Androgen Profiling by Liquid Chromatography–Tandem Mass Spectrometry (LC-MS/MS) in Healthy Normal-Weight Ovulatory and Anovulatory Late Adolescent and Young Women

Androgen Profiling by LiquidChromatography–Tandem Mass Spectrometry (LC-MS/MS) in Healthy Normal-Weight Ovulatory andAnovulatory Late Adolescent and Young Women

Flaminia Fanelli, Alessandra Gambineri, Ilaria Belluomo, Andrea Repaci,Valentina Diana Di Lallo, Guido Di Dalmazi, Marco Mezzullo, Olga Prontera,Gaia Cuomo, Laura Zanotti, Alexandro Paccapelo, Antonio Maria Morselli-Labate,Uberto Pagotto, and Renato Pasquali

Endocrinology Unit (F.F., A.G., I.B., A.R., V.D.D.L., G.D.D., M.M., O.P., G.C., L.Z., U.P., R.P.), andDepartment of Medical and Surgical Sciences and Center for Applied Biomedical Sciences (F.F., A.G.,I.B., A.R., V.D.D.L., G.D.D., M.M., O.P., G.C., L.Z., A.P., A.M.M.-L., U.P., R.P.), St Orsola-MalpighiHospital, University Alma Mater Studiorum, 40138 Bologna, Italy

Context: Physiological transient imbalance typical of adolescence needs to be distinguished fromhyperandrogenism-related dysfunction. The accurate determination of circulating androgens isthe best indicator of hyperandrogenism. However, reliable reference intervals for adolescent andyoung women are not available.

Objective: The aim of the study was to define androgen reference intervals in young women andto analyze the impact of the menstrual phase and ovulation efficiency over the androgen profileas assessed by reliable liquid chromatography–tandem mass spectrometry (LC-MS/MS) technique.

Participants: Female high school students aged 16–19 years were included in the study.

Main Outcome Measures: The study was performed on reference subjects properly selected amongan unbiased population. Normal-weight, drug and disease free, eumenorrheic females with nosigns of hyperandrogenism were included. The steroid hormone profile was determined by avalidated in-house LC-MS/MS method. A statistical estimation of overall and menstrual phase-specific reference intervals was performed. A subgroup of anovulatory females was identifiedbased on progesterone circulating levels. The impact of ovulation efficiency over hormonal profilewas analyzed.

Results: A total of 159 females satisfied healthy criteria. Androgen levels did not vary according tomenstrual phase, but a significantly higher upper reference limit was found for T in the luteal phasecompared to the follicular phase. Higher T and androstenedione levels were observed in anovu-latory compared to ovulatory females, paralleled by higher LH and FSH and lower 17-hydroxy-progesterone and 17�-estradiol levels.

Conclusions: This is the first study providing LC-MS/MS-based, menstrual phase-specific referenceintervals for the circulating androgen profile in young females. We identified a subgroup ofanovulatory healthy females characterized by androgen imbalance. (J Clin Endocrinol Metab 98:3058–3067, 2013)

ISSN Print 0021-972X ISSN Online 1945-7197Printed in U.S.A.Copyright © 2013 by The Endocrine SocietyReceived February 12, 2013. Accepted May 3, 2013.First Published Online June 18, 2013

Abbreviations: BMI, body mass index; CI, confidence interval; DHEA, dehydroepiandros-terone; E2, 17�-estradiol; FAI, free androgen index; HDL, high-density lipoprotein; IR, ionratio; IS, internal standard; LC-MS/MS, liquid chromatography–tandem MS; mF-G, mod-ified Ferriman-Gallwey (score); MS, mass spectrometry; 17OHP, 17-hydroxyprogesterone;PCOS, polycystic ovary syndrome.

O R I G I N A L A R T I C L E

E n d o c r i n e R e s e a r c h

3058 jcem.endojournals.org J Clin Endocrinol Metab, July 2013, 98(7):3058–3067 doi: 10.1210/jc.2013-1381

The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 16 January 2015. at 02:25 For personal use only. No other uses without permission. . All rights reserved.

Hormone determination in biological fluids constitutesa relevant issue in endocrinology. Clinical practice

strongly relies on the information that can be drawn fromanalytical results and, in this regard, the availability ofage- and gender-specific reference intervals built on a de-fined ethnic and geographic population should be man-datory. When commercial assays are used, clinicians arebound to rely on reference intervals provided by thecompanies; however, little information is usually givenon the criteria used to estimate these normative values(1). Concerning steroid hormones, the situation is evenmore complicated because nonextractive automatedimmunoassays are often used. These assays are usuallycharacterized by poor specificity, limiting the accuracywhen low circulating levels have to be detected, aswidely reported in the last decade for major steroidssuch as T, 17-hydroxyprogesterone (17OHP), proges-terone, and 17�-estradiol (E2) (2– 6).

Notably, the evidence on the poor consistency of im-munoassays for the detection of low-level T led The En-docrine Society to recommend the use of assays based onsample extraction and mass spectrometry (MS) (7). Liquidchromatography–tandem MS (LC-MS/MS) combineshigh sensitivity and specificity, practicability allowing theapplication over large number of samples, and a high in-formative power deriving from the possibility of analyzinga hormone profile rather than single hormones (8). Ourgroup recently developed and validated an LC-MS/MSmethod for the measurement of 9 serum steroids, high-lighting the dramatic incoherence of immunoassay resultsat low circulating steroid levels. The method was also pro-posed in association with reference intervals in healthynormal-weight adult males and females (6). However,these reference intervals cannot be transferred to adoles-cence and young age because this period of life is charac-terized by transient but significant changes in the hor-monal profile. In addition, adolescence and youth are theperiods of female life in which functional hyperandrogenicstates usually arise, the most common of them being thepolycystic ovary syndrome (PCOS). Nevertheless, men-strual irregularity, oligoanovulation, or enlarged ovarieswith polycystic appearance may also reflect normal ado-lescence dynamics. Therefore, the measurement of serumandrogens probably provides the best method to discernthe normal evolution processes from hyperandrogenic dis-orders, but specific reference intervals have to be estimatedbecause there is a physiological increase in androgen levelsduring adolescence (9).

