Selected performance characteristics of the Roche Elecsys® testosterone II assay on the Modular analytics E 170 analyzer

Selected performance characteristics of the Roche Elecsys® testosterone II assay on the Modular analytics E 170 analyzer

Clinica Chimica Acta 411 (2010) 1073–1079 Contents lists available at ScienceDirect Clinica Chimica Acta j o u r n a l h o m e p a g e : w w w. e l ...

1MB Sizes 15 Downloads 161 Views

Clinica Chimica Acta 411 (2010) 1073–1079

Contents lists available at ScienceDirect

Clinica Chimica Acta j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / c l i n c h i m

Selected performance characteristics of the Roche Elecsys® testosterone II assay on the Modular analytics E 170 analyzer William E. Owen a, Mindy L. Rawlins a, William L. Roberts b,⁎ a b

ARUP Institute for Clinical and Experimental Pathology, Salt Lake City, UT, United States Department of Pathology, University of Utah Health Sciences Center, Salt Lake City, UT, United States

a r t i c l e

i n f o

Article history: Received 2 November 2009 Received in revised form 29 March 2010 Accepted 29 March 2010 Available online 10 April 2010 Keywords: Imprecision Method comparison Interference Chemiluminescent Immunoassay

a b s t r a c t Background: Serum testosterone measurements have utility in diagnosis of clinical conditions characterized by both increased and decreased testosterone concentrations. Studies have indicated that testosterone immunoassays may give inaccurate results for women and children. We evaluated the performance of a second generation testosterone immunoassay from Roche Diagnostics. Methods: Testo II performed on a Modular Analytics E 170 analyzer is an automated random access electrochemiluminometric assay. We evaluated limit of blank (LoB), imprecision, linearity, interference, and method comparison with liquid chromatography-tandem mass spectrometry assay (LC-MS/MS). Method comparison included the current generation Roche testosterone assay (Testo I) and the Access 2 testosterone chemiluminometric assay (Beckman Coulter). Results for men and women were analyzed for analytic concordance. The relative % differences of immunoassay compared to LC-MS/MS results were evaluated. Results: The LoB was 0.07 nmol/l. Total imprecision was b 6%. The assay was linear from 0.2 to 46.6 nmol/l. Negative interference was observed for lipemia at concentrations N22.5 g/l. Analytic concordance showed improved specificity for women. Comparison of results to LC-MS/MS indicated comparable performance with other immunoassays for men and improved performance for women, boys, and girls with mean differences of 0.5%, − 0.7%, and 24.4%, respectively. Conclusions: The Roche Testo II assay demonstrated excellent precision. Comparison to 2 other automated immunoassays showed comparable performance for men and improved performance for women and children. However, challenges still exist for quantifying testosterone concentrations b 10.4 nmol/l for men and b 1.7 nmol/l for women and children by immunoassay. © 2010 Elsevier B.V. All rights reserved.

1. Introduction Testosterone is an androgen produced in Leydig cells of the testes in males. In females testosterone is produced principally in the ovaries. Normal concentrations of testosterone in females are approximately 10-fold less than in males. In males decreased testosterone may result from hypogonadism, hypopituitarism, hyperprolactinemia, renal failure, hepatic cirrhosis, or Kleinfelter's syndrome, while elevated testosterone can be caused by adrenal and testicular tumors. In females increased testosterone can be caused by polycystic ovary syndrome, stromal hyperthecosis, ovarian and Abbreviations: LoB, Limit of blank; Testo I, Current generation Roche Elecsys Testosterone assay; Testo II, New generation Roche Elecsys Testosterone II assay; Access 2, Beckman Coulter Testosterone assay; LC-MS/MS, High performance liquid chromatography-tandem mass spectrometry testosterone assay; ID-GC–MS, Isotope-dilution gas chromatography–mass spectrometry; CLSI, Clinical and Laboratory Standards Institute. ⁎ Corresponding author. c/o ARUP Laboratories, 500 Chipeta Way, Salt Lake City, UT 84108, United States. Tel.: + 1 801 583 2787x2086; fax: + 1 801 584 5207. E-mail address: [email protected] (W.E. Owen). 0009-8981/$ – see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.cca.2010.03.041

