Original Article Simultaneous determination of androstenedione and testosterone in human serum by liquid chromatography-tandem mass spectrometry L M Gallagher, L J Owen and B G Keevil Abstract Address Department of Clinical Biochemistry, South Manchester University Hospitals NHS Trust, Manchester M23 9LT, UK Correspondence Mrs L J Owen Email: laura.owen@smuht.nwest.nhs.uk Background We aimed to develop a sensitive assay to quantitate serum concentrations of both androstenedione and testosterone within the female range simultaneously, using liquid chromatography-tandem mass spectrometry (LC-MS/ MS) for use in the routine clinical laboratory and to compare this method with immunoassay. Method Samples (2 ml) were prepared by liquid liquid extraction using (1 ml) methyl-tert-butyl-ether. Deuterated androstenedione and testosterone were used as internal standards. Results The standard curve was linear to 5 nmol/l, the lower limit of quantitation was.25 nmol/l, and intra- and inter-assay coefficients of variation were o1% for both androgens over the range.3 35 nmol/l. There was a poor relationship between the LC-MS/MS and the radioimmunoassay methods for androstenedione with the LC-MS/MS generally giving lower results. For testosterone, the LC-MS/MS and immunoassay methods compared well at all concentrations. However, when female samples only were examined, the agreement deteriorated. Conclusions We have developed a sensitive and precise LC-MS/MS method, which gives more accurate results for all androstenedione measurements and low testosterone concentrations than immunoassay. Introduction Testosterone concentration is routinely requested in the investigation of oligo- or amenorrhoea, hirsutism and/or acne in women. The calculation of a free androgen index (FAI) using testosterone and sex-hormone-binding globulin (SHBG) concentrations has been shown to be a sensitive diagnostic marker of polycystic ovarian syndrome (PCOS). 1 The likelihood of a diagnosis of PCOS is increased if the androstenedione level is also raised. 2 Testosterone concentrations often do not correlate well with symptoms, but may help in prompting further investigation. A testosterone concentration of o5 nmol/ L measured by immunoassay is rarely associated with serious pathology, but higher concentrations warrant further investigation for the detection of an androgensecreting neoplasm 3 or congenital adrenal hyperplasia. 4 It has been documented that radioimmunoassays (RIAs) for androstenedione measurement may overestimate steroid concentrations in children and adults. 5,6 Commercial direct immunoassays for testosterone, which have almost completely replaced old extraction assays, can also lead to spuriously high measurements in women and children. 5,7--9 We report a liquid chromatography-tandem mass spectrometry (LC-MS/MS) method which is sensitive, accurate and precise to measure androstenedione and testosterone simultaneously. This method should prove useful in the assessment of female patients presenting with hirsutism, virilization or acne. Although LC-MS/ MS assays for testosterone do exist, 1,11 to our knowledge, none that can measure androstenedione alone or simultaneously has been reported. Materials and methods Androstenedione and testosterone were supplied by Sigma Diagnostics (Poole, Dorset, UK). Deuterated 48 r 27 The Association for Clinical Biochemistry
Simultaneous determination of androstenedione and testosterone 49 androstenedione (d 7 A,98% isotopic purity) and deuterated testosterone (d 2 T, 99% isotopic purity) were supplied by CDN Isotopes (Quebec, Canada). Stock androstenedione and testosterone solutions were prepared for standardization by dissolving pure compound in methanol and stored at 21C. A separate stock solution was prepared for quality controls (QCs) for both analytes in agreement with published guidelines. 