Anal. Chem. 2004, 76, 1008-1014
Development and Evaluation of a Candidate Reference Method for the Determination of Total Cortisol in Human Serum Using Isotope Dilution Liquid Chromatography/Mass Spectrometry and Liquid Chromatography/Tandem Mass Spectrometry Susan S.-C. Tai* and Michael J. Welch
Analytical Chemistry Division, National Institute of Standards and Technology, Gaithersburg, Maryland 20899-8392
Cortisol is an important diagnostic marker for the production of steroid hormones, and accurate measurements of serum cortisol are necessary for proper diagnosis of adrenal function. A candidate reference method involving isotope dilution coupled with liquid chromatography/ mass spectrometry (LC/MS) and liquid chromatography/ tandem mass spectrometry (LC/MS/MS) has been developed and critically evaluated. An isotopically labeled internal standard, cortisol-d3, was added to serum, followed by equilibration and solid-phase and ethyl acetate extractions to prepare samples for liquid chromatography/mass spectrometry electrospray ionization (LC/MSESI) and liquid chromatography/tandem mass spectrometry electrospray ionization (LC/MS/MS-ESI) analyses. (M + H)+ ions at m/z 363 and 366 for cortisol and its labeled internal standard were monitored for LC/MS. The transitions of (M + H)+ f [(M + H)+ - 2H2O] at m/z 363 f 327 and 366 f 330 were monitored for LC/MS/MS. The accuracy of the measurement was evaluated by a comparison of results of this candidate reference method on lyophilized human serum reference materials for cortisol [Certified Reference Materials 192 and 193] with the certified values determined by gas chromatography/mass spectrometry reference methods and by a recovery study for the added cortisol. The results of this method for total cortisol agreed with the certified values within 1.1%. The recovery of the added cortisol ranged from 99.8% to 101.0%. This method was applied to the determination of cortisol in samples of frozen serum pools. Excellent precision was obtained with within-set CVs of 0.3%-1.5% and between-set CVs of 0.04%-0.4% for both LC/MS and LC/MS/MS analyses. The correlation coefficients of all linear regression lines ranged from 0.998 to 1.000. The detection limits (at a signal-to-noise ratio of ∼3-5) were 10 and 15 pg for LC/MS and LC/MS/MS, respectively. This method, which demonstrates good accuracy and precision, and is free from interferences from structural analogues, qualifies as a candidate reference method and can be used as an alternative reference method to provide * To whom correspondence should be addressed. E-mail:
[email protected].
1008 Analytical Chemistry, Vol. 76, No. 4, February 15, 2004
an accuracy base to which the routine methods can be compared. Cortisol is a steroid hormone secreted by the adrenal cortex and is largely bound to protein in circulation. Cortisol is primarily involved in carbohydrate metabolism, but it has several other functions including inhibition of protein synthesis in peripheral tissues and promotion of antiinflammatory and immunosuppressive actions. Cortisol is an important diagnostic marker for the production of steroid hormones, and accurate measurements of serum cortisol are necessary for proper diagnosis of adrenal function. Many routine clinical methods for serum cortisol, primarily based on immunoassays, are positively biased as compared to mass spectrometric methods.1-4 There is a need for a critically evaluated reference method for cortisol to provide an accuracy base to which routine methods can be traceable. Various isotope dilution (ID) gas chromatography/mass spectrometry (GC/MS) definitive and reference methods with CVs of ∼2% or less have been published for the determination of cortisol in serum.5-8 Two lyophilized human serum reference materials for cortisol,9,10 Certified Reference Materials (CRMs) 192 and 193 [from the Community Bureau of Reference (BCR)] and a reference panel of fresh frozen single-donation sera for cortisol11 were certified using GC/MS reference methods for traceability assessment. (1) De Brabandere, V. I.; Veronique, I.; Thienpont, L. M.; Sto¨ckl, D.; De Leenheer, A. P. Clin. Chem. 1995, 41, 1781-1783. (2) Gosling, J. P.; Middle, J.; Siekmann, L.; Read, G. Scand. J. Clin. Lab. Invest. 1993, 53 (Suppl 216), 3-41. (3) Gaskell, S. J.; Collins, C. J.; Thorne, G. C.; Groom, G. V. Clin. Chem. 1983, 29, 862-867. (4) Bjo ¨rkhem, I.; Bergman, A.; Falk, O.; Kallner, A.; Lantto, O.; Svensson, L.; A° kerlo ¨f, E.; Blomstrand, R. Clin. Chem. 1981, 27, 733-735. (5) Thienpont, L. M.; Veronique, I.; De Brabandere, V. I.; Sto¨ckl, D.; De Leenheer, A. P. Anal. Biochem. 