adrenal tumors, or congenital adrenal hyperplasia [1–3]. Published studies have indicated that automated immunoassay methods for testosterone are acceptable for measuring testosterone concentrations seen in healthy men but may give inaccurate results for women and children due to poor specificity and precision at concentrations which are typically b1.7 nmol/l [4–7] 2. Materials and methods Testo I and Testo II (Roche Diagnostics, Indianapolis, IN) performed on a Modular Analytics E 170 analyzer are fully automated random access competitive immunoassays with electrochemiluminescent detection. Testo II is a second generation testosterone assay which differs from Testo I in 4 ways. First, it uses a high affinity sheep monoclonal antibody instead of a mouse monoclonal antibody that is used for Testo I. Second, it uses a smaller sample size of 20 μl compared to 50 μl for Testo I. Third, it uses a 2-morpholinoethanesulfonic acid buffer instead of the phosphate buffer used for Testo I. Fourth, the releasing reagent is 2bromoestradiol instead of 8-anilino-1-naphthalenesulfonic acid and norgestrel that are used for Testo I. Access 2 (Beckman Coulter, Brea, CA)

1074

W.E. Owen et al. / Clinica Chimica Acta 411 (2010) 1073–1079

Table 1 Summary of Testo II imprecision study results. Sample

Mean concentration nmol/l

Within-run CV (%)

Total CV (%)

HSP01a HSP03a HSP04a PCU2b HSP02a PCU1b HSP05a

1.2 2.6 8.0 8.7 17.6 21.2 45.8

2.6 2.4 1.8 2.0 2.1 1.6 2.2

5.1 3.5 2.6 2.5 2.7 2.3 2.8

a b

Patient pool. Quality control material.

Fig. 1. Intralipid interference study for Testo II. The x-axis shows the concentration of triglycerides tested, the y-axis shows the % recovery of testosterone in interference samples. The best fit regression line for the data is shown (r = 0.94).

is a fully automated random access competitive immunoenzymatic assay with chemiluminescent detection. All immunoassays were configured and calibrated on instruments in our research laboratory according to manufacturers' instructions. The 3 testosterone immu-

noassays were compared to LC-MS/MS to assess how performance of a second generation assay (Testo II) compared with that of the current Testo I assay and another commercially available assay (Access 2). The LoB study of Testo I and Testo II used a zero pool material from surplus female patient samples which had testosterone concentrations b0.03 nmol/l by LC-MS/MS. These samples were screened by the Testo I and Testo II assays and those with testosterone concentrations below the reporting limit for both assays were combined to obtain the zero pool material. The LoB was assessed by assaying 10 replicates of zero pool material and 3 replicates of a low non-zero testosterone CalCheck solution (Roche). The LoB study of the Access assay used the zero calibrator and the lowest non-zero calibrator in the Access testosterone calibrator set (Beckman Coulter). The zero calibrator was assayed in 10 replicates and the non-zero calibrator was assayed in 3 replicates. For all three immunoassays the raw instrument responses for the zero pool or zero calibrator and non-zero solution or non-zero calibrator were analyzed to determine the 2 SD limit from the mean of zero sample responses. Imprecision was tested according to CLSI 5 A2 guidelines using 2 quality control materials (PreciControl Universal Levels 1 and 2, Roche Diagnostics) and 5 pools prepared from surplus patient serum (normal male, low male, high female, normal female and low female). Materials for each level were pooled, aliquotted and frozen at − 70 °C until the time of testing. Sufficient aliquots of each pool were frozen to perform 2 runs per day in duplicate for 21 days for a total of 84 replicates per pool. Linearity studies were performed using dilutions of a high patient pool of samples from men with a low patient pool of samples from boys to the following percentages: 100, 75, 50, 25, 10, 5, 1.25 and 0 and were analyzed in duplicate by all three immunoassay methods. The expected concentrations for the dilutions were based on the mean results of the high and low pools. The mean result for each dilution sample was compared to the expected result and the % recovery was calculated as well as slope, intercept, and r.

