Papers in Press. Published August 1, 2018 as doi: /clinchem

Transcription

1 Papers in Press. Published August 1, 2018 as doi: /clinchem The latest version is at Clinical Chemistry 64: (2018) Endocrinology and Metabolism Diagnostic Thresholds for Androgen-Producing Tumors or Pathologic Hyperandrogenism in Women by Use of Total Testosterone Concentrations Measured by Liquid Chromatography Mass Spectrometry Anu Sharma, 1 Ekta Kapoor, 1 Ravinder J. Singh, 2 Alice Y. Chang, 1 and Dana Erickson 1* BACKGROUND: Previously defined thresholds for total testosterone (TT) concentrations to screen for androgenproducing tumors (APTs) have used RIA, which can be less accurate in women. We aimed to define diagnostic thresholds to screen for APTs or postmenopausal pathologic hyperandrogenism using TT concentrations measured by LC-MS/MS. METHODS: We performed a retrospective cohort study on all women with TT 3.5 nmol/l and all postmenopausal women presenting with hyperandrogenism between 2004 and 2014 at the Mayo Clinic in Rochester, MN. RESULTS: Of the 369 women with TT 3.5 nmol/l, 89 were included and subdivided into 3 groups based on their clinical diagnosis [21 (24%), APT; 16 (18%), postmenopausal pathologic hyperandrogenism; 52 (58%), polycystic ovary syndrome]. The source of the APT was more frequently ovarian (81%, n 17) than adrenal (19%, n 4). The diagnostic threshold using ROC analysis for TT to identify APT in women with severe biochemical hyperandrogenemia was 5.1 nmol/l (sensitivity, 90%; specificity, 81%). In a second analysis of a cohort of postmenopausal women only presenting with symptoms or signs of hyperandrogenism, median TT was significantly higher in the postmenopausal pathologic hyperandrogenism group (APT and ovarian hyperthecosis) vs the idiopathic hyperandrogenism group (4.9 vs 0.8 nmol/l; P 0.01). In postmenopausal women, the diagnostic threshold for pathologic hyperandrogenism was TT 2.2 nmol/l (sensitivity, 100%; specificity, 86%). CONCLUSIONS: The diagnostic threshold for TT concentration as measured by LC-MS/MS to identify APT in women with biochemical severe hyperandrogenemia was TT 5.1 nmol/l. In postmenopausal women, the diagnostic threshold for pathologic hyperandrogenism was lower (TT 2.2 nmol/l) American Association for Clinical Chemistry The prevalence of hirsutism has been estimated to be as high as 8% in women in the US (1). The evaluation of hirsutism often includes measurement of serum total testosterone (TT) 3 concentrations to determine the need to screen for androgen-producing tumors (APTs). A recent update to the Endocrine Society guidelines recommends measurement of circulating TT concentrations in all women presenting with hirsutism for the diagnosis of polycystic ovary syndrome (PCOS) and to screen for APTs (2). The inclusion of all women presenting with hirsutism extends TT evaluation to as many as an estimated 8% to 20% women with PCOS in the US (1, 3). Although APTs are rare (considered the cause of androgen excess in 1% of women), identification and surgical resection of an APT could have a major clinical impact leading to rapid resolution of symptoms (4). In postmenopausal women, ovarian hyperthecosis (a nonneoplastic process) can mimic an APT, presenting with severe hyperandrogenism (5). Oophorectomy for ovarian hyperthecosis can result in rapid resolution of symptoms. A TT threshold for pathologic hyperandrogenism in postmenopausal women has not been previously defined and would require separate analysis because of the substantial decline in ovarian androgen production after menopause (6). Previously defined discriminatory thresholds have ranged from 5.2 to 6.9 nmol/l for APTs (7 9). How- 1 Division of Endocrinology, Diabetes and Nutrition, Mayo Clinic, Rochester MN; 2 Department of Laboratory Medicine and Pathology, Mayo Clinic, Rochester MN. * Address correspondence to this author at: Mayo Clinic College of Medicine, 200 First St. SW, Rochester, MN Fax ; erickson.dana@mayo.edu. A.Y. Chang and D. Erickson contributed equally as senior author. Received April 17, 2018; accepted July 16, Previously published online at DOI: /clinchem American Association for Clinical Chemistry 3 Nonstandard abbreviations: TT, total testosterone; APT, androgen-producing tumor; PCOS, polycystic ovary syndrome; DHEA-S, dehydroepiandrosterone sulfate; US, ultrasound; CT, computed tomography; MRI, magnetic resonance imaging; IH, idiopathic hyperandrogenism. 1 Copyright (C) 2018 by The American Association for Clinical Chemistry

2 ever, these recommendations were based on measurement of TT by RIA, which is less accurate in women than men, who have concentrations 10-fold higher than women (10, 11). In addition, individual laboratory methods can influence intraassay variability substantially when immunoassay kits are used (12). Liquid chromatography mass spectrometry (LC-MS/MS) has the potential to provide specific and accurate measurement of TT concentrations and is being standardized in conjunction with the CDC (13, 14). Indeed, measurements of sex steroids in children and women by mass spectrometry are being recommended for use in research studies owing to the inaccuracy of RIA at lower concentrations and anticipation that sex steroid hormones will soon be clinically measured exclusively by mass spectrometry (15). Data on the appropriate LC-MS/MS TT threshold