The aim of the study is to estimate the reference intervalfor circulating T, androstenedione, dehydroepiandros-terone (DHEA), and 17OHP as measured by LC-MS/MS,

and for related parameters such as the SHBG and the freeandrogen index (FAI) in adolescent and young adult fe-males selected from a large population participating in anepidemiological study focusing on the prevalence of hy-perandrogenic states (10). In addition, the impact of themenstrual phase and of the ovulation efficiency on andro-gen circulating levels was investigated.

Subjects and Methods

SubjectsItalian Caucasian females aged 16–19 years were selected

among 2052 girls and young adults contacted at the high schoolsof 2 provinces of the Emilia Romagna region (Bologna and Forlì-Cesena) as part of a cross-sectional epidemiological study focus-ing on the prevalence of hyperandrogenic states (10). The studywas approved by the local Ethics Committee. The flow chart ofthe selection and stratification of normal subjects is reported inFigure 1. After informed consent was given, 519 girls (25.3%)gave blood samples for laboratory testing and steroid measure-ment. A trained physician conducted a physical examination inwhich height, weight, waist and hip circumference, and bloodpressure were registered. Hirsutism was scored through the mod-ified Ferriman-Gallwey score (mF-G) (11), and the presence orabsence of signs of androgenic alopecia or acanthosis nigricanswas assessed (12–14). Dates of menses of the previous year wererecorded and used to define menstrual irregularity. Blood pres-sure was measured twice in the supine position after 3 minutes ofrest, and the mean of the 2 recorded values was used. Body massindex (BMI) was computed as the ratio between weight (kilo-grams) and height (meters) squared. Blood was withdrawn in thefasting condition between 8 and 10 AM in Vacuette Z serum beadsclot activator tubes (Greiner Bio-One, Kremsmunster, Austria),allowed to settle for 20 minutes, and centrifuged (2000 � g for10 min at room temperature). Serum was stored at �80°C.

Of the initial sample, 104 girls were excluded because theywere on active treatment with oral contraceptives, 21 because ofthyroid dysfunction or hyperprolactinemia, 147 because theyhad BMI and height outside age-specific reference ranges or awaist circumference above 88 cm, and 88 girls because menstrualirregularities or hirsutism were assessed.

Thyroid dysfunction or hyperprolactinemia was defined inthe presence of TSH or prolactin levels outside the laboratoryreference ranges (0.1–4 �IU/mL and 2–35 ng/mL, respectively).Age-specific normal BMI range was defined within the 10th(18.1 kg/m2) and 75th (23.2 kg/m2) percentiles of Italian BMIgrowth charts (15). Similarly, normal height range was definedwithin the third (1.50 m) and 97th (1.74 m) percentiles (15).Menstrual regularity was defined for girls with a gynecologicalage above 2 years, reporting 11–13 cycles in the last year. Thenormal mF-G score was defined as � 8 (11, 16). The remaining159 subjects displayed no signs of androgenic alopecia or acan-thosis nigricans, were not taking any drug, and were not affectedby endocrine, hepatic, renal, tumoral, autoimmune, cardiovas-cular, hematological, neurological, or psychiatric diseases, sleepdisorders, or allergies requiring treatment.

The date of the last menstrual bleeding and the blood con-centrations of E2, LH, and FSH supported the definition of sub-

doi: 10.1210/jc.2013-1381 jcem.endojournals.org 3059


groups of subjects in the follicular and luteal phases. We definedthe follicular phase as between days 1 and 13, coinciding withlevels of E2 � 100 pg/mL, LH � 20 �IU/mL, and FSH � 10�IU/mL. The luteal phase was defined from day 16 on, coincid-ing with levels of LH � 20 �IU/mL and FSH � 10 �IU/mL.Accordingly, we obtained 53 subjects in the follicular phase and51 subjects in the luteal phase. No menstrual phase was assignedto the remaining 55 girls because the date of the last bleeding wasnot provided (22 subjects) or because the above stratificationcriteria were not satisfied (33 subjects).

In a previous study, we demonstrated the poor performanceof the immunoassays used in our Central Laboratory for low-level progesterone measurement (6); therefore, for the assess-ment of ovulation efficiency, we relied on LC-MS/MS for pro-gesterone determination. Because there is no accepted LC-MS/MS-based cutoff for progesterone levels below whichanovulation occurs, we applied a cluster analysis by using thecirculating progesterone levels in our luteal phase subjects inorder to define 2 groups at different levels of ovulation efficiency.

Metabolic and hormonal evaluationSerum hormones and metabolites (intra- and interassay co-

efficients of variation) were measured at the Central Laboratoryof S. Orsola-Malpighi Hospital (Bologna, Italy): glucose (1.1 and1.9%), triglycerides (�1.5% and 1.8%), total cholesterol(�1.0% and 2.7%), high-density lipoprotein (HDL) cholesterol(�0.95% and 1.3%), insulin (1.5 and 4.9%), TSH (3.0 and7.2%), prolactin (1.7 and 2.0%), LH (1.2 and 2.2%), FSH (2.8and 4.5%), and E2 (3.3 and 4.9%) were measured by ModularAnalytics E170 (Roche Diagnostics, Mannheim, Germany).SHBG (4.2 and 6.6%) was measured by Immulite 2000 (SiemensHealthcare Diagnostics, Deerfield, Illinois). FAI was calculatedas: T (ng/mL) � 3.47 � 100/SHBG (nmol/L) (17).