12 Mixed working standards were prepared by diluting the stock solutions of androstenedione and testosterone in phosphate-bu ered saline (PBS), ph 7.4, containing.1% bovine serum albumin (BSA), to give a range of concentrations (--5 nmol/l) for both androstenedione and testosterone. QCs were prepared in the same manner to give concentrations of.3--35 nmol/l for both analytes. The use of PBS-based QCs allowed us to calculate the percentage deviation from target concentrations, which would not have been possible had we used commercial QCs. Aliquots of 5 ml were stored in microcentrifuge tubes at 21C for up to six months. Internal standards were prepared by diluting d 7 A and d 2 T to concentrations of.1 and.2 mg/l, respectively, in methanol (high-performance liquid chromatography [HPLC] grade, Fisher, Loughborough, UK) and stored at 21C. Sample preparation Serum samples, standards or QC samples (2 ml) were transferred in singleton into polypropylene microcentrifuge tubes (1.5 ml). Internal standard (1 ml) and 1mL methyl-tert-butyl-ether (HPLC grade, Fisher, Loughborough, UK) were added. Following this, the tubes were stoppered and vortex mixed on a mechanical mixer for 4 min. The supernatant from each tube was then transferred into a glass tube, and the tubes were placed in a 41C heating block. The solvent was evaporated under a gentle air ow for approximately 15 min. The residue was then reconstituted with 1 ml of 5:5 mobile phase (water [A] and methanol [B: HPLC grade, Fisher, Loughborough, UK]), each containing 2 mmol/l ammonium acetate (Sigma, Poole, UK) and.1% (v/v) formic acid (Analar, VWR International, Poole, UK). The tubes were vortex mixed for 1min and extract was then transferred into a 96-well microtitre plate. The plate was sealed and transferred directly to the autosampler for analysis. Liquid chromatography-tandem mass spectrometry Chromatography was performed using a Shimadzu HPLC system comprising a SIL-HT sampler and a degasser with two 1-AD pumps (Shimadzu, Milton Keynes, UK). Extract (5 ml) was directly injected from the microtitre plate onto a Security Guard C18 4mm 2 mm column, connected to a Synergi 4 mm Hydro-RP column (5 mm 3 mm), both from Phenomenex (Maccles eld, UK). The column was maintained at ambient temperature. Following elution from the column, the sample was pumped directly to the electrospray probe of the mass spectrometer, with no splitting or solvent diversion. The mobile phase initially consisted of 7% B for 2 min, allowing isocratic elution. The column was then washed with 95% B for.5 min, then re-equilibrated with starting conditions for 1min. The ow rate was maintained at.6 ml/min. The total run time was 4.3 min injection to injection. A Quattro Micro tandem mass spectrometer (Waters, Manchester, UK) tted with a Z spray ion source was used for all analyses. The instrument was operated in electrospray positive ionization mode and was directly coupled to the HPLC system. System control and data acquisition were performed with MassLynx NT 4. software with automated data processing, using the MassLynx QuanLynx Program provided with the mass spectrometer. Calibration curves were constructed using least-squares regression with 1/x weighting. To tune the mass spectrometer, solutions of androstenedione and d 7 A (1mg/L in methanol) and testosterone and d 2 T (1mg/L in methanol) were infused into the ion source. The cone voltage was optimized to maximize the intensity of the precursor ions for androstenedione, d 7 A, testosterone and d 2 T, seen at m/z 287.3, m/z 294.4, m/z 289.3 and m/z 291.3, respectively. The collision energy was then adjusted to optimize the signal for the most abundant product ions, seen at m/z 96.8, m/z 99.8, m/z 19. and m/z 98.9, respectively. Tuning conditions for androstenedione and d 7 A were as follows: electrospray capillary voltage.9 kv, sample cone voltage 26 V, and collision energies 22 and 26 ev, respectively, at a collision gas pressure 3.62 1 3 mbar argon. For testosterone and d 2 T, tuning conditions were as follows: electrospray capillary voltage.9 kv, sample cone voltage 28 V and collision energy 28 ev, at a collision gas pressure 3.62 1 3 mbar argon. Desolvation gas ow and temperature were maintained at 6 L/h and 371C, respectively; source temperature was 141C. Transitions were monitored in multiple reaction monitoring (MRM) mode with a dwell time of.2 s. Validation The assay was validated against published acceptance criteria for linearity, precision, recovery and sample stability. 12 Linearity To evaluate linearity of the calibration curves, three curves were prepared and analysed in a single batch.