for further evaluation for an APT or pathologic cause of hyperandrogenism are needed and have not been previously evaluated. In the current project, we studied a large group of women with presentation of severe biochemical hyperandrogenemia (often presenting a dilemma for clinicians in practice) over a 10-year period beginning in 2004 when TT measurements were made by LC-MS/MS. Our overall aim was to identify the optimal discriminatory threshold for APT. In addition, we performed a secondary analysis that included all postmenopausal women presenting with symptoms or signs of hyperandrogenism during the same period to define a diagnostic threshold specific to the diagnosis of pathologic hyperandrogenism (both APT and ovarian hyperthecosis) in postmenopausal women. We hypothesized that the diagnostic TT threshold for APT and postmenopausal pathologic hyperandrogenism as measured by LC-MS/MS would be lower than the threshold previously referenced with RIA, and that the diagnostic threshold for postmenopausal pathologic hyperandrogenism would be even lower than for APT. Materials and Methods PARTICIPANTS This is a retrospective cohort study of (a) all women with severe biochemical hyperandrogenemia and (b) postmenopausal women presenting with new symptoms of hyperandrogenism at the Mayo Clinic in Rochester, MN, from January 1, 2004, to December 31, The initial dates of search coincided with exclusive use of LC-MS/MS assay in the Mayo laboratories for measurement of serum TT. After receiving Institutional Review Board committee approval, we identified cases by using an institutional clinical data repository query tool (Advanced Cohort Explorer) that rapidly searches text within laboratory results and all aspects of clinical notes (e.g., laboratory, radiology, and pathology notes) in the electronic health record at the Mayo Clinic (Rochester, MN). For the first cohort, we identified all women with a serum TT concentration of 3.5 nmol/l. A second postmenopausal cohort was created including all women with documentation in their clinical notes of postmenopausal status, symptoms or signs of hyperandrogenism (hirsutism, hyperandrogenism, or clitoromegaly), and a measured serum TT. Demographic, clinical, and laboratory data, imaging characteristics, and pathological diagnoses were collected. DIAGNOSTIC CRITERIA AND GROUP DEFINITIONS PCOS was assigned as the final diagnosis if the woman was premenopausal and there was clinical documentation of PCOS (based on the Rotterdam criteria: presence of at least 2 of the following: oligo/anovulation, clinical or biochemical evidence of hyperandrogenism and/or polycystic ovary morphology on ultrasound without other secondary causes of androgen excess and/or oligo/ anovulation) (Table 1) (16). In this cohort, postmenopausal hyperandrogenism was defined as postmenopausal women with biochemical hyperandrogenism but without an APT. Postmenopausal status was confirmed in the clinical notes by 1 year after the final menstrual period or history of bilateral oophorectomy. The diagnosis of APT required a radiologic image identifying a tumor, ovarian venous sampling localizing the source from 1 organ (1 adrenal or 1 ovary), or the pathologic diagnosis of tumor. In the postmenopausal cohort with hyperandrogenism, postmenopausal pathologic hyperandrogenism included women diagnosed with APT or ovarian hyperthecosis for the objective of defining a screening threshold for further diagnostic testing. Ovarian hyperthecosis was diagnosed by pathologic confirmation of stromal hyperplasia without a tumor present, resolution of hyperandrogenism after bilateral oophorectomy without a tumor identified by pathology or after medical ovarian suppression, or clinical diagnosis based on radiologic enlargement of both ovaries ( 4cm 3 ) (17), significant virilization or progressive pace of symptoms or unilateral ovarian enlargement without identification of a tumor, or increased TT in both ovarian veins on ovarian vein sampling. Idiopathic hyperandrogenism was defined as TT in the reference interval ( 2.1 nmol/l) and/or mild nonprogressive hirsutism and/or previous premenopausal diagnosis of PCOS. Participants were excluded if they had no clinical evaluation in the electronic health record, were receiving exogenous testosterone replacement, or were transgender, pregnant, 18 years old, or had another diagnosis that could influence TT concentrations (e.g., cirrhotic liver disease, adrenocortical carcinoma, Cushing s disease/syndrome, congenital adrenal hyperplasia, or andro- 2 Clinical Chemistry 64:11 (2018)