Steroid measurement by LC-MS/MSTestosterone, androstenedione, DHEA, 17OHP, and proges-

terone were measured by isotopic dilution-LC-MS/MS as de-tailed elsewhere (6). After protein precipitation of 0.9 ml of se-rum and addition of internal standards (IS), followed by solid-

phase extraction, samples were injectedinto a 2-dimension chromatographic sys-tem, purified by a perfusion column, andseparated on a Luna RP-C8 100 � 4.6mm, 5 �m (Phenomenex, Torrance, Cal-ifornia) in a 21-minute gradient run.Analytes then underwent atmosphericpressure chemical ionization and multi-ple reaction monitoring detection by anAPI4000 QTrap mass spectrometer (AB-Sciex, Toronto, Canada). Quantifier andqualifier precursor ion/fragment ionmass-to-charge transitions for T, andro-stenedione, DHEA, 17OHP, and proges-terone were: 289.2/97.1 and 289.2/109.1, 287.4/97.0 and 287.4/109.0,271.3/197.2 and 271.3/213.3, 331.1/97.0 and 331.1/109.3, and 315.6/97.1and 315.6/109.1, respectively. Ion ratios(IRs) between quantifier and qualifiertransitions were monitored in each sam-ple; data were excluded whenever the IRvalue exceeded � 20% of the expected

IR. 13C2-Testosterone (transition, 291.4/111.1; Cambridge Iso-tope Laboratories, Andover, Massachusetts) was used as IS forT, androstenedione, and DHEA; d8–17OHP (transition, 339.5/100.1; CDN Isotopes, Pointe Claire, Canada) was used as the ISfor 17OHP and progesterone. Data processing and quantitationwere performed by Analyst 1.4.2 software package by AB-Sciex.Intra- and interassay coefficients of variation were � 10% and �11%, respectively, and accuracy ranged between 83.7 and104.4% for all the analytes. The sensitivity in the serum matrixwas 0.019, 0.039, 0.781, 0.078, and 0.049 ng/mL for T, andro-stenedione, DHEA, 17OHP, and progesterone, respectively.

Data analysis and statisticsThe normality of androgens, SHBG, and FAI (x) distributions

was verified by means of the Kolmogorov-Smirnov test (18). Allthe source variables displayed a significant skewness coefficient(P values ranging between P � .001 and P � .038); hence, inorder to obtain robust reference intervals, data were transformedaccording to the formula log10 (x � k). The numeric constant (k)minimized the skewness of the distribution of the variables in thewhole population: values of 0.25, 43.0, �0.12, 0.45, and �0.65were chosen for T, SHBG, FAI, androstenedione, and DHEA,respectively. Due to the high extent of fluctuation of 17OHPlevels during the menstrual cycle (19), different k values wereapplied to normalize 17OHP distribution in the whole group(�0.040) and in the follicular (�0.025) and luteal (4.00) sub-groups to produce robust reference intervals. The descriptiveparameters of the distribution, as calculated on source variablesand on transformed data, are reported in Supplemental Table 1(published on The Endocrine Society’s Journals Online web siteat http://jcem.endojournals.org). The test of normal distributionshowed values ranging from P � .001 and P � .849 and valuesranging from P � .431 and P � 1.000, before and after variabletransformation, respectively.

The reference limits were estimated as the mean � 1.96 SD

values of the transformed variables and were reported as backtransformed to the original unit. Follicular and luteal referencelimits were compared by using the z distribution.

2,052 were offered participation

519 underwent clinical examination

1,533 not evaluable:- 546 declined participation- 37 crucial items not compiled- 431 refused clinical examination and blood withdrawal- 519 refused blood withdrawal

519 underwent clinical examinationand blood withdrawal

360 were excluded:- 104 because on active treatment with oral contraceptives;- 21 because of thyroid dysfunction or hyperprolactinemia;- 147 because BMI and height were out of the age specific normal ranges

for the Italian population;- 88 because of menstrual irregularities or mF-G score >8

159 subjects admitted to the study

53 follicular phase

51 luteal phase:- 2 no progesterone data available- 18 anovulatory- 31 ovulatory

55 undefined menstrual phase

Figure 1. Flow chart of subject participation and normal subject selection.

3060 Fanelli et al Androgen Profile in Young Women J Clin Endocrinol Metab, July 2013, 98(7):3058–3067


The cluster analysis was applied on progesterone levels byusing Ward’s methods based on squared Euclidean distances(SPSS for Windows, version 13.0; SPSS Inc, Chicago, Illinois) inorder to differentiate ovulatory and anovulatory luteal sub-groups. The Mann-Whitney U test was used to compare follic-ular and luteal phases, as well as ovulatory and anovulatoryluteal subgroups. Correlation analyses were performed by Spear-man’s rank test. Data analyses were performed on MedCalc ver-sion 9.3.7.0 (MedCalc Software, Mariakerke, Belgium). Two-tailed P values less than .05 were considered significant.

Results

Anthropometric and metabolic features of thepopulation

The anthropometric and biochemical features of thereference females are reported as median and interquar-tile range (IQR; ie, the 25th and 75th percentiles) inTable 1. Girls had a median (IQR) age of 17.5 (16.5–18.1) years and a gynecological age of 4.91 (3.93–5.96)years and showed a normal metabolic profile and nor-mal blood pressure values. No significant differenceswere detected between menstrual subgroups, except forhigher levels of prolactin (P � .015) and E2 (P � .001),and lower levels of FSH (P � .001) observed in the lutealcompared to the follicular phase.

Androgen, SHBG, and FAI reference intervals inthe whole population, in follicular and in lutealphases

Box-and-whiskers plots of T, SHBG, FAI, androstene-dione, DHEA, and 17OHP distribution in the whole pop-ulation and in menstrual phase subgroups are shown inFigure 2. The levels of 17OHP were higher in the lutealcompared to the follicular phase (median [IQR], 1.365[0.679–1.909] ng/mL vs 0.379 [0.248–0.602] ng/mL;P � .001). Androgen levels did not vary according to men-strual phase; however, T showed a tendency to be higherin the luteal compared to the follicular phase (median[IQR], 0.289 [0.218–0.399] ng/mL vs 0.247 [0.197–0.327] ng/mL, respectively; P � .080).