1996, 234, 204-209. (6) Siekmann, L.; Breur, H. J. Clin. Chem. Clin. Biochem. 1982, 20, 883-892. (7) Patterson, D. G.; Patterson, M. B.; Culbreth, P. H.; Fast, D. M.; Holler, J. S.; Sampson, E. J.; Bayse, D. D. Clin. Chem. 1984, 30, 619-626. (8) Jonckheere, J. A.; De Leenheer, A. P. Biomed. Mass Spec. 1983, 10, 197202. (9) Lawson, A. M.; Calam, D. H.; Colinet, E. S. EUR 9661 EN, 1985. (10) Thienpont, L. M.; Siekmann, L.; Lawson, A. M.; Colinet, E. S.; De Leenheer, A. P. Clin. Chem. 1991, 37, 540-546. 10.1021/ac034966f Not subject to U.S. Copyright. Publ. 2004 Am. Chem. Soc.
Published on Web 01/16/2004
Recent developments of highly sensitive liquid chromatography (LC)/MS instrumentation with robust interfaces such as electrospray ionization (ESI) or atmospheric pressure chemical ionization (APCI) make it possible to develop highly reproducible, selective, and sensitive LC/MS methods for low-concentration analytes.12,13 LC/MS methods offer a powerful tool for the determination of nonvolatile and thermally labile compounds without the need for derivatization (in contrast to the usual situation with GC/MS for many clinical analytes). Various ID LC/MS or LC/MS/MS methods using ESI or APCI have been reported for the determination of cortisol in urine14-16 and in placental perfusate.17 An LC/ MS/MS-ESI method for serum cortisol has been published.18 However, the CVs of this method ranged from 3.2% to 5.0%, and the results deviated from the target concentrations of external quality control materials (determined by a reference method) by +2.3% to -8.5%. This method is a good routine clinical method due to its short chromatographic run time, high throughput, and higher specificity than the routine immunoassays, but does not have sufficient precision and accuracy for use as a reference method. To the best of our knowledge, no LC/MS or LC/MS/ MS reference methods for serum cortisol with the precision and accuracy comparable to GC/MS reference methods has been published. In this work, we have developed a candidate reference method for total cortisol in serum involving ID coupled with LC/ MS-ESI and LC/MS/MS-ESI. The accuracy of the measurement was evaluated by a comparison of results of this candidate reference method on CRMs 192 and 193 with the certified values determined by GC/MS reference methods and by a recovery study for the added cortisol. This candidate reference method was found to be free from interferences by testing the structural analogues of cortisol under the LC conditions for cortisol measurement. This LC/MS-based method was applied to the determination of total cortisol in samples of frozen serum pools. The results of LC/MS and LC/MS/MS measurements for serum cortisol were compared. This candidate reference method was compared with routine clinical methods based on immunoassays from College of American Pathologists (CAP) surveys. EXPERIMENTAL SECTION Materials. SRM 921 cortisol (hydrocortisone), obtained from the Standard Reference Materials Program at NIST, was used as a cortisol reference compound for this work. SRM 921 is certified as 98.9% pure with an estimated uncertainty of 0.2%. Appropriate correction was made for impurities. A stable deuterium-labeled internal standard, cortisol-d3 (9,12,12-d3), with an isotopic purity of 95% was obtained from Cambridge Isotope Labs (Andover, MA). 3H-Labeled cortisol was obtained from DuPont NEN (Boston, MA). C18 Sep-Pak solid-phase extraction cartridges (20 mL, 500 mg) (11) Thienpont, L. M.; De Leenheer, A. P.; Siekmann, L.; Kristiansen, N.; Linsinger, T.; Schimmel, H.; Pauwels, J. EUR 19007 EN, 1999. (12) Tai, S. S.; Sniegoski, L. T.; Welch, M. J. Clin. Chem. 2002, 48, 637-642. (13) De Brabandere, V. I.; Hou, P.; Sto ¨ckl, D.; Thienpont, L. M.; De Leenheer, A. P. Rapid Commun. Mass Spectrom. 1998, 12, 1099-1103. (14) Taylor, R. L.; Machacek, D.; Singh, R. J. Clin. Chem. 2002, 48, 1511-1519. (15) Nassar, A. F.; Varshney, N.; Getek, T.; Cheng, L. J. Chromatogr. Sci. 2001, 39, 59-64. (16) Ohno, M.; Yamaguchi, I.; Saiki, K.; Yamamoto, I.; Azuma, J. J. Chromatogr., B 2000, 746, 95-101. (17) Dodds, H. M.; Taylor, P. J.; Cannell, G. R.; Pond, S. M. Anal. Biochem. 1997, 247, 342-347. (18) Vogeser, M.; Briegel, J.; Jacob, K. Clin. Chem. Lab. Med. 2001, 39, 944947.