3 Testosterone in Pathologic Hyperandrogenism Table 1. Group definitions and criteria. Definition PCOS as defined by the Rotterdam criteria (16) Postmenopausal hyperandrogenism Presence of at least 2: Criteria Oligo/anovulation Clinical or biochemical evidence of hyperandrogenism Polycystic ovary morphology on US Exclusion of other secondary diagnoses including APT Postmenopausal status Includes ovarian hyperthecosis and IH (see below) Absence of APT APT Presence of at least 1: Radiologic image localizing a tumor Ovarian venous sampling localizing the source from 1 organ (1 adrenal or 1 ovary) Pathologic diagnosis of tumor Ovarian hyperthecosis (34) Presence of at least one: Clinical diagnosis in the electronic health record Radiologic enlargement of both ovaries (>4 cm 3 ) without identification of a tumor Pathologic confirmation of stromal hyperplasia without a tumor present Resolution of symptoms of hyperandrogenism after bilateral oophorectomy without a tumor identified on pathology Postmenopausal pathologic Presence of at least 1: hyperandrogenism Ovarian hyperthecosis APT IH TT in the normal range (<2.1 nmol/l) and/or Inability to localize a pathologic source of excess androgens gen insensitivity syndrome) (Fig. 1). Fifty-three women in the second cohort were excluded owing to lack of documented symptoms of hyperandrogenism. In other words, phrases such as no hirsutism or denies symptoms of hyperandrogenism were used in their clinical chart (Fig. 1B). ANALYTIC TECHNIQUES TT was measured via extraction utilizing high-throughput liquid chromatography followed by conventional liquid chromatography and analysis on a tandem mass spectrometer (LC-MS/MS) (18). Intraassay and interassay CVs were 5%. Bioavailable testosterone was measured using an isotopic precipitation method. After precipitation, radioactivity was measured in the supernatant; the bioavailable testosterone results were expressed as a percentage of total testosterone (19). Free testosterone was measured via equilibrium dialysis of free-labeled testosterone through a semipermeable membrane to separate sex-hormone binding globulin and albumin-bound testosterone. After dialysis, radioactivity was measured both inside and outside the tubing; the free testosterone results were expressed as a percentage of TT. The percentage measured [free / (free sex-hormone binding globulin bound)] was then multiplied by TT (measured by LC- MS/MS), and an absolute free testosterone value was obtained (20). Our laboratory participates in the CDC Hormone Standardization Program, and based on the data, our values were accurate within 4% when 20 ng/dl but were within 20% when 20 ng/dl. Androstenedione was also measured by LC-MS/MS (interassay CV was 10.2%, 6.1%, 5.0%, and 4.1% at 55, 3.8, 25.5, and 69.8 nmol/l, respectively). The assay for dehydroepiandrosterone sulfate (DHEA-S) was an automated competitive chemiluminescent immunoassay (IMMULITE 2000) with an interassay CV of 6% at concentrations of 2.0 mol/l, 3.6 mol/l, and 13.6 mol/l. Clinical Chemistry 64:11 (2018) 3

4 Fig. 1. Cohort of women with biochemical severe hyperandrogenemia (A) and postmenopausal women with symptoms of hyperandrogenism (B). *Other: no clinical evaluation. CAH, congenital adrenal hyperplasia; ACC, adrenocortical carcinoma. STATISTICAL ANALYSIS Categorical variables were reported as percentages and analyzed with the Pearson 2 or Fisher exact test when applicable. Continuous variables were expressed as mean with SD when normally distributed and compared for group differences with the 2-sample t-test or ANOVA when 2 groups were compared. Continuous variables with a nongaussian distribution were reported as median with interquartile range (25th 75th quartiles) and analyzed with nonparametric testing (Wilcoxon rank-sum test). Because the reference interval for DHEA-S is established for specific age intervals, the percentage increase above the upper limit of the age-appropriate reference range was calculated for each participant and then compared by groups using nonparametric testing. A P value of 0.05 was considered significant. A ROC curve analysis was performed and used in determining the optimal threshold for the TT concentration to diagnose either an APT or postmenopausal pathologic hyperandrogenism. The thresholds on the ROC curve that provided the highest detection and prediction probabilities (sensitivity and specificity) were determined. Statistical analysis was performed using the JMP (JMP, SAS Institute) statistical package. Results COHORT OF WOMEN WITH SERUM TT CONCENTRATION 3.5 NMOL/L Clinical characteristics. Eighty-nine women out of 369 were included in the final analysis. They were subdivided into 3 groups based on their clinical diagnosis [21 (24%), APT; 16 (18%), postmenopausal hyperandrogenism; and 52 (58%), PCOS]. Twenty-nine (33%) women were postmenopausal [13 (62%), APT; 16 (100%), postmenopausal hyperandrogenism]. Women with APT were significantly older than the PCOS group and younger than women with postmenopausal hyperandrogenism (APT vs PCOS, P 0.01; APT vs postmenopausal hyperandrogenism, P 0.02; PCOS vs postmenopausal hyperandrogenism, P 0.01). Women with APT presented significantly earlier after onset of symptoms and were more likely to have male-pattern alopecia than women with PCOS (see Table 1 in the Data Supplement that accompanies the online version of this article at clinchem.org/content/vol64/issue11). Hormonal measurements. TT concentrations were significantly higher in the APT group than in women with PCOS or postmenopausal hyperandrogenism (see Table 1 in the online Data Supplement and Fig. 2A: APT vs PCOS, P 0.01; APT vs postmenopausal hyperandrogenism, P 0.01; and PCOS vs postmenopausal hyperandrogenism, P 0.01). In addition, bioavailable testosterone (Fig. 2B: APT vs PCOS, P 0.01; APT vs postmenopausal hyperandrogenism, P 0.07; and PCOS vs APT, P 0.4) and free testosterone (Fig. 2C: APT vs PCOS, P 0.01; APT vs postmenopausal hyperandrogenism, P 0.02; and PCOS vs postmenopausal hyperandrogenism, P 0.2) were significantly higher in APT compared with PCOS and postmenopausal hyperandrogenism groups [measured in n 50 4 Clinical Chemistry 64:11 (2018)