Lower and upper reference limits, with respective 90%confidence intervals (CIs), calculated for androgens,SHBG, and FAI in the whole group and in the follicularand luteal phase are reported in Table 2. No significantdifferences were observed for lower reference limits esti-mated in the follicular and luteal phases. However, the Tluteal upper reference limit (0.557 ng/mL; 90% CI,0.494–0.625 ng/mL) was significantly higher than the fol-licular upper reference limit (0.438 ng/mL; 90% CI,0.398–0.482 ng/mL) (P � .010). Accordingly, the FAIupper reference limit tended to be higher in the luteal phase(6.751; 90% CI, 5.287–8.629) compared to follicular

Table 1. Clinical, Anthropometric, and Biochemical Parameters of the Whole Population and of the MenstrualPhase Subgroups

All Follicular Luteal P Valuea

n 159 53 51Age, y 17.5 (16.5–18.1) 17.2 (16.4–18.0) 17.5 (16.5–18.3) .596Age at menarche, y 13.0 (12.0–13.0) 12.5 (12.0–13.0) 13.0 (12.0–13.8) .460Gynecological age, y 4.91 (3.93–5.96) 4.73 (3.84–5.97) 4.58 (3.94–5.60) .734Menses/year 12.0 (12.0–12.0) 12.0 (12.0–12.0) 12.0 (12.0–12.0) .735mF-G score 4.00 (2.00–6.00) 4.00 (3.00–5.00) 3.00 (2.00–4.75) .103Weight, kg 55.0 (52.0–59.2) 56.0 (52.2–60.8) 55.5 (51.9–58.8) .415Height, m 1.64 (1.60–1.67) 1.66 (1.62–1.69) 1.63 (1.60–1.67) .063BMI, kg/m2 20.8 (19.5–21.9) 20.8 (19.6–21.9) 21.0 (19.7–22.0) .951Waist circumference, cm 73.0 (70.0–76.8) 73.0 (69.0–77.0) 74.0 (71.0–77.8) .409Hip circumference, cm 91.0 (88.0–94.0) 92.0 (88.0–95.0) 92.0 (89.0–94.8) .833Waist/hip ratio 0.811 (0.777–0.835) 0.802 (0.775–0.843) 0.818 (0.793–0.835) .294Diastolic blood pressure, mm Hg 70 (60–75) 70.0 (60–71) 70.0 (60–79) .718Systolic blood pressure, mm Hg 110 (110–120) 110.0 (110–120) 110.0 (106–120) .603TSH, �IU/mL 1.65 (1.25–2.12) 1.58 (1.26–2.11) 1.67 (1.35–2.39) .516Prolactin, ng/mL 12.0 (9.0–16.0) 10.0 (7.8–15.3) 12.0 (10.0–16.8) .015FSH, �IU/mL 5.00 (3.80–6.50) 6.10 (4.88–6.73) 3.60 (2.43–4.65) �.001LH, �IU/mL 5.70 (3.50–8.95) 5.00 (3.48–6.10) 4.50 (2.83–8.43) .728E2, pg/mL 80.0 (40.0–157.5) 41.0 (25.0–58.3) 117.0 (83.0–190.0) �.001Insulin, �IU/mL 7.20 (5.43–9.38) 7.10 (5.65–9.45) 6.60 (5.30–9.20) .633Glucose, mg/dL 79.0 (72.3–85.0) 78.0 (74.5–86.5) 76.0 (69.8–83.3) .143Total cholesterol, mg/dL 162.0 (142.0–180.8) 166.0 (146.8–181.3) 162.0 (139.3–180.0) .743HDL-cholesterol, mg/dL 59.0 (53.0–69.0) 61.0 (53.0–69.0) 55.0 (50.0–63.0) .073Triglycerides, mg/dL 59.0 (47.0–77.8) 57.0 (44.8–74.0) 64.0 (49.3–82.8) .129

Data are expressed as median (IQR).a Comparison between follicular and luteal phases.



phase (4.707; 95% CI, 3.904–5.682) (P � .054). SHBG,androstenedione, and DHEA upper reference limits didnot vary between the follicular and luteal phases.

Impact of the ovulation efficiency on androgencirculating levels

In 2 subjects in the luteal phase, the measurement ofprogesterone was classified as nonspecific because of al-tered IR. Hence, 49 of 51 subjects in the luteal phase wereevaluated. None of the 49 subjects in the luteal phaseshowed a level of progesterone between 2 and 3 ng/mL.The cluster analysis on progesterone distribution revealedthe presence of 2 different subgroups: 18 (36.7%) subjectsshowing progesterone levels between 0.062 and 1.958 ng/mL, and 31 (63.3%) subjects showing progesterone levelsbetween 3.125 and 17.970 ng/mL. Thus, we defined an-ovulation as occurring at levels of progesterone below 2

ng/mL, whereas subjects displayinglevels of progesterone above 3 ng/mLare presumably ovulatory.

Anovulatory and ovulatory fe-males showed no differences in age,anthropometric parameters, gyneco-logical age, in the mean cycle day ofthe luteal phase and in the mF-Gscore. Anovulatory females showedhigher FSH (P � .001) and LH (P �.029) and lower E2 (P � .001) levelscompared to ovulatory girls (Table3). The androgen profile of anovula-tory and ovulatory females in the lu-teal phase is represented in Figure 3.Data are reported as median [IQR].The anovulatory group had higherlevels of T (0.385 [0.273–0.444]ng/mL vs 0.253 [0.208–0.344] ng/mL; P � .006) and androstenedione(1.010 [0.775–1.396] ng/mL vs0.735 [0.568–1.037] ng/mL; P �.019), and lower levels of 17OHP(0.666 [0.475–0.979] ng/mL vs1.694 [1.414–2.121] ng/mL; P �.001) compared to ovulatory girls,respectively. Anovulatory femalesalso showed a tendency to higher val-ues of FAI (2.989 [1.576–3.927] vs1.927 [1.315–2.709]; P � .059) andDHEA (8.60 [6.29–11.22] ng/mL vs4.86 [4.08–9.47] ng/mL; P � .065)with respect to the ovulatory girls.Moreover,anovulatoryfemalesshowedhigher cholesterol (P � .029) and

HDL-cholesterol (P � .001) levels (Table 3). Accordingly,a correlation analysis revealed that, in the luteal phasegroup, HDL-cholesterol was negatively associated withprogesterone (rho � �0.314; P � .030) and positivelyassociated with T (rho � 0.337; P � .017), but not withFAI, SHBG, and other androgens, whereas total choles-terol showed a positive correlation with T (rho � 0.289;P � .046) and a tendency to negatively correlate withprogesterone (rho � �0.243; P � .092) levels.