were obtained from Waters (Milford, MA). A Zorbax Eclipse XDBC18 column [15 cm × 2.1 mm (i.d.); 5-µm particle diameter] was obtained from Agilent Technologies (Wilmington, DE). Solvents used for LC/MS and LC/MS/MS measurements were HPLC grade, and all other chemicals were reagent grade. Samples of frozen serum pools were provided by the CAP from their 2001 survey. CRM 192, lyophilized human serum reference material for cortisol (unspiked), and CRM 193, lyophilized human serum reference material for cortisol (spiked), manufactured by European Commission Institute for Reference Materials and Measurements (formerly BCR), were obtained from Resource Technology Corp. (Laramie, WY). The following compounds were used for interference testing: 4-Pregnene-17R,21-diol-3,11,20-trione (cortisone), 1,4pregnadiene-11β,17R,21-triol-3,20-dione (prednisolone), 4-pregnene17R,20R,21-triol-3,11-dione (20R-dihydrocortisone), 17R,20β,21-trihydroxypregn-4-ene-3,11-dione (20β-dihydrocortisone), 4-pregnene11β,17R,20β,21-tetrole-3-one (20β-dihydrocortisol), 11β,17R,21trihydroxy-5R-pregnane-3,20-dione (5R-dihydrocortisol), 11β,17R,21-trihydroxy-5β-pregnane-3,20-dione (5β-dihydrocortisol), 3R,17R,21trihydroxy-5β-pregnane-11,20-dione (tetrahydrocortisone), 4-pregnene-11β,21-diol-3,18,20-trione (aldosterone), and 4-pregnene-11β,18,21-triol-3,20-dione (18-hydroxycorticosterone) were obtained from Sigma (St. Louis, MO). 4-Pregnen-11β,17,20R,21-tetrol-3-one (20R-dihydrocortisol), 5R-pregnane-17,21-diol-3,11,20-trione (5Rdihydrocortisone), 5β-pregnane-17,21-diol-3,11,20-trione (5β-dihydrocortisone), and 4-pregnene-6β,17,21-triol-3,20-dione (6β-hydroxycortexolone) were obtained from Steraloids (Newport, RI). Preparation of Calibrators. Two independently weighed stock solutions of cortisol were prepared. One milligram of the cortisol reference compound for each stock solution was accurately weighed on an analytical balance (Mettler ME22 with a readability of 1 µg) and dissolved in anhydrous ethanol in a 500-mL volumetric flask. The balance was calibrated and demonstrated to be accurate to 1 µg. The concentrations of cortisol in the stock solutions were ∼2 mg/L. A solution of isotopically labeled internal standard, cortisol-d3, at a concentration of ∼2 mg/L was prepared in the same way as the unlabeled cortisol. The concentrations of the two stock solutions of cortisol were cross-checked against each other by LC/MS and were in good agreement (within 0.5% of each other). Cortisol-d3 was added to two aliquots of each stock solution of cortisol, yielding four calibrators with mass ratios of unlabeled to labeled compound ranging from 0.69 to 1.32. The mixtures were dried and reconstituted with a solvent consisting of 1 mL/L acetic acid in water-methanol (54:46 by volume) to a cortisol concentration of ∼0.25 mg/L for LC/MS and LC/MS/MS analyses. The stock solution was aliquoted with a Rainin EDP-2 motorized pipet. The volumes were calibrated by weighing. Sample Preparation. Samples of frozen serum pools were prepared in three different sets (each set on a different day), each set consisting of two vials each of the three concentrations (83, 166, and 248 µg/L for concentrations 1, 2, and 3, respectively). Duplicate 2.0-mL aliquots were taken from each vial for sample workups. An appropriate amount of cortisol-d3 was added to each aliquot to give an ∼1:1 ratio of analyte to internal standard. Each sample was acidified to pH 1.5 with 6 mL of 0.5 mol/L phosphoric acid and equilibrated at room temperature (22 °C) for 1 h. After equilibration, cortisol was isolated from the serum matrix using Analytical Chemistry, Vol. 76, No. 4, February 15, 2004