5 Testosterone in Pathologic Hyperandrogenism Supplement). Of the 4 patients with adrenal APTs, androstenedione was measured in 3, with 2 of them having significantly increased androstenedione concentrations. The median TT/androstenedione ratio in patients with adrenal APT was 0.34 (n 3; range, ), and in ovarian APT, it was 2.69 (n 6; range, ). There was no statistical difference between the 2 groups, likely owing to limited numbers (P 0.06). DHEA-S was measured in 2 patients with adrenal APTs 1 was normal and the other increased. Fig. 2. Individual values of serum TT (A), bioavailable testosterone (B), and free testosterone (C) in women with severe biochemical hyperandrogenemia. *P < 0.05, **P < PH, postmenopausal hyperandrogenism. (56%) and n 47 (53%) individuals, respectively]. After ROC analysis, a TT concentration of 5.1 nmol/l had a sensitivity of 90% and a specificity of 81% for the diagnosis of APT (AUC, 0.90). DHEA-S concentrations were significantly higher in the PCOS group (see Table 1 in the online Data Supplement). However, after indexing increases in DHEA-S to age-specific ranges, there was no difference between groups. DHEA-S was mildly increased in 3 APT cases [1 adrenal tumor (7% above upper reference limit) and 2 ovarian tumors (4% and 19% above the upper reference limit)]. Androstenedione concentrations were available in only 26 (29%) cases. There was no statistical difference between groups in androstenedione or calculated TT/ androstenedione ratios (see Table 1 in the online Data Radiology and pathology. Eighty-one women underwent at least 1 imaging study with 18 (22%) identifying an APT. If imaging were performed only in women with a TT 5.1 nmol/l, 48 (60%; 40 PCOS and 8 postmenopausal hyperandrogenism) women without an underlying tumor would not have undergone imaging. Transabdominal or transvaginal ultrasonography (US) was the first radiologic test obtained in 75 (84%) women. This was followed by either computed tomography (CT) of the abdomen and pelvis (APT, 71%; postmenopausal hyperandrogenism, 63; PCOS, 19%; P 0.01) or magnetic resonance imaging (MRI) of the abdomen and pelvis (APT, 24%; postmenopausal hyperandrogenism, 13%; PCOS, 4%; P 0.01). The source of the APT was more frequently ovarian (81%, n 17) than adrenal (19%, n 4) (Table 2). Imaging failed to identify the APT in 3 cases (all had US and MRI). All 3 women were postmenopausal and underwent bilateral oophorectomy with final pathology identifying ovarian tumors (Table 2, cases 10, 20, and 21). Pelvic US failed to identify the APT in 4 of 16 cases (25%); CT scan failed to identify 2 of 15 (13%) tumors; and MRI did not identify any tumors that were not already identified by another imaging technique (Table 2). Three women (1 premenopausal and 2 postmenopausal) required ovarian and adrenal vein sampling to diagnose an ovarian source. POSTMENOPAUSAL WOMEN WITH SYMPTOMS OF HYPERANDROGENISM (SECOND COHORT) Clinical characteristics. Seventy-one individuals (20 were in the first cohort) were included in the final analysis of the second cohort [36 idiopathic hyperandrogenism (IH) and 35 postmenopausal pathologic hyperandrogenism]. There were no statistically significant differences in age, ethnicity, frequency of symptoms of hyperandrogenism (acne, alopecia, or deepening of the voice), or duration of symptoms. Male-pattern alopecia and the distribution of terminal hair growth differed significantly between groups, being more severe in the postmenopausal pathologic hyperandrogenism group (see Table 2 in the online Data Supplement). Clinical Chemistry 64:11 (2018) 5

6 Table 2. Clinical characteristics and pathological diagnoses in the 21 women with an APT as the cause of severe biochemical hyperandrogenemia. Case Age, years TT, nmol/l Pelvic US CT MRI Pathology a Fibrothecoma a Not available a Cystic granulosa cell tumor a + a Sertoli Leydig cell tumor a + a Leydig cell tumor a Adrenocortical adenoma a + a Steroid cell tumor a Ovarian metastasis from neuroendocrine tumor a + a Leydig cell tumor Leydig cell tumor Not available a Sertoli Leydig cell tumor a + a + a Serous cystadenoma a + a Sertoli Leydig cell tumor a + a Sertoli Leydig cell tumor a + a Transitional cell carcinoma of Mullerian origin a Adrenocortical adenoma a + Sertoli Leydig cell tumor a Granulosa cell tumor Leydig cell tumor Bilateral steroid cell tumor a Denotes imaging was diagnostic (+, imaging was performed;, imaging was not performed). Hormonal measurements. TT concentrations were significantly higher in the postmenopausal pathologic hyperandrogenism group compared with the IH group [Fig. 3A: 4.9 ( ) vs 0.8 ( ) nmol/l; P 0.01]. In addition, bioavailable testosterone concentrations [Fig. 3B: 0.6 ( ) vs 0.1 ( ) nmol/l; P 0.01] and free testosterone concentrations [Fig. 3C: 0.1 ( ) vs 0.02 ( ) nmol/l; P 0.01] were significantly higher in those with postmenopausal pathologic hyperandrogenism compared with the IH group (measured in 62% and 55% of patients, respectively). DHEA-S concentrations were not found to be different between groups [1.4 ( ) vs 1.1 ( ) mol/l; P 0.7]. DHEA-S was significantly increased in adrenal APT (n 1; DHEA-S, 6.0 mol/l) compared with ovarian APTs (n 28; median, 1.4 mol/l; interquartile range, mol/l). Interestingly, despite higher androstenedione concentrations [measured in 21 (30%) of patients] in the postmenopausal pathologic hyperandrogenism group, the TT/androstenedione ratio was also greater in the postmenopausal pathologic hyperandrogenism group (see Table 2 in the online Data Supplement). Luteinizing hormone and follicle-stimulating hormone concentrations were similar between the 2 groups. ROC analysis of TT concentrations revealed the most optimal diagnostic threshold was 2.2 nmol/l (sensitivity, 100%; specificity, 86%) for the diagnosis of postmenopausal pathologic hyperandrogenism (AUC, 0.95). Radiology and pathology. Forty-four (62%) women in the postmenopausal cohort with symptoms of hyperandrogenism underwent radiological investigation [25% (n 9) in the IH group and 100% (n 35) in the postmenopausal pathologic hyperandrogenism group]. Most women in the postmenopausal pathologic hyperandrogenism group received pelvic US as their first radiological investigation (94% vs 17%; P 0.01), followed by CT scan (60% vs 11%; P 0.01) and MRI scan (23% vs 3%; P 0.01). The source of pathologic hyperandrogenism was largely ovarian (97%; n 34) with only 1 woman s source being adrenal (2.2 cm right adrenal adenoma on pathology) (Table 3). Of the 35 individuals in the postmenopausal pathologic hyperandrogenism group, 20 (57%) had a clinical or pathological diagnosis of ovarian hyperthecosis, and 15 (43%) met the criteria for an APT. 6 Clinical Chemistry 64:11 (2018)

7 Testosterone in Pathologic Hyperandrogenism Discussion Fig. 3. Individual values of serum TT (A), bioavailable testosterone (B), and free testosterone (C) in postmenopausal women with symptoms of hyperandrogenism. **P < PPH, postmenopausal pathologic hyperandrogenism. Of those with a diagnosis of ovarian hyperthecosis, median ovarian volumes on pelvic US were 4.3 ml (range, ml) on the right and 5.3 ml (range, ml) on the left. Pelvic US identified 53% (n 8) of ovarian APT. CT scan identified 1 additional tumor when pelvic US failed to do so. There was pathological confirmation of an APT in 12 women. Eleven (92%) had an ovarian tumor and 1 (8%) had an adrenal tumor. In those without imaging or pathologic abnormality (n 6), postoperative testosterone concentrations after bilateral oophorectomy were all within the normal range with resolution of hyperandrogenism symptoms at follow-up evaluation. APTs have been previously described as having a rapid onset of virilizing symptoms with increased TT concentrations 6.9 nmol/l by RIA (21, 22). The results of our study show that when measured by LC-MS/MS, TT of 5.1 nmol/l was the most discriminatory diagnostic threshold for the diagnosis of APT in women with severe biochemical hyperandrogenemia (defined as TT 3.5 nmol/l). In postmenopausal women presenting with symptoms of hyperandrogenism, however, we found that a lower diagnostic threshold for TT ( 2.2 nmol/l) optimally identified postmenopausal pathologic hyperandrogenism (APT and ovarian hyperthecosis). Given the predicted shift of all TT measurements to LC-MS/MS in women (15) and the recent updated recommendation by the Endocrine Society to measure TT in all women with hirsutism (2), these results provide an important update to clinical practice. To our knowledge, this is the first study using LC-MS/MS to characterize TT diagnostic thresholds for APT in women with hyperandrogenemia and for pathologic hyperandrogenism in postmenopausal women. Previous studies evaluating the optimal diagnostic threshold for TT concentrations to identify APTs used less accurate methods for measuring TT in women. Derksen et al. (23) used RIA in 100 women with androgen excess (27 with virilizing adrenal tumors and 73 with nonneoplastic hirsutism) and proposed a higher diagnostic threshold (TT 6.9 nmol/l; sensitivity, 100%; specificity, 50%). However, their cohort comprised adrenocortical carcinoma, including those with cosecretion of cortisol, as the majority of virilizing tumors (12 of 14). We excluded adrenocortical carcinoma and Cushing s syndrome, as the clinical picture often differs, with the diagnostic evaluation guided by the Cushing s diagnosis. In addition, Yance et al. (24) suggested a significantly higher TT concentration of 10.8 nmol/l (sensitivity, 76.9%; specificity, 90.5%) to differentiate ovarian APT from ovarian hyperthecosis. However, their approach has several limitations in applicability to clinical practice. The immunofluorometric assay used to measure TT concentrations is subject to increased variability, and results differ depending on the method of collection (25). Application of a diagnostic threshold of 6.9 nmol/l in our cohort would have resulted in 6 missed cases of APT and 22 missed cases of postmenopausal pathologic hyperandrogenism. This supports our hypothesis that the LC- MS/MS diagnostic threshold for APT would need to be lower than RIA based on previous data that RIA overestimates TT concentrations in women (10, 11, 26). Kaltas et al. (27) showed that once adrenocortical carcinoma is excluded, an ovarian tumor is the most common cause of an APT, which is consistent with our results (81% ovarian origin). This is an important distinction Clinical Chemistry 64:11 (2018) 7