Discussion

This work focused on the analysis of the androgen profilein a group of healthy, finely characterized women in lateadolescence by using a highly sensitive and specific tech-nique such as LC-MS/MS. We were able to estimate po-

Figure 2. Box-and-whiskers plots of T (A), SHBG (B), FAI (C), androstenedione (D), DHEA (E),and 17OHP (F) distribution in the whole female cohort and in menstrual phases. Central boxrepresents values ranging between the 25th and 75th percentiles, the middle line represents themedian, and the lower and upper whiskers extend from (25th percentile � 1.5 � IQR) to (75thpercentile � 1.5 � IQR), respectively.



tential reference intervals for androgens and for relatedindicators such as SHBG and FAI to be used for clinical aswell as for research purposes, also providing referencelimits specific for the follicular and luteal phases. More-over, the improved informative power of LC-MS/MS al-lowed a keen characterization of the hormonal profile ofgirls who were otherwise normal but who showed a re-duced ovulation efficiency.

Reference intervals are essential for the interpretationof laboratory data. Because LC-MS/MS has been pro-posed for replacement of immunoassays in clinical set-tings, it is mandatory for laboratories undergoing this in-novation to define new reference limits based on well-characterized populations (1). Several LC-MS/MSmethods for the determination of T, sometimes includingother androgens, were published in the last few years.However, few attempts have been made to generate ref-erence intervals (6, 20–24). Differences in the samplingconditions, in the sample size, and in the statistical ap-proach represent sources of variability among these stud-ies. In addition, the criteria used to define the level ofhealth of the reference population are highly variable and,very often, parameters known to influence sex hormone

levels, such as fat distribution or clinical indicators of hy-perandrogenism, are not taken into proper account. Fe-male adolescence implies additional difficulties in the def-inition of health due to the overlap between frequentphysiopathological states arising during the maturation ofthe ovulatory system and features associated with hy-perandrogenism, such as menstrual irregularities, acne,and metabolic imbalance. Nonetheless, the androgen cir-culating level is deemed to be the most reliable indicator todistinguish between transient manifestations and hy-perandrogenic disorders (9).

Our reference population consisted of a very homoge-neous group of 159 girls. Although such a sample size canbe seen as a limitation of the study, our reference popu-lation was derived from an unselected cohort of 2052 stu-dents thanks to the application of rigorous inclusion andexclusion criteria. Notably, a careful clinical assessmentwas conducted aimed at excluding any condition poten-tially affecting androgen levels or referable to PCOS, in-cluding clinical hyperandrogenism, menstrual irregulari-ties, and obesity (16, 25). Moreover, in order to guaranteea robust calculation of reference limits also in menstrualsubgroups, in which the sample size is smaller, we used a

Table 2. Androgen, SHBG, and FAI Low and High Reference Limits Estimated in the Whole Group and in Follicularand Luteal Phases

Lower Reference Limit (90% CI) Upper Reference Limit (90% CI)

Testosterone, ng/mLAll 0.130 (0.115–0.145) 0.506 (0.476–0.536)Follicular 0.124 (0.102–0.148) 0.438 (0.398–0.482)Luteal 0.114 (0.086–0.145) 0.557 (0.494–0.625)P (F vs L) 0.658 0.010

SHBG, nmol/LAll 19.3 (16.7–22.0) 88.3 (82.8–94.1)Follicular 19.5 (14.9–24.3) 88.7 (79.2–99.0)Luteal 18.9 (14.2–24.1) 91.5 (81.2–102.6)P (F vs L) 0.898 0.758

FAIAll 0.791 (0.712–0.879) 5.900 (5.224–6.666)Follicular 0.794 (0.676–0.938) 4.707 (3.904–5.682)Luteal 0.701 (0.573–0.866) 6.751 (5.287–8.629)P (F vs L) 0.437 0.054

Androstenedione, ng/mLAll 0.396 (0.352–0.442) 1.671 (1.561–1.786)Follicular 0.393 (0.323–0.469) 1.546 (1.381–1.727)Luteal 0.329 (0.251–0.417) 1.748 (1.527–1.995)P (F vs L) 0.347 0.247

DHEA, ng/mLAll 2.39 (2.15–2.66) 21.63 (18.82–24.88)Follicular 2.46 (2.05–2.99) 24.29 (18.91–31.26)Luteal 2.18 (1.82–2.65) 21.97 (16.92–28.56)P (F vs L) 0.462 0.650

17OHP, ng/mLAll 0.173 (0.151–0.199) 3.064 (2.565–3.663)Follicular 0.155 (0.132–0.184) 0.995 (0.818–1.211)Luteal 0.087 (-0.124–0.309) 2.853 (2.500–3.226)

Abbreviations: F, follicular; L, luteal. 17OHP limits in follicular and luteal phases cannot be statistically compared because the variable neededdifferent transformations by using a specific formula in each phase.



parametric approach based on data transformation, yield-ing an unskewed distribution of the variables.

Our data did not highlight differences in androgen,SHBG, or FAI levels according to the menstrual phase.Nonetheless, T tended to be higher in the luteal phasecompared to the follicular phase. Accordingly, the esti-mated T upper reference limit was found to be significantlyhigher in the luteal phase, and FAI showed a similar ten-dency. Data produced in the immunoassay era yieldedconflicting results about the fluctuation of androgen levelsbetween the luteal and follicular phases (26, 27); LC-MS/MS now provides improved sensitivity to definitivelyelucidate this issue. Rothman et al (28) assessed sex hor-mones by LC-MS/MS in the early follicular, midcycle, andmidluteal phases in 14 premenopausal adult women, ob-serving that T and estrogens peaked at midcycle and thatin the luteal phase they remained higher than the follicularphase. In our study, 2 different subsets of subjects in dif-ferent menstrual phases were compared, rather than in-traindividual fluctuations during the menstrual cycle;hence, definitive conclusions about this issue could not betaken. Nonetheless, our observations stress the impor-tance of interpreting T measurement by applying men-strual phase-specific reference limits, with special con-cerns to daily clinical practice.