Table 2 Analytic concordance for men and women. LC-MS/MS (cutoff 10.4 nmol/l)a

Men Testo I Negative Positive Total Men Testo II Negative Positive Total Men Access 2 Negative Positive Total

a b c d

Specificityb (%)

Agreement (%)

Cohen's kappa

Negative

Positive

Total

73 0 73

5 47 52

78 47 125

90.4

100.0

96.0 (91.0–98.3)c

0.917 (0.845–0.988)c

73 0 73

5 47 52

78 47 125

90.4

100.0

96.0 (91.0–98.3)

0.917 (0.845–0.988)

71 2 73

2 50 52

73 52 125

96.2

97.3

96.8 (92.1–98.7)

0.934 (0.871–0.998)

LC-MS/MS (cutoff 1.7 nmol/l)d

Women Testo I Negative Positive Total Women Testo II Negative Positive Total Women Access 2 Negative Positive Total

Sensitivityb (%)

Sensitivityb (%)

Specificityb (%)

Agreement (%)

Cohen's kappa

Negative

Positive

Total

61 15 76

1 51 52

62 66 128

98.1

80.3

87.5 (80.7–92.2)

0.751 (0.638–0.865)

69 7 76

3 49 52

72 56 128

94.2

90.8

92.2 (86.2–95.7)

0.840 (0.745–0.935)

60 16 76

1 51 52

61 67 128

98.1

78.9

86.7 (79.8–91.5)

0.737 (0.620–0.853)

Positive results for men are b 10.4 nmol/l. The sensitivity and specificity are calculated relative to the LC-MS/MS comparison method. 95% confidence limits. Positive results for women are N1.7 nmol/l.

W.E. Owen et al. / Clinica Chimica Acta 411 (2010) 1073–1079

Studies were performed to determine interference from bilirubin, hemoglobin, and triglycerides (Intralipid) using stock interferent solutions prepared according to procedures previously described [8]. A low (0%) pool was prepared from surplus male samples and high (100%) interferent pools were made by adding bilirubin stock solution to the low pool to a final concentration of 0.7 g/l, hemolysate stock was added to a final hemoglobin concentration of 40.0 g/l, and

1075

Intralipid stock was added to a final triglycerides concentration of 40.0 g/l. These high (100%) interferent pools were then mixed with the low pool according to a preparation protocol previously published to create 75, 50 and 25% interferent pools [9]. The concentrations of bilirubin, triglycerides were measured using a Modular Analytics P800 analyzer (Roche Diagnostics, Indianapolis, IN) with Roche reagents. Hemoglobin concentrations were estimated using the H index from a

Fig. 2. Method comparison % difference plots for three immunoassays vs LC-MS/MS for adults. The y-axis is the % difference calculated as described in the Materials and methods section. The solid line represents the mean % difference. The dashed lines indicate the parametric 95% limits of agreement about the mean. Panels A–C show results for men. For panels A and C, n = 125 and for panel B, n = 123. Panels D–F show results for women. For panel D, n = 124 for panel E, n = 121 and for panel F, n = 116. Immunoassays include Testo I (panels A and D), Testo II (panels B and E) and Access 2 (panels C and F).

1076

W.E. Owen et al. / Clinica Chimica Acta 411 (2010) 1073–1079

Modular Analytics P800 analyzer, which is roughly equivalent to hemoglobin concentration in mg/dL. Testosterone was measured in triplicate and the recovery in samples with interferent added that was between 90 and 110% of the mean concentration in the interferent free pool was considered acceptable.

The comparison method was a high performance liquid chromatography-tandem mass spectrometry assay (LC-MS/MS) used in our facility for testing of samples from women and children [10,11]. Serum samples from 125 men, 128 women, 120 boys (ages 0–17 y), to 110 girls (ages 0– 17 y) were used for method comparison studies. These were surplus

Fig. 3. Method comparison % difference plots for three immunoassays vs LC-MS/MS for children. The y-axis is the % difference calculated as described in the Materials and methods section. The solid line represents the mean % difference. The dashed lines indicate the parametric 95% limits of agreement about the mean. Panels A–C show results for boys. For panel A, n = 117 and for panels B and C, n = 101. Panels D–F show results for girls. For panel D, n = 109 for panel E, n = 101 and for panel F, n = 103. Immunoassays include Testo I (panels A and D), Testo II (panels B and E) and Access 2 (panels C and F).