8 Table 3. Clinical characteristics and pathological diagnoses of the 35 postmenopausal women with symptoms of hyperandrogenism and severe biochemical hyperandrogenemia. Cases Age, years TT, nmol/l Right ovary size, cm Right ovarian volume, cm 3 Pelvic US Left ovary size, cm Left ovarian volume, cm 3 Pathology Ovarian hyperthecosis Ovaries without diagnostic abnormalities Not available Not available Not available Ovaries without diagnostic abnormalities Not available Ovaries without diagnostic abnormalities Not available Ovaries without diagnostic abnormalities Bilateral ovarian stromal hyperplasia Not available Ovaries without diagnostic abnormalities Bilateral ovarian stromal hyperplasia Not available Not available Bilateral ovarian stromal hyperplasia Not available Bilateral ovarian nodular stromal hyperplasia Ovaries without diagnostic abnormalities Bilateral ovarian stromal hyperplasia APT Adrenocortical adenoma Right ovarian luteinized fibrothecoma Not available Right ovarian Leydig cell tumor Right Sertoli Leydig cell tumor Left ovarian serous cystadenoma Bilateral metastases from a neuroendocrine tumor Right ovarian Leydig cell tumor Left ovarian Leydig cell tumor Not available Right ovarian serous cystadenoma Right ovarian Sertoli Leydig cell tumor Bilateral ovarian Leydig cell tumors Bilateral ovarian steroid cell tumors Not available from other cohorts, in which adrenocortical carcinoma with or without Cushing s comprised 59% of cases of hyperandrogenemic patients (23, 28). Therefore, pelvic imaging should be considered as the first step after biochemical confirmation of severe hyperandrogenemia, especially if the clinical picture does not suggest hypercortisolism. Once an adrenal source has been excluded, the evaluation should focus on localization of an ovarian source and/or consideration of bilateral oophorectomy in postmenopausal women. POSTMENOPAUSAL HYPERANDROGENISM Severe androgen excess in women can lead to significant psychosocial distress (29), decreased quality of life (30), 8 Clinical Chemistry 64:11 (2018)

9 Testosterone in Pathologic Hyperandrogenism and possible increased risk for insulin resistance (31), cardiovascular disease (32, 33), and endometrial adenocarcinoma in postmenopausal women (34). Therefore, the benefits of surgical intervention for postmenopausal women with ovarian hyperthecosis or APT often outweigh the risks (35). Thus, we combined both ovarian hyperthecosis and APT groups for postmenopausal women as a meaningful diagnostic category for decisionmaking about further diagnostic testing and possible surgical intervention. To our knowledge, we are the first to demonstrate with any assay that the diagnostic threshold for pathologic hyperandrogenism in postmenopausal women should be lower (TT 2.2 nmol/l) to ensure inclusion of clinically significant ovarian hyperthecosis. Because this threshold is close to the upper limit of normal, this suggests that the normal reference range for TT should be redefined for menopausal women. The diagnostic criteria for ovarian hyperthecosis are yet to be clearly defined. Although the gold standard is a histological diagnosis with nests of steroidogenically active luteinized stromal cells throughout the ovarian stroma (36), other definitions include postmenopausal status with symptoms of hyperandrogenism (33), TT 5.2 nmol/l (8), and US findings of bilateral enlarged ovaries (37). Ovarian hyperthecosis cases have also been described with normal ovarian volume (37) and histology (37 39). Immunohistochemistry in reported cases with normal histology has revealed the presence of P450c17 -positive cells, indicating persistence of steroidogenic luteinizing hormone-responsive cells in postmenopausal women (38, 39). The clinical diagnosis of ovarian hyperthecosis was assigned in our cohort despite normal imaging and ovarian pathology in 45% of cases (see the online Data Supplement). Consistent with previous reports (24), ovarian volume was increased in some but not all individuals in the ovarian hyperthecosis group (7 of 20 in our cohort and 8 of 21 in the previously reported cohort) (24). This highlights the limitation of histologic and radiologic diagnosis of ovarian hyperthecosis. The clinical diagnosis of ovarian hyperthecosis identified women who responded to medical or surgical therapy, confirming the ovarian source of hyperandrogenism. STRENGTHS AND LIMITATIONS Our study has several strengths, including the use of LC- MS/MS (gold standard) for all testosterone measurements and a thorough systematic search of the Mayo Clinic laboratory and clinical electronic health record, which had a high volume of exclusive LC-MS/MS TT measurements in women during the specified period. This cohort also includes the greatest number of ovarian APTs to date. As a retrospective review, this study might have missed cases of APT at an even lower TT concentration in premenopausal women if the evaluating provider did not order a TT. However, a separate confirmatory search was performed for APT, and there were no cases of APT below a TT concentration of 3.5 nmol/l. Because of the tertiary care practice at the Mayo Clinic, some diagnosed APT and ovarian hyperthecosis cases were not resected or managed clinically at our center. As a result, some patients were lost to follow-up [36 (22.5%) patients in both cohorts combined (1 woman in the APT group, 6 women in the postmenopausal pathologic hyperandrogenism group)], and not all cases had pathologic confirmation. However, any misclassification between ovarian hyperthecosis and APT