The reference limits we reported for T are consistentwith previous reports on females of similar age: 0.090–

0.580 ng/mL (22), 0.41–2.29 nmol/L (0.118–0.661 ng/mL) (29), �0.400 ng/mL (23). Our reference limits forandrostenedione are similar to those reported by Kushniret al (22), 0.350–2.120 ng/mL; but are lower than thosereported by Rauh et al (29), 2.37–10.7 nmol/L (0.679–3.066 ng/mL). Moreover, our DHEA reference limits arehigher than those previously reported, 1.42–9.00 ng/mL(22). As a major source of variability in androgen levels,ethnicity should be taken into consideration when esti-mating, using, or comparing reference limits. Normativevalues estimated on a specific ethnic group cannot be ap-plied to other ethnic groups. Our reference intervals wereestimated on Italian Caucasian girls. Kushnir et al (22)reported that in their study, more than 90% of subjectswere white, but no ethnic indications were given in theother reports (23, 29). Subjects included in these studieswere apparently not selected for criteria related to hy-perandrogenism or adiposity, possibly explaining thelarger intervals obtained by some of them for T and an-drostenedione (22, 29). Other sources of variability mayderive from differences in sample size and from the lack ofharmonization between different LC-MS/MS methodsand the lack of certified material, particularly for andro-stenedione and DHEA (30, 31). Until a standardization ofLC-MS/MS assays is achieved, or unless harmonization ofresults among different laboratories is achieved, reference

Table 3. Clinical, Anthropometric, and Biochemical Features of Luteal Phase Girls Subdivided According toProgesterone Level

Anovulatory Ovulatory P Value

n 18 31Age, y 17.3 (17.0–17.9) 17.7 (16.3–18.4) .507Age at menarche, y 13.0 (12.0–13.0) 12.0 (12.0–14.0) .796Gynecological age, y 4.67 (3.98–5.31) 4.33 (3.77–6.09) .648Menses/year 12.0 (12.0–12.0) 12.0 (12.0–12.0) .756Cycle day 24.0 (17.0–30.0) 23.0 (20.0–28.0) .534mF-G score 3.00 (2.00–4.00) 3.00 (2.00–5.00) .764Weight, kg 56.0 (50.5–62.0) 55.1 (53.4–58.0) .950Height, m 1.64 (1.60–1.65) 1.63 (1.59–1.68) 1.000BMI, kg/m2 20.9 (19.2–21.6) 21.1 (19.7–22.1) .796Waist circumference, cm 72.5 (70.0–77.0) 74.0 (71.3–78.0) .290Hip circumference, cm 91.0 (89.0–97.0) 92.0 (89.3–94.0) .590Waist/hip ratio 0.811 (0.742–0.830) 0.819 (0.804–0.836) .130Diastolic blood pressure, mm Hg 73 (60–80) 65 (60–75) .152Systolic blood pressure, mm Hg 115 (110–120) 110 (100–115) .115TSH, �IU/mL 1.69 (0.99–2.47) 1.67 (1.40–2.32) .788Prolactin, ng/mL 13.5 (10.0–16–0) 12.0 (9.3–19.0) .671FSH, �IU/mL 4.60 (3.90–6.20) 2.60 (2.23–3.78) �.001LH, �IU/mL 8.10 (3.10–11.30) 4.20 (2.70–6.45) .029E2, pg/mL 69.5 (40.0–101.0) 161.0 (115.3–210.3) �.001Insulin, �IU/mL 8.40 (5.30–10.60) 6.00 (5.35–8.38) .315Glucose, mg/dL 74.0 (64.0–85.0) 78.0 (71.0–83.3) .304Total cholesterol, mg/dL 175 (155–198) 157 (136–172) .029HDL-cholesterol, mg/dL 62.0 (56.0–73.0) 53.0 (47.5–56.5) �.001Triglycerides, mg/dL 65.0 (49.0–80.0) 64.0 (54.0–82.8) .740

Data are expressed as median (IQR).



intervals estimated with an assay cannot be transferred toother assays.

Anovulatory cycles are very frequent in the first yearsafter menarche and have been associated with increasedandrogen levels (32). The use of circulating progesteronefor the assessment of ovulation is largely used both inclinical practice and in scientific reports (33). Unfortu-nately, progesterone normative data based on LC-MS/MSdetermination are still not available. Nonetheless, theanalysis of the distribution of progesterone levels revealedthe presence of 2 distinct populations within the lutealphase group: an anovulatory subgroup, showing proges-terone levels below 2 ng/mL, and an ovulatory subgroup,displaying progesterone levels above 3 ng/mL. The per-centage of healthy anovulatory girls that we reported,36.7%, is in agreement with previous studies by Apter andVihko (34), who observed that the frequency of anovula-tory cycles decreases from 85% in the first year, to 59% inthe third year, to 25% in the sixth year after menarche.

Therefore, we confirmed that in oth-erwise healthy, normal-weight eu-menorrheic young females, a condi-tion of defective ovulation can bedetected, characterized by a derange-ment of the androgen profile in theluteal phase, such as elevated T andandrostenedione and reduced 17OHPlevels. The androgen imbalance isparalleled by higher levels of LH andFSH, almost doubled in anovulatorycompared to ovulatory girls. Inter-estingly, Apter and colleagues re-ported a similar hormonal derange-ment in the luteal phase in adultwomen defined as aluteal or ovula-tory based on serum progesteronebelow or above 6 nmol/L (1.887 ng/mL), but no differences were ob-served between the same groups inthe follicular phase (35). In anovula-tory girls, the increase in T levels isparalleled by reduced levels of E2,less than 50% of the median level re-ported for ovulatory girls, and can beexplained by reduced activity of theP450 aromatase, converting T to E2,and by a general subnormal activityof the corpus luteum (36). The con-comitant increase in FSH and LHlevels does not support the presenceof an imbalance typical of adultPCOS, in which the LH increase isaccompanied by an FSH decrease

(37, 38), but could be explained by the insufficient feed-back operated by E2 and progesterone at the hypothala-mus-pituitary level.

Finally, we found increased total and HDL-cholesterollevels in healthy anovulatory girls, an observation not sup-porting the presence of a metabolic impairment often as-sociated with hyperandrogenemia, as confirmed by thefact that no evidence of glucose metabolism or insulinfunction impairment was observed in the same group (39).