5 Gallagher et al. The ratios of analyte peak height to internal standard peak height were plotted against androgen concentration in nmol/l. Calibration curves were judged linear if the correlation coe cient (r 2 )wasbetterthan.99as calculated by weighted linear regression. Six serum samples were subject to serial dilution with PBS containing.1% BSA to assess assay linearity. Imprecision Imprecision of the method was assessed against a range of concentrations using QC samples. These samples were analysed daily for 15 days to calculate interassay imprecision. To determine intra-assay imprecision, the same samples were analysed 12 times within one batch. Percentage deviation was calculated from the di erence between mean observed and nominal concentrations to assess bias. Limit of quantitation The lower limit of quantitation (LLOQ) was de ned as the concentration for which 1 replicates of PBS-based samples, prepared with low concentrations of androstenedione and testosterone, gave a coe cient of variation (CV) of less than 2% and bias of less than 2%. Recovery The recovery was determined by comparing the concentration of each androgen in patient samples, before and after the addition of a known amount of the androgen. We took six serum samples with androstenedione and testosterone concentrations ranging from.8 to 4.5 nmol/l and from.6 to 18 nmol/l, respectively. The concentrations of androstenedione and testosterone added to each sample were, 7.5, 15 and 3 nmol/l, and,1, 2 and 4 nmol/l, respectively. Sample stability We assessed stability by carrying out: (1) analysis of extract before and after 24 h storage at 41C, (2) analysis of a single extract injected every 8 min over 15 h, (3) analysis of serum samples after one to ve freeze-- thaw cycles. Samples were judged to be stable if the change in response was less than 1%. Ion suppression Ion suppression is a matrix e ect, which occurs when compounds in a sample compete with the analyte for ionization in the source. To investigate this, we infused the following solutions directly into the mass spectrometer to give a constant background signal: 34 nmol/l deuterated testosterone in methanol and 34 nmol/l of deuterated androstenedione in methanol. An extracted serum sample was injected simultaneously via the autosampler. A reduction in the background signal is observed when ion suppression is occurring. The ion suppression is deemed signi cant if a reduction in signal of 41% is observed where the compound of interest elutes. 13 Specificity Solutions of various related steroids, both natural and synthetic, were prepared in 5% methanol/water to a nal concentration of 1 nmol/l and injected directly, without extraction. This excess concentration was chosen to allow easy identi cation of any potentially interfering peaks. Any compounds found to give a signal in the speci c channels for testosterone, androstenedione or their deuterated counterparts were subsequently prepared to the same concentration in PBS containing.1% BSA and extracted along with a full standard curve and QCs to allow quantitation. Steroids tested were levonorgestrel, 19 nortestosterone, estrone, dehydroepiandrosterone, oestradiol, epitestosterone, cyproterone acetate, desogestrel, ethinyloestradiol and norethisterone. Accuracy To assess the accuracy of our method, we compared our assay with an established LC-MS/MS method 1 (n ¼ 28) and three pools provided by the United Kingdom National QualityAssessment Scheme (UKNEQAS) with GC-MS assigned values (n ¼ 4) for testosterone. Comparative assays We compared the results obtained by LC-MS/MS with those of a non-extracted RIA method for androstenedione (DSL-38 Active, Diagnostic Systems Laboratories Inc., TX, USA) and those of the Roche E17 electrochemiluminescence immunoassay (ECLIA) for testosterone (Roche Diagnostics, Lewes, UK). Samples from patients were stored at 31C until analysis, rst by these methods and then by LC-MS/MS. The samples taken for analysis for androstenedione and testosterone were subject to local ethical approval. Androstenedione concentrations were measured in 46 female and 46 male anonymized sera by RIA and LC-MS/MS. Testosterone concentrations were measured in 77 female and 52 male anonymized sera by both ECLIA and LC-MS/MS.We then assessed the accuracy of the ECLIA method by analysing low concentra-
Simultaneous determination of androstenedione and testosterone 51 tion (.3 and 1. nmol/l) QC samples 1 times. These QCs were made up in PBS containing.1% BSA to reduce any matrix e ects and possible crossreactivities. Although the LC-MS/MS method is aimed at the routine measurement of female samples, male samples were included in the comparison to allow us to assess the linear range of the assay. Including male samples in the comparison would also allow us to use this method for any male samples that gave anomalous testosterone results by immunoassay. Statistical analysis All statistical analyses were performed using Analyseit software (Analyse-it Software Ltd, Leeds, UK). Results Validation Using the chromatographic conditions above, androstenedione and testosterone co-eluted with their deuterated counterparts. Androstenedione and d 7 A, and testosterone and d 2 T were found to have retention times of 1.8 and 2.3 min, respectively. There were no interfering peaks (Figure 1). Figure 2 shows the result of investigation for ion suppression. The peak for androstenedione is shown in Figure 2b, eluting at 1.8 min. Figure 2a represents the constant infusion of deuterated androstenedione, showing that the main ion suppression occurs at.6 and 3. min. The peak for testosterone (Figure 2d) is shown to elute at 2.3 min. The main ion suppression occurs at.5--1. min and 3. min (Figure 2c). Both androgens are seen to elute away from the main areas of ion suppression. The standard curve was linear to 5 nmol/l for both androgens, with a correlation coe cient (r 2 )of.999 (n ¼ 3) (Figure 3). Serial dilution of six serum samples demonstrated dilutional linearity (Table 1). The lower limit of quanti cation was.25 nmol/l for both androgens. At this concentration, the %CVand bias of1 replicates were 1.9% and 11.5%, respectively, for androstenedione and 11.9% and 7.5%, respectively, for testosterone. Concentrations below.25 nmol/l result in %CVs greater than the acceptable limit of 2%. Both inter- and intra-assay imprecision were below 1% CV for androstenedione and testosterone at all concentrations, with less than 15% bias for all samples (Table 2). The mean recovery of androstenedione was 99% (range 93--14%) and for testosterone 93% (range 89--95%). In all experiments for stability, there was a decrease in response of less than 5.5% in measured concentrations for both androgens. There was no loss of sensitivity over repeat injection for 15 h; the extract wasstablefor24hat41c and samples were stable for ve freeze--thaw cycles. None of the steroids tested for speci city gave a signi cant signal in the speci c MRM channels for androstenedione, testosterone or the internal standards. Figure 1 A patient sample showing a typical chromatogram for each androgen and their deuterated form, at concentrations of 2.1 and 1.2 nmol /L for androstenedione and testosterone, respectively. (a) d 7 androstenedione, (b) d 2 testosterone, (c) testosterone and (d) androstenedione