W.E. Owen et al. / Clinica Chimica Acta 411 (2010) 1073–1079

patient samples previously analyzed by our LC-/MS/MS method and stored at −20 °C as clinical samples for a maximum of 4 weeks. These were retrieved from storage, de-identified, and stored in a research freezer at −70 °C for up to 2 months until the time of testing. It was the goal of our collections to obtain samples with testosterone concentrations covering the full measurement range of the immunoassays. This was met for all subject groups except for girls. Samples were thawed, centrifuged and tested by all immunoassays and stored at 2–8 °C until re-analysis by LC-MS/MS. All samples were re-analyzed by LC-MS/MS within 48 h of immunoassay testing and then stored frozen. Samples were considered discrepant if some method results were within the reference interval and other method results fell outside the reference interval and/or any results differed by N1.3 nmol/l. By these criteria, four samples were determined to be discrepant. These discrepant samples were thawed and retested by all methods to confirm results. Repeat results for all four discrepant samples were within 10% of the original. Therefore, the original results were used in subsequent data analyses. Results for samples from men and women were evaluated for analytic concordance compared to LC-MS/ MS. Positive cutoffs of b10.4 nmol/l for men and N1.7 nmol/l for women were used for analyzing concordance. Because of the variable nature of results from children, it was not possible to determine suitable cutoffs for use in analytic concordance. For this reason we did not perform analytic concordance analyses for results from boys or girls. The relative % difference of immunoassay compared to LC-MS/MS was calculated for every sample with a measurable testosterone result. Samples whose immunoassay results were outside the analytic measurement range were eliminated from further analysis. The % difference was plotted against the

1077

LC-MS/MS testosterone concentrations. Also the mean % difference and the parametric 95% limits of agreement about the mean were determined. In addition we compared Testo II to Testo I and Access 2 using % difference plots as previously described [12]. Approval of the Institutional Review Board of the University of Utah Health Sciences Center, Salt Lake City, was obtained for all studies in which samples from human subjects were used. EP Evaluator Release 8 software (D.G. Rhoads Associates Inc., Kennett Square, PA) was used to determine LoB, linearity, and analytic concordance statistics including sensitivity, specificity, agreement and Cohen's kappa statistic. 3. Results The LoB for Teso II was 0.07 nmol/l, compared with 0.12 and 0.11 nmol/l for Testo I and Access 2, respectively. Imprecision data are summarized in Table 1. Testosterone concentrations ranging from 1.2 to 45.8 nmol/l showed within-run CV's of 1.6 to 2.6% and total CV's of 2.3 to 5.1%. The Testo II assay was linear over the range of 0.2 to 46.6 nmol/ l (slope= 1.01, intercept= −0.1 nmol/l and r = 0.9997). The maximum deviation from the target recovery was 17.7% at a concentration of 0.6 nmol/l and the average recovery was 95.7%. The Testo I assay was linear over the range of 0.4 to 43.1 nmol/l (slope = 1.02, intercept= −0.3 nmol/l, and r = 0.9997). The maximum deviation from the target recovery was 20.8% at a concentration of 2.0 nmol/l and the average recovery was 93.8%. The Access 2 assay was linear over the range of 0.5 to 40.2 nmol/l (slope= 1.03, intercept= 0.1 nmol/l, and r = 0.9990).

Fig. 4. Method comparison % difference plots for Testo II vs two immunoassays for adults. The x-axis is the mean testosterone concentration. The y-axis is the % difference. The solid line represents the mean % difference. The dashed lines indicate 2 standard deviations from the mean. Panels A and B show results for men. Panels C and D show results for women. For panel A, n = 123, for panel B, n = 123, for panel C, n = 121, and for panel D, n = 116. Panels A and C show Testo II vs Testo I. Panels B and D show Testo II vs Access 2.