1009
C18 Sep-Pak solid-phase extraction cartridges.19-21 Each sample was loaded at a rate of 3-4 mL/min onto a cartridge previously conditioned by wetting with 5 mL of methanol, followed by 5 mL of water. The cartridge was then washed with 12 mL of water, followed by 5 mL of water-acetonitrile (80:20 by volume). The cortisol was eluted from the cartridge with 2 mL of methanol. The eluate was dried under nitrogen at 40 °C and reconstituted with 0.5 mL of water. The cortisol was then extracted from the water layer with 1.0 mL of ethyl acetate. Each tube was thoroughly mixed using a vortex mixer. The layers were allowed to separate, and the upper ethyl acetate layer was transferred to another tube. The aqueous layer was extracted once more with another 1 mL of ethyl acetate. The combined ethyl acetate extract was dried under nitrogen at 40 °C and reconstituted with a solvent consisting of 1 mL/L acetic acid in water-methanol (54:46 by volume) to a cortisol concentration of ∼0.25 mg/L for LC/MS and LC/MS/ MS analyses. 3H-Labeled cortisol was used as a tracer to evaluate the recovery of cortisol from serum using this extraction method. Equilibration. Vials of frozen serum samples were combined, and six 2.0-mL aliquots were taken for the equilibration study. A given amount of cortisol-d3 was added to each aliquot, and the aliquots were equilibrated at room temperature for various times (0.5, 1, 1.5, 2, 3, 4 h). The samples were processed as described above for LC/MS measurement. Recovery of the Added Cortisol. Vials of frozen serum samples were combined, and eight 2.0-mL aliquots were taken for a study of the accuracy of the method. A known amount of unlabeled cortisol was added to six of the eight aliquots at three concentrations, two each with 58.4, 97.2, and 136.1 µg/L cortisol. No cortisol was added to the other two aliquots. A given amount of cortisol-d3 was added to each aliquot, and the aliquots were then processed as described above for LC/MS measurement. LC/MS-ESI Analysis. LC/MS-ESI analysis was performed on a Hewlett-Packard 1100 Series LC/MSD from Agilent Technologies with an ESI interface. An autosampler was used. Aliquots (25 µL) of calibrators or sample extracts (∼6 ng of cortisol) were separated by LC on a Zorbax Eclipse XDB-C18 column at room temperature with a gradient mobile phase consisting of 1 mL/L acetic acid in water-methanol. The gradient was initially set at water-methanol (54:46 by volume) for 24 min, ramped to 100% methanol at 27 min, and held for 13 min. The flow rate was 0.25 mL/min. The mass spectrometer was operated in the positive ion mode. (M + H)+ ions at m/z 363 and 366 were monitored for cortisol and cortisol-d3, respectively. The nitrogen drying gas temperature was set at 350 °C and the flow at 12 L/min. The nebulizer pressure was set at 172 kPa (25 psi), Vcap at 3000 V, and fragmentor at 80 V. Interference Testing. The metabolites of cortisol17,22-24 and other corticosteroids2,5 having the molecular mass of 360, 362, or 364 Da were tested for interference using the LC/MS-ESI method described above for cortisol measurement. (M + H)+ ions at m/z 361, 363, and 365 were monitored. (19) Hsu, F. F.; Wang, L.; Bier, D. M. Anal. Biochem. 1994, 216, 401-405. (20) Kasuya, Y.; Furuta, T.; Hirota, N. Biomed. Environ. Mass Spectrom. 1988, 16, 309-311. (21) Cannell, G. R.; Galligan, J. P.; Mortimer, R. H.; Thomas, M. J. Clin. Chim. Acta 1982, 122, 419-423.
