Androgens During the Reproductive Years: What Is Normal for Women?

Marina A. Skiba; Robin J. Bell; Rakibul M. Islam; David J. Handelsman; Reena Desai; Susan R. Davis


J Clin Endocrinol Metab. 2019;104(11):5382-5392. 

In This Article


Study Participants

The study participants were women aged 18 to 39 years, who had been recruited from the eastern Australian states of Victoria, New South Wales, and Queensland to the Grollo-Ruzzene Young Women's Health Study.[10] Recruitment to the study has been previously described in detail.[10] In brief, 6986 women were recruited from two large, community-based, continuously refreshed databases from November 11, 2016, to July 21, 2017. The national representativeness of the study sample was confirmed by comparison with the Australian Bureau of Statistics census data for Australian women of the same age range.[10] All study participants completed a comprehensive online questionnaire that included extensive socio-demographic and health information and self-reported height and weight, medication use, and a previous diagnosis of polycystic ovary syndrome (PCOS).

On completion of the study questionnaire, the 3689 women who were not using systemic hormones, currently or recently (within 3 months) pregnant or lactating, and not postmenopausal were invited to provide a blood sample. Sonic Healthcare, the largest pathology provider in Australia, has a large network of collection centers across the included states, and the women were directed to their closest site. Other than an elevated progesterone level as a strong indicator of ovulation, a single hormone level in isolation is of little value in classifying a woman as at a certain menstrual cycle stage. We used a stepwise process to classify women using their self-reported last menstrual period (LMP) and their hormone profile.

When blood samples were taken, each participant was asked to report the number of days since the start of their LMP. This enabled us to classify women as being in the "early follicular" phase for cycle days 1 to 7, "midcycle" phase for cycle days 8 to 21, or "luteal" phase for cycle days 22 to 35, or as ">35 days since the LMP or unknown." We then tabulated the full hormone profile for every woman, including gonadotrophins, with progesterone and estradiol measured using LC-MS/MS. We then reviewed the hormones and the reported cycle stage for every individual and confirmed the cycle classification. A progesterone level of >12 nmol/L was used to distinguish the luteal phase from the midcycle phase. Therefore, women who reported being at day 8 to 21 but who had a high progesterone level were classified as being in the luteal phase. However, because the late luteal phase is characterized by a decrease in progesterone, women who reported being in cycle day 22 to 35 were not reclassified because they might have been about to menstruate. We defined a "reference group" of women from within the study population to establish normative sex steroid values by the menstrual cycle phase, BMI, and age. Participants were excluded from the reference group if any of the following applied at the time of their blood sampling: they had become pregnant; had begun taking systemic sex steroids, glucocorticoids, or antiandrogen therapy; had reported a diagnosis of PCOS with ultrasound confirmation; had not menstruated within the preceding 35 days; or were amenorrheic with a reproductive hormone profile indicating premature ovarian insufficiency. The Monash University Human Research Ethics Committee approved the present study (approval no. CF16/2322-2016001166 [7703]). The online invitation to participate included a link to the study explanatory statement and the study consent form. Women were required to select the "agree"option to progress to the questionnaire.

Sample Size

The total sample size for the Grollo-Ruzzene Young Women's Health Study was based on the precision of the estimate of the prevalence of PCOS.[10] The final number of participants in the present study was determined by the number of women invited to provide a blood sample who had then attended a sample collection center.

Biochemical Analysis

Serum aliquots were stored at −80°C until analysis. Sex steroids were measured in a single sample of serum using LCMS/MS at the ANZAC Research Institute (University of Sydney, Sydney, News South Wales, Australia).

All steroid standards and internal standards (ISs) were obtained from the National Measurement Institute (Sydney, New South Wales, Australia), except for 11KT, 11KDHT, 11KA, and d4-E1 (Steraloids, Inc., Newport, RI), E2, E1, and progesterone (P4) (Cerillant Corp., Round Rock, TX), d9-P4 (CDN Isotopes, Point-Claire, QC, Canada), d4-E2 (Cambridge Isotope Laboratories, Tewksbury, MA), d3-11KT (RTI International, Research Triangle Park, NC), and d3-11KDHT (Sapphire Biosciences, Redfern, Australia). All ISs were deuterated (d3) isotopes, except for d2-DHEA, d4-E2, d4-E1, and d9-P4. For 11KA, in the absence of any commercially available IS, d3-11KT was used instead. Where available, certified reference materials were used as assay standards from the National Measurement Institute (testosterone, DHT, DHEA, androstenediol, P4) or Cerillant (E2, E1). Aliquots (200 μL) of serum, standards, and quality control samples in 5-mL glass tubes were fortified with 50 μL of deuterated steroid IS, well mixed, and left at 4°C for 15 minutes. Extraction was performed by adding 1 mL of methyl tert-butyl ether, followed by vigorous mixing for 1 minute to extract steroids into the organic layer. The tubes were then covered and allowed to phase separate at 4°C for 1 hour, followed by freezing of the lower aqueous layer at −80°C for 30 minutes. The upper organic layer was decanted into clean glass tubes, and the solvent allowed to evaporate at 37°C in a fume hood overnight. The dried extracts were resuspended in 75 μL of 20% methanol/water and mixed for 1 minute before 50 μL of extract was transferred into a 96-well microtiter plate for injection into the LC-MS/MS/MS equipment. The ultra-pressure LC conditions included elution of steroids using a methanol/water gradient from a Kinetex Phenyl Hexyl column (100 mm × 2.1 mm × 1.7 μm; part no. 00D-4500-AN) with a Phenomenex guard cartridge (ultra-high performance LC phenyl for 2.1-mm ID columns, part no. AJ0-8788, Security Guard ULTRA Cartridges; Phenomenex, Lane Cove, New South Wales, Australia) at a column temperature of 45°C and flow rate of 0.35 mL/min. The eluant was introduced into the mass spectrometer without splitting. Chromatography provided baseline separation for each steroid, with a total run time of 13 minutes. The LC running conditions were split into three periods to match the ionization requirements of the steroids. The run times in the first period for the ketoandrogens (for the positive ionization mode) were 4.49 minutes for 11KT, 4.68 minutes for 11KA, and 5.14 minutes for 11KDHT. The run times in the second period for estrogens (for the negative ionization mode) included 5.74 minutes for E2 and 6.56 minutes for E1. The remaining steroids were run in the third period (for the positive ionization mode) and included testosterone (7.07 minutes), androstenedione (7.25 minutes), DHEA (7.39 minutes), DHT (8.73 minutes), and P4 (10.88 minutes).