1078

W.E. Owen et al. / Clinica Chimica Acta 411 (2010) 1073–1079

The maximum deviation from the target recovery was 7.4% at a concentration of 32.5 nmol/l and the average recovery was 102.2%. Interference studies with bilirubin and hemoglobin showed no significant interference up to concentrations of 0.7 g/l and 40.0 g/l, respectively. Negative interference, with testosterone concentrations below 90% of the expected recovery, was seen for Intralipid triglyceride concentrations greater than 22.5 g/l (Fig. 1). Statistics for analytic concordance studies of the three immunoassays' testosterone results for men and women compared to LC-MS/MS are summarized in Table 2. The Testo II assay showed performance comparable to the Testo I and Access 2 assays for men and improved performance in terms of specificity and overall agreement for women. Method comparison results of immunoassay vs LC-MS/MS as difference plots are shown for men and women (Fig. 2). For men the mean difference of 9.8% for Testo II (Fig. 2, panel B) was comparable (within 95% limits of agreement) to Testo I and Access 2 with mean differences of 6.8% and 1.2%, respectively (Fig. 2, panels A and C). The range of Testo II differences was 18.1% to 106.7% which was similar to Testo I and Access 2 with ranges of −33.3% to 102.4% and −32.8% to 68.3%, respectively. Improvement in Testo II specificity relative to LC-MS/MS for women was seen in a mean difference of 0.5% (Fig. 2, panel E) vs 37.2% and 47.9% for Testo I and Access 2, respectively (Fig. 2, panels D and F). Also, the range of differences was improved for Testo II, −80.0% to 179.2% vs −54.8% to 679.2% and −48.2% to 342.1% for Testo I and

Access 2, respectively. These data show that the Testo II assay does tend to underestimate testosterone more than Testo I and Access 2 at concentrations of b1.7 nmol/l by LC-MS/MS. Difference plots for boys and girls are shown (Fig. 3). Data for boys indicate improvement in Testo II performance with a mean difference of −0.7% (Fig. 3, panel B) vs 24.7% and 41.7% for Testo I and Access 2 respectively (Fig. 3, panels A and C). Also, the range of differences was −87.5% to 223.5% which was lower than Testo I and Access 2 with ranges of − 72.7% to 482.4% and − 47.4% to 341.2%, respectively. However, the largest difference in Testo II vs LC-MS/MS for boys was 223.5%, which was the highest seen for any subject group in our study. Also, the absolute magnitude of the most negative difference of −87.5% was greater than those of Testo I and Access 2 with values of −72.7% and −47.4%, respectively, indicating lower recovery at very low concentrations of testosterone. Differences for girls indicate improvement in trueness with a mean difference of 24.4% (Fig. 3, panel E) vs 95.4% and 94.8% for Testo I and Access 2, respectively (Fig. 3, panels D and F). This was also seen in the range of differences of −50.0% to 204.8% for Testo II vs −42.9% to 600.0% and −35.4% to 460.0% for Testo I and Access 2, respectively. Percent difference plots of Testo II compared to Testo I and Access 2 with results for adults and children are shown in Figs. 4 and 5, respectively. The mean biases relative to the Testo I and Access 2 assays for samples from men were 3.2% and 8.2%, respectively (Fig. 4, panels A and B). For samples from

Fig. 5. Method comparison % difference plots for Testo II vs two immunoassays for children. The x-axis is the mean testosterone concentration. The y-axis is the % difference. The solid line represents the mean % difference. The dashed lines indicate 2 standard deviations from the mean. Panels A and B show results for boys. Panels C and D show results for girls. For panel A, n = 101, for panel B, n = 98, for panel C, n = 101, and for panel D, n = 100. Panels A and C show Testo II vs Testo I. Panels B and D show Testo II vs Access 2.