Taken together, these data suggest that the androgen im-balance observed in healthy anovulatory girls may be due toan incomplete level of maturation of the hypothalamus-pi-tuitary-ovary axis, rather than to a hormonal-metabolic im-pairment typical of PCOS (34, 37). Nonetheless, an associ-ation between sex hormones, gonadotropins, and lipidprofile was highlighted. It is well known that the adrenalcontribution to circulating androgens is higher in pubertaland late pubertal age than in adult age (40). Accordingly, the

Figure 3. Box-and-whiskers plots of T (A), SHBG (B), FAI (C), androstenedione (D), DHEA (E),and 17OHP (F) distribution in luteal ovulatory and anovulatory females. Central box representsvalues ranging between the 25th and 75th percentiles, the middle line represents the median,and the lower and upper whiskers extend from (25th percentile � 1.5 � IQR) to (75th percentile� 1.5 � IQR), respectively.



fact that DHEA tended to follow the increase of T and an-drostenedione in anovulatory girls might support the con-cept that the adrenal contribution could be significantlyhigher in the anovulatory girls (41).

Because endocrine features manifested at an early ageare presumed to persist in adult life, a prospective study isneeded to understand whether these girls have a higherrisk of developing hyperandrogenism, metabolic dysfunc-tions, or infertility in adult life (35, 42, 43).

In conclusion, our LC-MS/MS method demonstratesproper sensitivity to measure androgen at low circulatinglevels and adequate specificity and precision to highlightminimal but significant alterations, even when they occurwithin the physiological range. We provide adolescence-specific reference intervals and show that differences existbetween the follicular and luteal phases, thus stressing theimportance of evaluating T in awareness of the menstrualphase by applying specific cutoffs. We also show that evenin healthy, normal-weight, and eumenorrheic girls, smallbut significant alterations of the androgen profile are de-tectable and are related to suboptimal ovulationefficiency.

Acknowledgments

We thank Fondazione Cassa di Risparmio, Bologna, Italy, forsupporting the Centre of Applied Biomedical Research. We alsothank Dr Pasquale Chieco for the laboratory organizationand Dr Susan West for language editing of the manuscript.

Address all correspondence and requests for reprints to:Renato Pasquali, MD, Division of Endocrinology, Depart-ment of Medical and Surgical Sciences, St. Orsola-MalpighiHospital, Via Massarenti 9, 40138 Bologna, Italy. E-mail:[email protected].

This research received an unrestricted grant from a not-for-profit foundation, “Fondazione Centro Studi sulla Calvizie Ce-sare Ragazzi Laboratories,” Zola Predosa, Bologna, Italy.

Disclosure Summary: The authors have nothing to disclose.

References

1. Steinberger E, Ayala C, Hsi B, et al. Utilization of commercial lab-oratory results in management of hyperandrogenism in women. En-docr Pract. 1998;4:1–10.

2. Herold DA, Fitzgerald RL. Immunoassays for testosterone in wom-en: better than a guess? Clin Chem. 2003;49:1250–1251.

3. Taieb J, Mathian B, Millot F, et al. Testosterone measured by 10immunoassays and by isotope-dilution gas chromatography-massspectrometry in sera from 116 men, women, and children. ClinChem. 2003;49:1381–1395.

4. Wang C, Catlin DH, Demers LM, Starcevic B, Swerdloff RS. Mea-surement of total serum testosterone in adult men: comparison ofcurrent laboratory methods versus liquid chromatography-tandemmass spectrometry. J Clin Endocrinol Metab. 2004;89:534–543.

5. Yang DT, Owen WE, Ramsay CS, Xie H, Roberts WL. Performance

characteristics of eight estradiol immunoassays. Am J Clin Pathol.2004;122:332–337.

6. Fanelli F, Belluomo I, Di Lallo VD, et al. Serum steroid profiling byisotopic dilution-liquid chromatography-mass spectrometry: com-parison with current immunoassays and reference intervals inhealthy adults. Steroids. 2011;76:244–253.

7. Rosner W, Auchus RJ, Azziz R, Sluss PM, Raff H. Position state-ment: utility, limitations, and pitfalls in measuring testosterone: anEndocrine Society position statement. J Clin Endocrinol Metab.2007;92:405–413.

8. Vogeser M, Seger C. LC-MS/MS in clinical chemistry. J ChromatogrB Analyt Technol Biomed Life Sci. 2012;883–884:1–2.

9. Roe AH, Dokras A. The diagnosis of polycystic ovary syndrome inadolescents. Rev Obstet Gynecol. 2011;4:45–51.

10. Gambineri A, Fanelli F, Prontera O, et al. Prevalence of hyperan-drogenic states in late adolescent and young women: epidemiolog-ical survey on high-school students in Northern Italy. J Clin Endo-crinol Metab. 2013;98:1641–1650.

11. Hatch R, Rosenfield RL, Kim MH, Tredway D. Hirsutism: impli-cations, etiology, and management. Am J Obstet Gynecol. 1981;140:815–830.

12. Ludwig E. Classification of the types of androgenetic alopecia (com-mon baldness) occurring in the female sex. Br J Dermatol. 1977;97:247–254.

13. Burke BM, Cunliffe WJ. The assessment of acne vulgaris—the Leedstechnique. Br J Dermatol. 1984;111:83–92.

14. Burke JP, Hale DE, Hazuda HP, Stern MP. A quantitative scale ofacanthosis nigricans. Diabetes Care. 1999;22:1655–1659.

15. Cacciari E, Milani S, Balsamo A, et al. Italian cross-sectional growthcharts for height, weight and BMI (2 to 20 yr). J Endocrinol Invest.2006;29:581–593.

16. van Hooff MH, Voorhorst FJ, Kaptein MB, Hirasing RA, Koppe-naal C, Schoemaker J. Endocrine features of polycystic ovary syn-drome in a random population sample of 14–16 year old adoles-cents. Hum Reprod. 1999;14:2223–2229.

17. Vermeulen A, Verdonck L, Kaufman JM. A critical evaluation ofsimple methods for the estimation of free testosterone in serum.J Clin Endocrinol Metab. 1999;84:3666–3672.

18. Siegel S. Nonparametric statistics for the behavioral sciences. To-kyo, Japan: McGraw-Hill; 1956:193.

19. Aedo AR, Landgren BM, Cekan Z, Diczfalusy E. Studies on thepattern of circulating steroids in the normal menstrual cycle. 2. Lev-els of 20�-dihydroprogesterone, 17-hydroxy-progesterone and 17-hydroxypregnenolone and the assessment of their value for ovula-tion prediction. Acta Endocrinol (Copenh). 1976;82:600–616.

