/MS % CV

SPAE LC/MS % CV

SPAE LC/MS/MS % CV

Prot Pcpta LC/MS/MS % CV

1.2 2.0 0.60 0.70 0.29 0.40 0.62 0.53

0.55 2.8 1.2 2.0 1.4 1.8 0.83 2.1

0.35 1.1 0.41 1.5 0.69 1.7 0.91 1.6

3.5 2.8 1.9 2.5 1.6 3.0 1.7 2.7

a Prot Pcpt ) protein precipitation sample preparation. b Within-run variation is a comparison of four sample preparations of each type of SRM serum from 1 day with two injections per sample preparation (n ) 8). c Total variation is a comparison of eight sample preparations of each type of SRM serum from two separate days with two injections per sample preparation (n ) 16).

the variability between injections. None of the other factors was a chemically significant source of variability. The quantitative precision of the total homocysteine values was explored by investigating within-run variation (see Table 2), in which eight values determined from a single day of four sample preparations from two vials of an SRM serum with two injections per sample preparation using the same method were compared. Total variation was determined by comparing 16 possible homocysteine values obtained by two separate-day sample preparations.

Although specific standards for analytical precision in measurements of homocysteine have not been set, standards based on the biological variation of total homocysteine levels can be used, as was done by Pfeiffer et al. in the comparison of total homocysteine measurements from 14 laboratories.29 For optimum performance analytical imprecision as determined by the percent coefficient of variance (% CV) should be