would still include the woman in the appropriate group postmenopausal pathologic hyperandrogenism the clinically relevant group to identify with a diagnostic threshold for further evaluation, treatment, and/or surgery. In addition, other biochemical markers of hyperandrogenemia (bioavailable testosterone, free testosterone, DHEA-S, and androstenedione) were not measured in all patients. Given the rarity of postmenopausal pathologic hyperandrogenism, 20 women were included in both cohorts. These findings would be strengthened by future confirmatory studies utilizing larger cohorts. Conclusion Using LC-MS/MS measurements of TT serum concentrations in women, the optimal screening threshold to diagnose APT was a TT of 5.1 nmol/l. To our knowledge, this is also the first study to describe a TT diagnostic threshold for postmenopausal pathologic hyperandrogenism derived from a postmenopausal cohort. We demonstrated that a lower threshold (TT 2.2 nmol/l) identifies symptomatic postmenopausal women who should have further detailed imaging evaluation for postmenopausal pathologic hyperandrogenism that could be cured with surgical intervention rather than medications. Clinical history appeared to be of limited use in identifying women with APT or postmenopausal pathologic hyperandrogenism. Author Contributions: All authors confirmed they have contributed to the intellectual content of this paper and have met the following 4 requirements: (a) significant contributions to the conception and design, acquisition of data, or analysis and interpretation of data; (b) drafting or revising the article for intellectual content; (c) final approval of the published article; and (d) agreement to be accountable for all aspects of the article thus ensuring that questions related to the accuracy or integrity of any part of the article are appropriately investigated and resolved. A. Sharma, designed the project, performed chart review and data collection, analyzed the data and wrote the manuscript, critically reviewed and edited the manuscript; E. Kapoor, designed the project, performed chart review and data collection, critically reviewed and edited the manuscript; R. Singh, designed the project, validated the laboratory measurements, critically reviewed and edited the manuscript; A. Clinical Chemistry 64:11 (2018) 9

10 Chang, designed the project, analyzed the data and wrote the manuscript, critically reviewed and edited the manuscript; D. Erickson, designed the project, performed chart review and data collection, critically reviewed and edited the manuscript. Authors Disclosures or Potential Conflicts of Interest: Upon manuscript submission, all authors completed the author disclosure form. References Disclosures and/or potential conflicts of interest: No authors declared any potential conflicts of interest. Role of Sponsor: No sponsor was declared. Acknowledgments: The authors thank June Oshiro for her help with image formatting. 1. Knochenhauer ES, Key TJ, Kahsar-Miller M, Waggoner W, Boots LR, Azziz R. Prevalence of the polycystic ovary syndrome in unselected black and white women of the southeastern United States: a prospective study. J Clin Endocrinol Metab 1998;83: Martin KA, Anderson RR, Chang RJ, Ehrmann DA, Lobo RA, Murad MH, et al. Evaluation and treatment of hirsutism in premenopausal women: an Endocrine Society clinical practice guideline. J Clin Endocrinol Metab 2018;103: Chang AY, Ayers C, Minhajuddin A, Jain T, Nurenberg P, de Lemos JA, et al. Polycystic ovarian syndrome and subclinical atherosclerosis among women of reproductive age in the Dallas Heart Study. Clin Endocrinol (Oxf) 2011;74: Azziz R, Sanchez LA, Knochenhauer ES, Moran C, Lazenby J, Stephens KC, et al. Androgen excess in women: experience with over 1000 consecutive patients. J Clin Endocrinol Metab 2004;89: Vaikkakara S, Al-Ozairi E, Lim E, Advani A, Ball SG, James RA, Quinton R. The investigation and management of severe hyperandrogenism pre- and postmenopause: nontumor disease is strongly associated with metabolic syndrome and typically responds to insulin-sensitization with metformin. Gynecol Endocrinol 2008;24: Lasley BL, Santoro N, Randolf JF, Gold EB, Crawford S, Weiss G, et al. The relationship of circulating dehydroepiandrosterone, testosterone, and estradiol to stages of the menopausal transition and ethnicity. J Clin Endocrinol Metab 2002;87: Moltz L, Pickartz H, Sorensen R, Schwartz U, Hammerstein J. Ovarian and adrenal vein steroids in seven patients with androgen-secretingovarianneoplasms:selectivecatheterization findings. Fertil Steril 1984;42: Meldrum DR, Abraham GE. Peripheral and ovarian venous concentrations of various steroid hormones in virilizing ovarian tumors. Obstet Gynecol 1979;53: O Driscoll JB, Mamtora H, Higginson J, Pollock A, Kane J, Anderson DC. A prospective study of the prevalence of clear-cut endocrine disorders and polycystic ovaries in 350 patients presenting with hirsutism or androgenic alopecia. Clin Endocrinol (Oxf) 1994;41: Rosner W, Auchus RJ, Azziz R, Sluss PM, Raff H. Utility, limitations, and pitfalls in measuring testosterone: an Endocrine Society position statement. J Clin Endocrinol Metab 2007;92: Tosi F, Fiers T, Kaufman JM, Dall Alda M, Moretta R, Giagulli VA, et al. Implications of androgen assay accuracy in the phenotyping of women with polycystic ovary syndrome. J Clin