1010
Analytical Chemistry, Vol. 76, No. 4, February 15, 2004
LC/MS/MS-ESI Analysis. LC/MS/MS-ESI analysis was performed on a Micromass Quattro Ultima equipped with a Waters 2795 LC. Aliquots (40 µL) of calibrators or sample extracts (∼10 ng of cortisol) were separated by LC on the same column with a gradient mobile phase consisting of 1 mL/L acetic acid in watermethanol. The gradient was initially set at water-methanol (54: 46 by volume) for 20 min, ramped to 100% methanol at 21 min, and held for 13 min. The flow rate was 0.25 mL/min. The column temperature was set at 30 °C. ESI in the positive ion and multiple reaction monitoring (MRM) modes were used for LC/MS/MS. The transitions of (M + H)+ f [(M + H)+ - 2H2O] at m/z 363 f 327 and 366 f 330 were monitored for cortisol and cortisol-d3, respectively. The collision gas was argon at a collision cell pressure of 0.25 Pa (2.5 × 10-3 mbar), and the collision energy was 14 eV. The dwell times were 0.3 s for MRM. Measurement Protocol. The following measurement protocol was used for LC/MS and LC/MS/MS analyses. For each set of samples, a single analysis of each of the four calibrators was run first. Subsequently, duplicate analyses of each sample were run. Finally, the four calibrators were run again in reverse order. By combining the data of calibrators run before and after the samples, a composite linear regression was calculated, which was used to convert the measured intensity ratios of analyte to mass ratios. The mass ratios were then used along with the amounts of the internal standard added to calculate analyte concentrations. RESULTS AND DISCUSSION Evaluation of Critical Parameters. For a method to be considered as a reference method, it must be thoroughly tested for sources of bias. The following sections discuss the evaluation of critical parameters that potentially could bias the results. (1) Equilibration. Complete equilibration of the cortisol in serum with the added internal standard is necessary for accurate measurement. The time required for cortisol to be liberated from its protein binding and then equilibrated with the internal standard, cortisol-d3, was investigated. It was found that the equilibration was complete within 30 min at room temperature, and the ratio of cortisol to cortisol-d3 was unchanged for up to 4 h. One hour was chosen as the time for equilibration. (2) Extraction. The combination of solid-phase and ethyl acetate extractions produced a clean extract for LC/MS and LC/ MS/MS analyses. 3H-Labeled cortisol was used as a tracer to evaluate the extraction procedures. The overall recovery of cortisol from the serum with this extraction method was determined to be ∼75%. With complete equilibration between the labeled and unlabeled forms, as discussed above, the absolute recovery is not critical because it is the ratio of unlabeled to labeled cortisol that is measured. (3) Recovery of the Added Cortisol. The accuracy of measurement for cortisol in serum was determined through a recovery study for the added cortisol. The results are shown in Table 1. The mean concentration of endogenous cortisol in the pooled serum was 63.5 µg/L. The amounts of cortisol recovered and added were in good agreement for all three concentrations with the mean recoveries of 100.9%, 99.8%, and 101.0% for 58.4, 97.2, and 136.1 µg/L added cortisol, respectively. (22) Abel, S. M.; Maggs, J. L.; Back, D. J.; Park, B. K. J. Steroid Biochem. Mol. Biol. 1992, 43, 713-719. (23) Jenkins, J. S. J. Endocrinol. 1966, 34, 51-56. (24) Meigs, R, A.; Engel, L. L. Endocrinology 1961, 69, 152-62.