52 Gallagher et al. a 294.4 > 99.9 1.8e3 b 287.3 > 97 1.12e4 Time..5 1. 1.5 2. 2.5 3. 3.5 c 291.3 > 99 1.85e3 289.3 > 19 6.59e3 d..5 1. 1.5 2. 2.5 3. 3.5 Figure 2 Ion suppression. (a) Represents the constant infusion of deuterated androstenedione, while (b) is a serum sample showing a peak where androstenedione elutes. (c) Represents the constant infusion of deuterated testosterone, while (d) is a serum sample showing a peak where testosterone elutes Bland--Altman analysis of female samples analysed for testosterone by this method and an established LC- MS/MS method in routine use in the SAS Centre for Steroid Hormones (Leeds), showed good agreement (Figure 4). Our method demonstrated a small negative bias of.5 nmol/l. Analysis of UKNEQAS samples also demonstrated good agreement for all four replicates for each of the three pools (Table 3). Comparison with other methods Androstenedione When results obtained by LC-MS/MS were compared with those by RIA, Passing and Bablok regression analysis showed LC-MS/MS (nmol/l) ¼.58 RIA (nmol/l).21, r 2 ¼.74. Bland--Altman agreement analysis showed the RIA to have an overall bias of 1.97 nmol/l
A Androstenedione/d 7 - androstenedione peak height B 3 25 2 15 1 5 1 2 3 4 5 6 Androstenedione concentration (nmol/l) Simultaneous determination of androstenedione and testosterone 53 Table 2 Intra-assay imprecision for androstenedione (a) and testosterone (b), and inter-assay imprecision for androstenedione (c) and testosterone (d) Bias (%) Imprecision CV (%) (a) Androstenedione nominal concentration (nmol/l).35 14.3 6.1 3.5 2 2.5 35 3 (b) Testosterone nominal concentration (nmol/l).3 9 6.9 3. 8.2 2.5 29 3.4 2.6 Testosterone/d 2 - testosterone peak height 14 12 1 8 6 4 2 1 2 3 4 5 6 Testosterone concentration (nmol/l) Figure 3 Typical standard curves for (a) androstenedione and (b) testosterone Table 1 Testosterone (a) and androstenedione (b) concentrations in serum samples diluted 1 in 2, 1 in 5 and 1 in 1 with PBS containing.1% BSA Sample number Undiluted 1 in 2 1 in 5 1 in 1 (a) Testosterone (nmol/l) 1 14.2 12.7 15. 14.2 2 18.5 18.3 18.1 18.1 3 15.8 16. 16.2 15. 4 5.6 5.5 4.8 4.6 5 21.8 22.1 22.6 2.4 6 12.2 12.2 11.7 11.2 (b) Androstenedione (nmol/l) 1 6.6 6.5 6.7 6.4 2 6.3 6.5 6.5 6.2 3 5.6 5.4 5.5 5 4 9.6 9.6 9.5 9.4 5 8.7 8.8 8.6 8.3 6 5.1 5.5 5.1 4.7 (c) Androstenedione nominal concentration (nmol/l).35 14.3 6.3 3.5 1.7 3.2 35 2. 4.8 (d) Testosterone nominal concentration (nmol/l).3 13 9.8 3. 9.7 3.8 29 1.9 3.2 n=12 LC-MS/MS comparative assay.2.15.1.5.5.1.15.2.25.3 (Figure 5a). However a concentration-dependent bias was observed. At androstenedione concentrations of approximately 1nmol/L, the bias was close to zero and it increased with increasing androstenedione concentrations up to a bias of 6 nmol/l at approximately 1 nmol/l androstenedione. Testosterone 1 2 3 Mean of all methods Zero bias Figure 4 Bland Altman analysis comparing this method with an established LC-MS/MS assay for testosterone only When results obtained by LC-MS/MS were compared with those by ECLIA, Passing and Bablok regression analysis showed LC-MS/MS (nmol/l) ¼.98 ECLIA (nmol/l).5, r 2 ¼.97 and Bland--Altman analysis revealed a bias of.22 nmol/l (Figure 5b).