MS was performed using an API-5000 triple-quadrupole mass spectrometer (Applied Biosystems, Foster City, CA; or MDS SCIEX, Concord, ON) equipped with an atmospheric pressure photoionization source that operated in both positive and negative ion modes (probe settings, X = 5, Y = 10, Z = 7). The atmospheric pressure photoionization system consists of a 10 eV krypton discharge lamp for photoionization, with dopant (toluene) delivery set to 75 μL/min. Nitrogen was used as the curtain, nebulizer, collision, and lamp gases. Multiple reaction monitoring with qualifier and quantifier transitions was used to quantify the steroids. The settings for the various transitions were optimized by infusing pure standard material into the mass spectrometer. Unit mass resolution was set in both mass-resolving quadrupole Q1 and Q3. A dwell time of 200 ms was used for the target analytes and their corresponding isotopically labeled (deuterated) IS had a dwell time of 100 ms, except for the ketoandrogens, for which the dwell time was 100 ms for both the standard and the IS. The quantifier transitions and collision energies (eV) were testosterone (289 to >109 and 35, respectively), DHT (273 to >123 and 31, respectively), E2 (271 to >145 and −57, respectively), E1 (269 to >145 and −57, respectively), androstenedione (287 to >97 and 34, respectively), P4 (315 to >97 and 34, respectively), DHEA (253 to>197 and 28, respectively), 11KT (306 to>123 and 40, respectively), 11KDHT (308 to >190 and 26, respectively), and 11KA (301.5 to>121 and 30, respectively). For all analytes, the external recovery was 88% to 96%, and the matrix effect was 85% to 93%.

The limits of detection (LODs), limits of quantitation, and between-run (15 replicates) and within-run (10 replicates) reproducibility (range of coefficient of variation percentage for three quality control levels) for each analyte were as follows: testosterone, 0.03 and 0.09 nmol/L, 2% to 9%, and 4% to 8%; DHT, 0.17 and 0.34 nmol/L, 4% to 12%, and 4% to 9%; E2, 9.18 and 18.36 pmol/L, 4% to 8%, and 5% to 8%), E1, 9.24 and 18.49 pmol/L, 3% to 6%, and 5% to 9%); androstenedione, 0.025 and 0.05 ng/mL, 4% to 10%, and 6% to 9%; P4, 0.06 and 0.16 nmol/L, 4% to 7%, and 4% to 9%; DHEA, 0.07 and 0.17 nmol/L, 3% to 6%, and 8% to 12%; 11KT, 0.07 and 0.17 nmol/L, 3% to 9%, and 4% to 8%; 11KDHT, 0.16 and 0.33 nmol/L, 5% to 7%, and 5% to 7%; and 11KA, 0.07 and 0.17 nmol/L, 3% to 8%, and 5% to 7%, respectively.

Statistical Analysis

DHT and 11KDHT concentrations were below the LOD in 104 (17.7%) and 436 (74.1%) samples respectively, and, therefore, were not reported. E2 was below the LOD in six samples (1.0%), E1 in one sample (0.2%), and 11KT in one sample (0.2%). A result below the LOD was not considered to indicate the absence of a hormone but, rather, that the hormone was not distinguishable from background noise. Therefore, for these steroids, any values below the LOD were allocated a value half way between the LOD and zero. The median and range of the serum levels for the C-19 steroids testosterone, 11KT, androstenedione, 11KA, and DHEA and the ratios between the sex steroids are reported.

The associations between the hormone levels and menstrual cycle stage, age, and BMI (kg/m2) are presented graphically as box and whisker plots. Fitted locally weighted scatterplot smoothing (LOWESS) curves are also shown for age. Further analyses were performed on the log-transformed data. The associations between cycle phase, age, and BMI and the hormone levels were explored using linear regression, with the results reported as β coefficients and 95% CIs. To compare the C19 steroid levels among all age groups (rather than only the youngest women compared with each of the other groups), factorial ANOVAs, which included BMI and cycle phase, with Bonferroni post hoc analyses, were performed. The analyses were performed using SPSS, version 24.0 (IBM Corp., Armonk, NY).