Figure 1. LC/MS chromatogram of a mixture of corticosteroid standards: 1, 6β-hydroxycortexolone; 2, 20R-dihydrocortisone; 3, aldosterone; 4, 20β-dihydrocortisone; 5, 20R-dihydrocortisol; 6, 18-hydroxycorticosterone; 7, cortisone; 8, 20β-dihydrocortisol; 9, cortisol; 10, prednisolone; 11, 5R-dihydrocortisone; 12, 5R-dihydrocortisol. Table 1. Recovery of Added Cortisol added, µg/L 0 58.4 97.2 136.1 a
expected, µg/L naa 121.9 160.7 199.6
dectected, µg/L 63.5 123.0 160.3 201.5
recovery, % naa 100.9 99.8 101.0
CV, % (n ) 4) 0.7 0.6 0.3 0.6
na, not applicable.
(4) Interference Testing. If a structural analogue of cortisol were to coelute with cortisol or cortisol-d3, it could bias the results. The metabolites of cortisol and other corticosteroids having the molecular mass of 360, 362, or 364 were tested to determine whether they could interfere with the measurement of cortisol. All the compounds tested were completely resolved from cortisol except for prednisolone. The retention times for these compounds under the LC conditions for cortisol measurement are listed in Table 2, and an LC/MS chromatogram of 12 of these compounds is shown in Figure 1. Compounds with longer retention times are not included in Figure 1. Prednisolone is a synthetic corticosteroid with a molecular mass of 360. Because of the natural isotope effects for prednisolone, the intensity of the ion at m/z 363, which corresponds with the ion monitored for cortisol, will be 3%-4% of the (M + H)+ ion at m/z 361. If prednisolone and cortisol were to coelute and significant amounts of prednisolone were present in the sample extracts, the results for cortisol could be biased. As shown in Figure 1, prednisolone was partially separated from cortisol using the mobile phase consisting of 1 mL/L acetic acid in water-methanol (54:46 by volume). We found no prednisolone present in the sample extracts from monitoring m/z 361; therefore, we used the mobile phase of water-methanol (54:46 by volume) specified in the Experimental Section to maintain a reasonable retention time. If significant amounts of prednisolone were to be
Table 2. Ion Monitored and Retention Time for Each Compound of Interest compound
ion monitored, m/z
retention time, min
6β-hydroxycortexolone 20R-dihydrocortisone aldosterone 20β-dihydrocortisone 20R-dihydrocortisol 18-hydroxycorticosterone cortisone 20β-dihydrocortisol cortisol prednisolone 5R-dihydrocortisone 5R-dihydrocortisol 5β-dihydrocortisone 5β-dihydrocortisol tetrahydrocortisone
363 363 361 363 365 363 361 365 363 361 363 365 363 365 365
4.7 9.7 10.2 11.4 11.7 12.1 13.2 15.4 16.8 17.3 19.2 23.1 >25 >25 >25
present in the sample extracts, modified LC conditions with less polar mobile phase and higher column temperature (30 °C) could be used to achieve better separation. (5) Comparison to Certified Values of CRMs 192 and 193. As another accuracy assessment for the candidate reference method, samples of CRMs 192 and 193 were analyzed as described in the Experimental Section and results were compared to the certified values determined using GC/MS reference methods.9,10 The mean concentrations of cortisol in CRMs 192 and 193 measured by the candidate reference method were 97.7 (0.8% CV, n ) 6), and 273.9 µg/L (0.2% CV, n ) 4) for CRM 192 and CRM 193, respectively. These results agreed with the certified values (98.8 ( 2 µg/L for CRM 192 and 277 ( 5 µg/L for CRM 193) within 1.1%. Measurement of Frozen Serum Materials. The LC/MS-ESI method for the determination of cortisol was applied to samples Analytical Chemistry, Vol. 76, No. 4, February 15, 2004
1011
Figure 2. Ion chromatograms by LC/MS-ESI for cortisol and cortisol-d3 from a serum sample, concentration 1. Table 3. LC/MS-ESI Measurements of Cortisol in Serum (µg/L) conc
set
mean
SDa
CV, %
1 1 1 2 2 2 3 3 3
1 2 3 1 2 3 1 2 3
82.8 83.6 83.3 165.2 165.7 165.5 247.4 247.9 249.1
0.42 0.32 0.27 0.57 0.70 0.52 1.46 1.33 1.76
0.5 0.4 0.3 0.3 0.4 0.3 0.6 0.5 0.7
overall mean
overall SDb
overall CV, %
83.2
0.26
0.3
165.5
0.16
0.1
248.1
0.51
0.2
a Standard deviation (SD) of a single measurement within a set. b SD of the mean for that level.