20. Shiraishi S, Lee PW, Leung A, Goh VH, Swerdloff RS, Wang C.Simultaneous measurement of serum testosterone and dihydrotes-tosterone by liquid chromatography-tandem mass spectrometry.Clin Chem. 2008;54:1855–1863.

21. Turpeinen U, Linko S, Itkonen O, Hämäläinen E. Determination oftestosterone in serum by liquid chromatography-tandem mass spec-trometry. Scand J Clin Lab Invest. 2008;68:50–57.

22. Kushnir MM, Blamires T, Rockwood AL, et al. Liquid chromatog-raphy-tandem mass spectrometry assay for androstenedione, dehy-droepiandrosterone, and testosterone with pediatric and adult ref-erence intervals. Clin Chem. 2010;56:1138–1147.

23. Salameh WA, Redor-Goldman MM, Clarke NJ, Reitz RE, CaulfieldMP. Validation of a total testosterone assay using high-turbulenceliquid chromatography tandem mass spectrometry: total and freetestosterone reference ranges. Steroids. 2010;75:169–175.

24. Haring R, Hannemann A, John U, et al. Age-specific referenceranges for serum testosterone and androstenedione concentrationsin women measured by liquid chromatography-tandem mass spec-trometry. J Clin Endocrinol Metab. 2012;97:408–415.

25. McCartney CR, Blank SK, Prendergast KA, et al. Obesity and sexsteroid changes across puberty: evidence for marked hyperandro-genemia in pre- and early pubertal obese girls. J Clin EndocrinolMetab. 2007;92:430–436.



mailto:[email protected]

26. Schijf CP, van der Mooren MJ, Doesburg WH, Thomas CM, Rol-land R. Differences in serum lipids, lipoproteins, sex hormone bind-ing globulin and testosterone between the follicular and the lutealphase of the menstrual cycle. Acta Endocrinol (Copenh). 1993;129:130–133.

27. Elliott KJ, Cable NT, Reilly T, Diver MJ. Effect of menstrual cyclephase on the concentration of bioavailable 17-� oestradiol and tes-tosterone and muscle strength. Clin Sci (Lond). 2003;105:663–669.

28. Rothman MS, Carlson NE, Xu M, et al. Reexamination of testos-terone, dihydrotestosterone, estradiol and estrone levels across themenstrual cycle and in postmenopausal women measured by liquidchromatography-tandem mass spectrometry. Steroids. 2011;76:177–182.

29. Rauh M, Gröschl M, Rascher W, Dörr HG. Automated, fast andsensitive quantification of 17 �-hydroxy-progesterone, androstene-dione and testosterone by tandem mass spectrometry with on-lineextraction. Steroids. 2006;71:450–458.

30. Thienpont LM, Van Uytfanghe K, Blincko S, et al. State-of-the-artof serum testosterone measurement by isotope dilution-liquid chro-matography-tandem mass spectrometry. Clin Chem. 2008;54:1290–1297.

31. Vesper HW, Bhasin S, Wang C, et al. Interlaboratory comparisonstudy of serum total testosterone [corrected] measurements per-formed by mass spectrometry methods. Steroids. 2009;74:498–503.

32. Pasquali R, Gambineri A. Polycystic ovary syndrome: a multifaceteddisease from adolescence to adult age. Ann NY Acad Sci. 2006;1092:158–174.

33. Nestler JE, Jakubowicz DJ, Evans WS, Pasquali R. Effects of met-formin on spontaneous and clomiphene-induced ovulation in thepolycystic ovary syndrome. N Engl J Med. 1998;338:1876–1880.

34. Apter D, Vihko R. Hormonal patterns of the first menstrual cycles.In: Venturoli S, Flamigni C, Givens JR, eds. Adolescence in females.Chicago. Illinois: Year Book Medical Publisher; 1985:215–238.

35. Apter D, Vihko R. Endocrine determinants of fertility: serum an-drogen concentrations during follow-up of adolescents into the thirddecade of life. J Clin Endocrinol Metab. 1990;71:970–974.

36. Morishima A, Grumbach MM, Simpson ER, Fisher C, Qin K. Aro-matase deficiency in male and female siblings caused by a novelmutation and the physiological role of estrogens. J Clin EndocrinolMetab. 1995;80:3689–3698.

37. Venturoli S, Porcu E, Fabbri R, et al. Longitudinal evaluation of thedifferent gonadotropin pulsatile patterns in anovulatory cycles ofyoung girls. J Clin Endocrinol Metab. 1992;74:836–841.

38. Blank SK, McCartney CR, Marshall JC. The origins and sequelae ofabnormal neuroendocrine function in polycystic ovary syndrome.Hum Reprod Update. 2006;12:351–361.

39. Rosenfield RL. Clinical review: identifying children at risk for poly-cystic ovary syndrome. J Clin Endocrinol Metab. 2007;92:787–796.

40. Parker LN. Adrenarche and puberty. In: Parker LN, eds. Adrenalandrogens in clinical medicine. San Diego, California: AcademicPress, Inc; 1989:98–117.

41. Abraham GE. Ovarian and adrenal contribution to peripheral an-drogens during the menstrual cycle. J Clin Endocrinol Metab. 1974;39:340–346.

42. Pasquali R, Gambineri A, Anconetani B, et al. The natural historyof the metabolic syndrome in young women with the polycysticovary syndrome and the effect of long-term oestrogen-progestagentreatment. Clin Endocrinol (Oxf). 1999;50:517–527.

43. Gambineri A, Patton L, Altieri P, et al. Polycystic ovary syndrome isa risk factor for type 2 diabetes: results from a long-term prospectivestudy. Diabetes. 2012;61:2369–2374.

Save the Date for Pediatric Endocrine Board Review (PEBR), September 24–25, 2013, Hyatt Regency New Orleans New Orleans, LA

www.endo-society.org/CEU2013



Androgen Profiling by Liquid Chromatography–Tandem Mass Spectrometry (LC-MS/MS) in Healthy Normal-Weight Ovulatory and Anovulatory Late Adolescent and Young Women

Documents