54 Gallagher et al. Table 3 Testosterone concentrations (nmol/l) measured by LC-MS/MS of UKNEQAS samples with GC-MS assigned values Pool GC-MS target LC-MS/MS individual values LC-MS/MS mean 31A.7361.66;.67;.61;.6.64 87 31B 1.225 1.23;1.14;1.28;1.23 1.22 1 31C 2.72 2.9;2.81;2.82;2.89 2.86 15 LC-MS/MS as a percentage of GC-MS target Table 4 Table showing greater variability at low concentrations when measured by ECLIA (Roche) than by LC-MS/MS.3 nmol/l 1. nmol/l Testosterone concentration LC-MS/MS ECLIA LC-MS/MS ECLIA Mean.3.16.99.86 SD.2.1.4.3 CV (%) 6.9 8.39 4 3.97 Bias (%) 47 1 14 On examining results from female samples alone (n ¼ 77), we found the correlation (r 2 )was.25.the regression equation was LC-MS/MS (nmol/ L) ¼.74 ECLIA (nmol/l) þ.1. Bland--Altman analysis revealed a bias of.24 nmol/l, i.e., concentrations obtained by ECLIA were higher than those obtained by LC-MS/MS (Figure 5c). One signi cant outlier was observed in which the ECLIA testosterone concentration was almost 6 nmol/l higher than the LC- MS/MS. Repeat analysis of PBS-based samples using the LC-MS/MS and automated immunoassay for testosterone showed that the ECLIA underestimated testosterone considerably in the female range (Table 4). Discussion The chromatograms that are produced with this method are clear with no interfering peaks seen near the relevant retention times. Despite the use of liquid-- liquid extraction, which theoretically should remove many substances that may result in ion suppression, regions of ion suppression are observed. The chromatography has been optimized to ensure that both testosterone and androstenedione elute away from these areas, and therefore negligible ion suppression is seen. The assay s LLOQ and good linearity at high concentrations would make the method, particularly for testosterone, applicable to both female and male samples. The high working range will also cover both the normal and pathological concentrations; therefore the need for sample dilution is extremely rare. Stability studies showed that both androstenedione and testosterone are stable for 24 h when extracted and stored in the fridge. There was no systematic loss in sensitivity upon repeat injection over 15 h, demonstrating that long assays may be performed without a decline in performance. It was also observed that both testosterone and androstenedione were stable for at least ve freeze--thaw cycles. None of the steroids tested showed signi cant interference even at high concentrations. However, the authors acknowledge that only a limited number of steroids were tested. Metabolites of steroids or drugs may possibly cause interfering peaks, but as these are not commercially available, it would be impractical to test these and the vast number of other steroids that are available. By comparing our method with both LC-MS/MS and GC-MS (UKNEQAS samples), we are con dent that our assay gives accurate results for testosterone. Unfortunately, it was not possible to conduct such experiments for androstenedione as GC-MS target pools and routine LC-MS/MS assays are not, to our knowledge, available for this analyte. Although we appreciate that comparing an extracted LC-MS/MS assay to non-extracted immunoassays is not comparing like with like, our intention was to compare the results obtained by our new assay (intended for routine use) with the methods that were in routine use in our laboratory at the time. These comparisons showed that much lower results were obtained by the LC-MS/MS method for androstenedione. This is perhaps due to the speci city of the LC-MS/MS assay because the calibrators from the RIA kit agreed closely with those of the LC-MS/MS assay.