W.E. Owen et al. / Clinica Chimica Acta 411 (2010) 1073–1079

women the mean biases relative to the Testo I and Access 2 assays were −26.8% and −32.8%, respectively (Fig. 4, panels C and D). The mean biases relative to the Testo I and Access 2 assays for samples from boys were −17.7% and −26.2%, respectively (Fig. 5, panels A and B). For sample from girls the mean biases relative to the Testo I and Access 2 assays were −39.6% and −41.0%, respectively (Fig. 5, panels C and D). A serum based secondary reference preparation for testosterone was not available at the time of our study to determine what bias may have been present in our LC-MS/MS method. Based on a study of LCMS/MS methods including our own compared to an ID-GC–MS reference measurement procedure, we assumed a 10% level of uncertainty in our LC-MS/MS data [11]. The biologically derived total error limit for testosterone is 14% [11]. We therefore calculated the percentage of samples exceeding a total error of 25% which is a standard that most LC-MS/MS methods can meet [11]. For men the Testo II had 9.5% outside the 25% total error limit vs 9.6% and 4.8% for Testo I and Access 2 respectively. Samples for women showed a value of 30.6% for Testo II vs 56.5% and 51.7% for Testo I and Access 2 respectively. For boys Testo II had 27.7% of samples outside the 25% total error limit compared with 32.5% and 33.7% for Testo I and Access 2 respectively. 49.5% of Testo II results for girls exceeded the limit compared with 83.5% and 88.6% for Testo I and Access 2, respectively. 4. Discussion The LoB for Testo II was improved compared to the Testo I and Access 2 assays. Imprecision results indicated the Testo II assay has excellent reproducibility. It compared well with reported performance for immunoassay and LC-MS/MS methods [4,6,10,11,13–16]. We saw similar imprecision for patient serum pools and control material at comparable testosterone concentrations. Linearity studies indicated acceptable performance comparable to other immunoassays and LC-MS/MS over the analytic measurement range of the assay. The Testo II assay showed no significant interference from bilirubin or hemoglobin up to very high interferent concentrations of 0.7 g/l and 40.0 g/l, respectively. Significant negative interference occurred from triglycerides (Intralipid), but only at high concentrations (N22.5 g/l). This resistance to interference was notable since the Testo II assay utilizes a delayed competitive test method with no wash step, but it incorporates a high affinity sheep monoclonal antibody allowing a small sample size and uses a high protein concentration in the buffer to minimize effects from possible interfering substances in the sample matrix [17]. For analytic concordance our decision cutoffs were the lower reference limit for men (10.4 nmol/l) and the upper reference limit for women (1.7 nmol/l). We defined positive based on the primary use of testosterone testing i.e. hypogonadism in men and androgen excess in women. Results of analytic concordance (Table 2) for men indicated comparable performance for the Testo II assay compared with Testo I and Access 2. Analytic concordance studies indicated that the specificity for women was 90.8%, which was improved over the Testo I and Access 2 assays which had specificities 80.3 and 78.9%, respectively. These results parallel those that were reported previously [17]. These analytic concordance results for men and women were reflected in our method comparison data (Fig. 2). These data do indicate that Testo I and Testo II on average slightly over estimate testosterone in samples from men relative to LC-MS/MS. This has been reported previously for Testo I compared to LC-MS/MS but contrasts with results reported for Testo I compared to isotope-dilution gas chromatography–mass spectrometry (ID-GC–MS) [7,16]. The mean differences and range of differences for individual subjects indicated that the Testo II assay has improved accuracy for women and children and shows comparable performance for men compared to Testo I. These findings were also reflected in the comparisons of Testo II to Testo I and Access 2 (Figs. 4 and 5). The negative biases we saw for women and children were most pronounced at low concentrations of testosterone. We speculate this was due to