of three frozen serum pools. Samples were prepared and analyzed in three different sets, each set consisting of three concentrations of cortisol. The results are shown in Table 3. Excellent reproducibility was obtained for all three concentrations with within-set CVs ranging from 0.3% to 0.7%, and between-set CVs ranging from 0.1% to 0.3%. A linear regression line was generated for each set of samples. Excellent linearity was also obtained with the correlation coefficients (R) of all linear regression lines (measured intensity ratios vs mass ratios) ranging from 0.9996 to 0.9998. The same sample extracts were also analyzed by LC/MS/MS for the added specificity provided by tandem mass spectrometry. The transitions of (M + H)+ f [(M + H)+ - 2H2O] at m/z 363 1012 Analytical Chemistry, Vol. 76, No. 4, February 15, 2004
f 327 and 366 f 330 were monitored for cortisol and cortisol-d3, respectively. A daughter ion at m/z 121 was the most intense ion from a collision-induced fragmentation of the cortisol molecular ion. However, this ion, representing the ring-A moiety common to many steroids with a similar chemistry structure, is relatively nonspecific, and is the common product ion yielded from the molecular ions of both cortisol and cortisol-d3 (The deuterium labels of cortisol-d3 do not remain on the daughter ion.). Therefore, the slightly less intense, but more specific daughter ions at m/z 327 and 330 for cortisol and cortisol-d3, respectively, were chosen for measurement. The results of LC/MS/MS measurement are shown in Table 4. Excellent reproducibility was obtained for all three concentrations with within-set CVs ranging from 0.7% to 1.5% and between-set CVs ranging from 0.04% to 0.4%. A linear regression line was generated for each set of samples. Excellent linearity was also obtained with the correlation coefficients (R) of all linear regression lines (measured intensity ratios vs mass ratios) ranging from 0.998 to 1.000. The LC/MS and LC/MS/MS results were in very good agreement (within 0.2%). The detection limits (at a signal-to-noise ratio of ∼3-5) were 10 and 15 pg for LC/MS and LC/MS/MS, respectively. The quantities injected (∼6 ng for LC/MS and ∼10 ng for LC/MS/MS) are far above (∼600 times) the detection limits. Ion chromatograms for cortisol and cortisol-d3 are shown in Figure 2 for LC/MS and Figure 3 for LC/MS/MS.
Table 4. LC/MS/MS-ESI Measurements of Cortisol in Serum (µg/L) conc
set
mean
SDa
CV, %
1 1 1 2 2 2 3 3 3
1 2 3 1 2 3 1 2 3
83.4 83.5 83.4 165.5 165.7 165.5 248.7 246.3 249.2
0.58 1.15 0.92 1.19 1.65 2.44 1.46 2.34 2.47
0.7 1.4 1.1 0.7 1.0 1.5 0.6 0.9 1.0
overall mean
overall SDb
overall CV, %
83.4
0.05
0.1
165.6
0.06
0.04
248.1
0.90
0.4
a SD of a single measurement within a set. b SD of the mean for that level.
Table 5. Calculation of Expanded Uncertainties for Serum Cortisol by LC/MS-ESI and LC/MS/MS-ESI Measurements uncertainty table
conc 1
conc 2
conc 3
LC/MS SD, µg/L LC/MS/MS SD, µg/L difference between methods, µg/L purity, µg/L volumetric, µg/L combined SD uncertainty, µg/L degrees of freedom k factor expanded uncertaintya, µg /L mean, µg /L relative expanded uncertainty, %
0.26 0.05 0.21 0.17 0.83 0.91 315 2 1.83 83.3 2.2
0.16 0.06 0.11 0.33 1.66 1.70 26298 2 3.40 165.5 2.1
0.51 0.90 0.06 0.50 2.48 2.73 157 2 5.46 248.1 2.2
a
Uncertainty of 95% confidence interval.