Simultaneous determination of androstenedione and testosterone 55 A 1 1 LC-MS/MS RIA 3 5 7 5 1 Mean of all methods Zero Bias Testosterone results by both methods compared well with a high correlation coe cient and low bias (.22 nmol/l); however, female samples alone showed a much poorer correlation and a slightly larger bias. As we are con dent of the LC-MS/MS method s accuracy and precision at these low concentrations, we can conclude that the automated immunoassay s performance is questionable in this range. The nding of a high bias by the immunoassay from the nominal testosterone concentration in an arti cial matrix supports this. This method has been used routinely for female samples in our laboratory for 13 months, measuring in excess of 12 samples in this time. The method is also often used to measure testosterone in male serum, including samples from other hospitals in the region, when the results of immunoassay are under question. B LC-MS/MS ECLIA 6 4 2 2 4 Zero bias Summary We have developed a simple, sensitive, robust, accurate and precise LC-MS/MS method for simultaneous determination of androstenedione and testosterone. Using a preparatory liquid--liquid extraction step, combined with adequate chromatography, this method separates androstenedione and testosterone from possible interfering substances, thereby removing inherent problems seen with direct immunoassay. This method will prove useful in clinical practice. C 6 8 1 2 3 Mean of all methods Acknowledgements We acknowledge Dr H Field, Leeds SAS Centre for Steroid Hormones, for kindly analysing a series of samples for this comparison and DrJ Middle, UKNEQAS, for providing specimens for analysis. 2 References LC-MS/MS Roche 1.5 1.5.5 1 1.5 2 2.5 3 1 2 3 4 5 Mean of all methods Zero bias 1 Fox R, Corrigan E, Thomas PA, Hull MG. The diagnosis of polycystic ovaries in women with oligo-amenorrhoea: predictive power of endocrine tests. Clin Endocrinol (Oxf) 1991; 34: 127 31 2 Robinson S, Rodin DA, Deacon A, Wheeler MJ, Clayton RN. Which hormone tests for the diagnosis of polycystic ovary syndrome? Br J Obstet Gynaecol 1992; 99: 232 8 3 O Driscoll JB, Mamtora H, Higginson J, Pollock A, Kane J, Anderson DC. A prospective study of the prevalence of clear-cut Figure 5 Bland Altman difference plots of: (a) Androstenedione concentrations for all male and female samples measured by LC-MS/MS and RIA. (b) Testosterone concentrations measured by LC-MS/MS and ECLIA for all male and female samples. (c) Female testosterone samples alone measured by LC-MS/MS and ECLIA. Dotted lines indicate two standard deviations (2 SD) from the mean
56 Gallagher et al. endocrine disorders and polycystic ovaries in 35 patients presenting with hirsutism or androgenic alopecia. Clin Endocrinol 1991; 41: 231 6 4 Tietz NW. Fundamentals of Clinical Chemistry. 3rd edn. Philadelphia: WB Saunders, 1996; 1553 5 Wudy SA, Wachter UA, Homoki J, Teller WM. 17 Alphahydoxyprogesterone, 4-androstenedione, and testosterone profiled by routine stable isotope dilution/gas chromatography-mass spectrometry in plasma of children. Pediatr Res 1995; 38: 76 8 6 Dorgan JF, Fears TR, McMahon RP, Aronson Friedman L, Patterson BH, Greenhut SF. Measurement of steroid sex hormones in serum: a comparison of radioimmunoassay and mass spectrometry. Steroids 22; 67: 151 8 7 Wheeler MJ, Lowy C. Warning on serum testosterone measurement. Lancet 1987; 2: 514 5 8 Boots LR, Potter S, Potter D, Azziz R. Measurement of total serum testosterone levels using commercially available kits: high degree of between-kit variability. Fertil Steril 1998; 69: 286 92 9 Taieb J, Mathian B, Millot F, et al. Testosterone measured by 1 immunoassays and by isotope dilution gas chromatography mass spectrometry in sera from 116 men, women, and children. Clin Chem 23; 49: 1381 95 1 Cawood ML, Field HP, Ford CG, et al. Testosterone measurement by isotope-dilution liquid chromatography-tandem mass spectrometry: validation of a method for routine clinical practice. Clin Chem 25; 51: 1472 9 11 Wang C, Catlin DH, Demers LM, Starcevic B, Swerloff S. Measurement of total serum testosterone in adult men: comparison of current laboratory methods versus liquid chromatography-tandem mass spectrometry. J Clin Endocrinol Metab 24; 89: 534 43 12 Health and Human Services, US. Department of Food and Drug Administration, Center for Drug Evaluation and Research (CDER), Center for Veterinary Medicine (CVM). Guidance for Industry Bioanalytical Method Validation. Rockville. MD: FDA/Center for Drug Evaluation and Research, May 21; 1 22 13 Annesley TM. Ion suppression in mass spectrometry. Clin Chem 23; 49: 141 4 Accepted for publication 7 September 26