Endocrinol Metab 2016;101: Scaglia HE, Buccini G, Chichizola C, Colombani ME, Corazza N, Corthey C, et al. A comparison of serum total testosterone (TT) measurements by various methods, employing current validated techniques, and liquid chromatography-tandemmassspectrometry(lc-ms/ms). A multicentric study possible interferences affecting laboratories variability in TT measurements. Endocr Rev 2017; Salameh WA, Redor-Goldman MM, Clarke NJ, Mathur R, Azziz R, Reitz RE. Specificity and predictive value of circulating testosterone assessed by tandem mass spectrometry for the diagnosis of polycystic ovary syndrome by the National Institutes of Health 1990 criteria. Fertil Steril 2014;101: Travison TG, Vesper HW, Orwoll E, Wu F, Kaufman JM, Wang Y, et al. Harmonized reference ranges for circulating testosterone levels in men of four cohort studies in the United States and Europe. J Clin Endocrinol Metab 2017;102: Handelsman DJ, Wartofsky L. Requirement for mass spectrometry sex steroid assays in the Journal of Clinical Endocrinology and Metabolism. J Clin Endocrinol Metab 2013;98: Revised 2003 consensus on diagnostic criteria and long-term health risks related to polycystic ovary syndrome. Fertil Steril 2004;81: Giacobbe M, Mendes Pinto-Neto A, Simoes Costa-Paiva LH, Martinez EZ. The usefulness of ovarian volume, antral follicle count and age as predictors of menopausal status. Climacteric 2004;7: Wang C, Catlin DH, Demers LM, Starcevic B, Swerdloff RS. Measurement of total serum testosterone in adult men: comparison of current laboratory methods versus liquid chromatography-tandem mass spectrometry. J Clin Endocrinol Metab 2004;89: Wheeler MJ. The determination of bio-available testosterone. Ann Clin Biochem 1995;32(Pt 4): Vermeulen A, Stoica T, Verdonck L. The apparent free testosterone concentration, an index of androgenicity. J Clin Endocrinol Metab 1971;33: FriedmanCI,SchmidtGE,KimMH,PowellJ.Serumtestosterone concentrations in the evaluation of androgenproducingtumors. Am J Obstet Gynecol 1985;153: Monteagudo A, Heller D, Husami N, Levine RU, Mc- Caffrey R, Timor-Tritsch IE. Ovarian steroid cell tumors: sonographic characteristics. Ultrasound Obstet Gynecol 1997;10: Derksen J, Nagesser SK, Meinders AE, Haak HR, van de Velde CJ. Identification of virilizing adrenal tumors in hirsute women. N Engl J Med 1994;331: Yance V, Marcondes JAM, Rocha MP, Barcellos CRG, Dantas WS, Avila AFA, et al. Discriminating between virilizing ovary tumors and ovary hyperthecosis in postmenopausal women: clinical data, hormonal profiles and image studies. Eur J Endocrinol 2017;177: Kohek M, Leme C, Nakamura IT, De Oliveira SA, Lando V, Mendonca BB. Effects of EDTA and sodium citrate on hormone measurements by fluorometric (FIA) and immunofluorometric (IFMA) methods. BMC Clin Pathol 2002;2: Taieb J, Mathian B, Millot F, Patricot M-C, Mathieu E, Queyrel N, 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: Kaltsas GA, Mukherjee JJ, Kola B, Isidori AM, Hanson JA, Dacie JE, et al. Is ovarian and adrenal venous catheterization and sampling helpful in the investigation of hyperandrogenic women? Clin Endocrinol (Oxf) 2003;59: Kaltsas GA, Isidori AM, Kola BP, Skelly RH, Chew SL, Jenkins PJ, et al. The value of the low-dose dexamethasone suppression test in the differential diagnosis of hyperandrogenism in women. J Clin Endocrinol Metab 2003;88: Ekback MP, Lindberg M, Benzein E, Arestedt K. Healthrelated quality of life, depression and anxiety correlate with the degree of hirsutism. Dermatology 2013;227: Khomami MB, Tehrani FR, Hashemi S, Farahmand M, Azizi F. Of PCOS symptoms, hirsutism has the most significant impact on the quality of life of Iranian women. PLoS One 2015;10:e Corbould A. Chronic testosterone treatment induces selective insulin resistance in subcutaneous adipocytes of women. J Endocrinol 2007;192: Macut D, Antic IB, Bjekic-Macut J. Cardiovascular risk factors and events in women with androgen excess. J Endocrinol Invest 2015;38: Markopoulos MC, Kassi E, Alexandraki KI, Mastorakos G, Kaltsas G. Management of endocrine disease: hyperandrogenism after menopause. Eur J Endocrinol 2015;172:R79 R Zhang C, Sung CJ, Quddus MR, Simon RA, Jazaerly T, Lawrence WD. Association of ovarian hyperthecosis with endometrial polyp, endometrial hyperplasia, and endometrioid adenocarcinoma in postmenopausal women: a clinicopathological study of 238 cases. Hum Pathol 2017;59: Alpañés M, González-Casbas JM, Sánchez J, Pián H, Escobar-Morreale HF. Management of postmenopausal virilization. J Clin Endocrinol Metab 2012;97: Brown DL, Henrichsen TL, Clayton AC, Hudson SB, Coddington CC 3rd, Vella A. Ovarian stromal hyperthecosis: sonographic features and histologic associations. J Ultrasound Med 2009;28: Rousset P, Gompel A, Christin-Maitre S, Pugeat M, Hugol D, Ghossain MA, Buy JN. Ovarian hyperthecosis on grayscale and color Doppler ultrasound. Ultrasound Obstet Gynecol 2008;32: Sasano H, Fukunaga M, Rojas M, Silverberg SG. Hyperthecosis of the ovary. Clinicopathologic study of 19 cases with immunohistochemical analysis of steroidogenic enzymes. Int J Gynecol Pathol 1989;8: Sarfati J, Bachelot A, Coussieu C, Meduri G, Touraine P. Impact of clinical, hormonal, radiological, and immunohistochemical studies on the diagnosis of postmenopausal hyperandrogenism. Eur J Endocrinol 2011;165: Clinical Chemistry 64:11 (2018)