Table 6. Comparison of Combined Results from LC/MS and LC/MS/MS Measurements with Routine Clinical Methods, Mean Results
conc
LC/MS and LC/MS/MS (this study), µg/L
routine methods,a µg/L
relative difference, %
1 2 3
83.3 ( 1.8b 165.5 ( 3.4b 248.1 ( 5.5b
99.8 ( 14.2c 191.6 ( 20.3c 282.7 ( 31.7c
19.8 15.8 13.9
a Private communication from the CAP. b Uncertainty of 95% confidence interval. c Standard deviation of the composite results from all the methods.
Figure 3. Ion chromatograms by LC/MS/MS-ESI for cortisol and cortisol-d3 from a serum sample, concentration 1.
Statistical Analysis. The overall uncertainty in the measurements includes the measurement repeatability (type A) and other factors (type B) that include the bias between the two measurement approaches, uncertainty in the purity of the reference compound, and uncertainty associated with the use of volumetric additions. Components of uncertainty and their magnitude are listed in Table 5. The various components were combined quadratically25 to determine the standard uncertainty for each level. For the measurements using each method, analyses of variance testing found that set-set differences were significant in most cases; thus, they were assumed to be significant for both methods across all three materials. The largest component of the uncertainty is associated with volumetric additions, estimated to be 1.0%, based upon gravimetric measurements which showed that errors in volumetric steps can approach 1%. Because this factor and the other type B components contribute most of the uncertainty and have very large degrees of freedom, the overall degrees of freedom, as calculated using the Welch-Satterthwaite approximation, for each level are high, resulting in a k factor near (25) Levenson, M. S.; Banks, D. L.; Eberhardt, K. R.; Gill, L. M.; Guthrie, W. F.; Liu, H. K.; Vangel, M. G.; Yen, J. H.; Ahang, N. F. J. Res. Natl. Inst. Stand. Technol. 2000, 105, 571-579.
2. The degrees of freedom for the middle level was a particularly large number, because the high measurement precision for this level made the type A uncertainty very small compared with the type B factors. A value of 2 was used for multiplying the standard uncertainty for each level to calculate the expanded uncertainty, which is intended to represent a 95% confidence interval. Comparison to Routine Clinical Methods (Based on Immunoassays from CAP Surveys). The overall results of LC/ MS-ESI and LC/MS/MS-ESI measurements and the mean values of 10 routine clinical methods (based on immunoassays from ∼500 laboratories from CAP surveys) are compared in Table 6. The interlaboratory CVs for a given immunoassay ranged from 5.8% to 10.2%, and the overall CVs ranged from 10.6% to 14.2% for the three concentrations measured in this study. It appears that the routine clinical method means may be biased high by 14%-20% across the concentrations. This bias probably results from crossreactivities of immunoassays with other corticosteroids in serum. CONCLUSIONS This method is based on sound theoretical principles, is free from interferences from structural analogues, demonstrates quantitative recoveries of added cortisol, produces results confirmed by LC/MS/MS with relatively small uncertainties, and compared well with certified values determined using GC/MS reference methods. Therefore, this method qualifies as a candidate reference method and can be used as an alternative method to provide an accuracy base to which routine methods can be compared. NIST is planning to use this method to certify the concentration of cortisol in a new SRM for hormones in frozen Analytical Chemistry, Vol. 76, No. 4, February 15, 2004
1013
serum pools, which is intended to provide a means for laboratories to test the accuracy of their methods, calibrators, and controls and to demonstrate traceability of their measurements to the NIST reference method. ACKNOWLEDGMENT We gratefully acknowledge professor Linda Thienpont for a gift of a corticosteroid, 18-hydroxycorticosterone, for interference testing, and Polly Ellerbe for her initial participation in the
1014
Analytical Chemistry, Vol. 76, No. 4, February 15, 2004
development of this candidate reference method. Certain commercial equipment, instruments, and materials are identified in this paper to adequately specify the experimental procedure. Such identification does not imply recommendation or endorsement by NIST nor does it imply that the equipment, instruments, or materials are necessarily the best available for the purpose. Received for review August 18, 2003. Accepted December 3, 2003. AC034966F