1079

reduced cross-reactivity of the Testo II assay antibody with steroids present in these samples that are structurally similar to testosterone. Determining the nature of the cross-reacting substances was beyond the scope of this study. A review of the package inserts for Testo I and Testo II provided some limited information. Decreased cross-reactivity was reported for Testo II compared to Testo I for dehydroepiandrosterone sulfate (decreased by 67%), 5-α-dihydrotestosterone (decreased by 54%), and 11-ketotestosterone (decreased by 69%). Increased cross-reactivity for Testo II compared to Testo I was reported for androstenedione (increased by 175%), 5-α-androstane-3β,17β-diol (increased by 314%), and 11-β-hydroxytestosterone (increased by 116%). Another possible explanation could be interactions between the labeled testosterone derivative in the assay, the assay antibody, and sex-hormone binding globulin which has previously been reported to cause overestimation of testosterone concentrations with other direct immunoassays [7]. Testo II assay showed excellent precision and acceptable performance for LoB, linearity and interferences. This assay gave comparable performance for samples from men and improved performance for samples from women and children compared to other immunoassays. This assay was not optimal for measuring testosterone concentrations b10.4 nmol/l in men or b1.7 nmol/l in women and children. Acknowledgements Roche Diagnostics provided reagents and support for this study. E 170 and Access 2 analyzers were provided by Roche Diagnostics and Beckman Coulter, respectively. The ARUP Institute for Clinical and Experimental Pathology provided support for this study. References [1] Barbieri RL. Hyperandrogenic disorders. Clin Obstet Gynecol 1990;33:640–54. [2] Bhasin S. Testicular disorders. In: Kronenberg H, Melmed S, Polonsky KS, Larsen PR, editors. Williams Textbook of Endocrinology Philadelphia. Saunders/Elsevier; 2008. p. 645–80. [3] Bulun S, Adashi EY. The physiology and pathology of the female reproductive axis. In: Kronenberg H, Melmed S, Polonsky KS, Larsen PR, editors. Williams Textbook of Endocrinology. Philadelphia: Saunders/Elsevier; 2008. p. 569–90. [4] Cawood ML, Field HP, Ford CG, et al. Testosterone measurement by isotopedilution liquid chromatography-tandem mass spectrometry: validation of a method for routine clinical practice. Clin Chem 2005;51:1472–9. [5] Herold DA, Fitzgerald RL. Immunoassays for testosterone in women: better than a guess? Clin Chem 2003;49:1250–1. [6] Moal V, Mathieu E, Reynier P, Malthiery Y, Gallois Y. Low serum testosterone assayed by liquid chromatography-tandem mass spectrometry. Comparison with five immunoassay techniques. Clin Chim Acta 2007;386:12–9. [7] Taieb J, Mathian B, Millot F, et al. Testosterone measured by 10 immunoassays and by isotope-dilution gas chromatography–mass spectrometry in sera from 116 men, women, and children. Clin Chem 2003;49:1381–95. [8] Owen WE, Rawlins ML, La'ulu SL, Roberts WL. Performance characteristics of the Access pregnancy-associated plasma protein-A assay. Clin Chim Acta 2008;398:165–7. [9] McEnroe RJ, Burrit MF, Powers DM. Interference Testing in Clinical Chemistry; Approved Guideline- Second Edition (EP7-A2). Wayne: Clinical and Laboratory Standards Institute; 2008. [10] Kushnir MM, Rockwood AL, Roberts WL, et al. Performance characteristics of a novel tandem mass spectrometry assay for serum testosterone. Clin Chem 2006;52:120–8. [11] Thienpont LM, Van Uytfanghe K, Blincko S, et al. State-of-the-art of serum testosterone measurement by isotope dilution-liquid chromatography-tandem mass spectrometry. Clin Chem 2008;54:1290–7. [12] Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;1:307–10. [13] Levesque A, Letellier M, Swirski C, Lee C, Grant A. Analytical evaluation of the testosterone assay on the Bayer Immuno 1 system. Clin Biochem 1998;31:23–8. [14] Sanchez-Carbayo M, Mauri M, Alfayate R, Miralles C, Soria F. Elecsys testosterone assay evaluated. Clin Chem 1998;44:1744–6. [15] Singh RJ. Validation of a high throughput method for serum/plasma testosterone using liquid chromatography tandem mass spectrometry (LC-MS/MS). Steroids 2008;73:1339–44. [16] Wang C, Catlin DH, Demers LM, Starcevic B, Swerdloff RS. Measurement of total serum testosterone in adult men: comparison of current laboratory methods vs liquid chromatography-tandem mass spectrometry. J Clin Endocrinol Metab 2004;89:534–43. [17] Oldekamp J, Hirzel K, Schneider E, Gassner D. Development of an Elecsys® Testosterone II Immunoassay with an improved performance for measurement of testosterone in women. Endocr Abstr 2008;16:631.