{"paper_id":"63c32840-f163-4fee-be69-a465975fb363","body_text":"Epithelial ovarian cancer: testing the\n‘androgens hypothesis’\nCatherine M Olsen 1,2, Ade` le C Green 1, Christina M Nagle 1, Susan J Jordan 1,2,\nDavid C Whiteman 1, Christopher J Bain 2, Penelope M Webb 1 on behalf of the\nAustralian Cancer Study Group (Ovarian Cancer) and the Australian Ovarian\nCancer Study Group\n1Cancer and Population Studies Group, Queensland Institute of Medical Research, Brisbane, 4029, Australia\n2School of Population Health, The University of Queensland, Brisbane, 4029, Australia\n(Correspondence should be addressed to C M Olsen; Email: catherine.olsen@qimr.edu.au)\nAbstract\nIn 1998, Risch proposed a hypothesis for the pathogenesis of ovarian cancer relating to the role of\nandrogens in stimulating epithelial cell proliferation. Although this hypothesis has been widely\ndiscussed, direct evidence to support it is scant. To address this issue, we have conducted a detailed\nanalysis of factors possibly associated with high circulating levels of androgens, including polycystic\novary syndrome (PCOS), hirsutism and acne (all clinically associated with hyperandrogenism) using\nthe data collected in an Australia-wide, population-based case-control study. Cases aged 18–79\nyears with a new diagnosis of invasive epithelial ovarian cancer (nZ1276) or borderline malignant\ntumour ( nZ315) were identified through a network of clinics and cancer registries throughout\nAustralia. Controls (nZ1508) were selected from the National Electoral Roll. Women self-reported a\nhistory of PCOS, acne, hirsutism and also use of testosterone supplements or the androgenic\nmedication Danazol. We found no evidence that a history of PCOS, acne or hirsutism was associated\nwith ovarian cancer overall, or with specific subtypes, with the exception of serous borderline tumours\nthat were positively associated with a history of PCOS (OR 2.6; 95% CI 1.0–6.1). Women who had\never used testosterone supplements had an increased risk of ovarian cancer (OR 3.7; 95% CI\n1.1–12.0); however, use of the androgenic medication Danazol did not increase risk (OR 1.0; 95% CI\n0.4–2.9). Overall, our results do not support the hypothesis that androgen-related disorders increase\nthe risk of ovarian cancer.\nEndocrine-Related Cancer (2008) 15 1061–1068\nIntroduction\nIn 1998, Risch (1998) put forward a hypothesis for the\npathogenesis of ovarian cancer relating to the role of\nandrogens in stimulating epithelial cell proliferation.\nAlthough widely discussed in the aetiologic literature,\nthere is a scant evidence to support this hypothesis.\nWhile a number of in vitro and animal experiments\nsuggest a role for androgens in the development of\novarian cancer, epidemiol ogical evidence is less\nconvincing. Four prospective studies have examined\nthe prediagnostic serum levels of androgens but, aside\nfrom the first very small study ( nZ31 cases;\nHelzlsouer et al . 1995 ), none has reported any\nsignificant associations with ovarian cancer risk\n(Helzlsouer et al. 1995, Lukanova et al. 2003, Rinaldi\net al . 2007 , Tworoger et al . 2007 ). These studies are,\nhowever, based on a single measure of androgens that\nmay not accurately reflect long-term exposure. Others\nhave assessed markers of high androgen levels such as\npolycystic ovary syndrome (PCOS), a disorder of\nfunctional androgen excess ( Azziz 2003 ), and hirsut-\nism and acne which are associated with high\ncirculating levels of androgens ( Lucky 1995 ), but\nthese studies have also been limited by small numbers\nof exposed cases. One case-control study has reported\nan increased risk of ovarian cancer among women with\nPCOS ( Schildkraut et al . 1996 ), although this finding\nwas based on few exposed cases ( nZ7), and has not\nbeen confirmed in prospective studies ( Coulam et al .\n1983, Pierpoint et al . 1998 ); and only one previous\nstudy has investigated the association between acne or\nhirsutism and ovarian c ancer, noting a positive\nassociation based on 13 and 3 exposed cases\nrespectively ( Wynder et al . 1969 ). These associations\nEndocrine-Related Cancer (2008) 15 1061–1068\nEndocrine-Related Cancer (2008) 15 1061–1068\n1351–0088/08/015–001061 q 2008 Society for Endocrinology Printed in Great Britain\nDOI: 10.1677/ERC-08-0075\nOnline version via http://www.endocrinology-journals.org\nDownloaded from Bioscientifica.com at 06/13/2026 12:40:48AM\nvia free access\n\n\nhave not been examined by tumour behaviour or\nhistological subtype despite some known differences in\nrisk factors ( Risch et al . 1996 , Titus-Ernstoff et al .\n2001, Purdie et al .2 0 0 3).\nTo address this issue, we have conducted a detailed\nanalysis of factors that have been clinically associated\nwith high circulating levels of androgens, including\nPCOS, hirsutism and acne, in relation to risk of the\nmajor histological subtypes of ovarian cancer using the\ndata collected from an Australia-wide, population-\nbased case-control study. We also examined whether\nuse of testosterone hormone therapy or the androgenic\nmedication Danazol was related to ovarian cancer.\nDanazol (17- a-ethinltestosterone; marketed as Dano-\ncrine in the US) is a synthetic androgen that binds to\nandrogen receptors and sex hormone-binding globulin\nresulting in a threefold increase in free testosterone and\nis commonly used for the treatment of endometriosis\n(Olive & Pritts 2001 ). We assessed potential\ninteraction between body mass index (BMI) and\nPCOS and ovarian cancer risk, since overweight or\nobese women with PCOS appear to suffer from a more\nsevere form of hyperandrogenism than those of normal\nweight with PCOS ( Gambineri et al . 2002 ).\nMethods\nStudy participants\nThe Australian Ovarian Cancer Study was an Australia-\nwide population-based case-control study of epithelial\novarian cancer; full details of study design and\nparticipant recruitment have been reported previously\n(Merritt et al . 2008 ). Cases were women aged 18–79\nyears living in Australia with histologically confirmed\nepithelial ovarian, fallopian tube or primary peritoneal\ncancer newly diagnosed between January 2002 and June\n2005. Cases were recruited by nurses who liaised with\nthe treatment clinics, physicians and state cancer\nregistries throughout Australia. Of the 3550 women\nidentified with suspected ovarian cancer, 307 died\nbefore contact could be made, physicians refused to\ngive consent to contact 133, usually because they were\ntoo sick or unable to give informed consent and 194\nwomen could not be contacted. A further 171 (5%) were\nexcluded on the basis of language difficulties (70),\nmental incapacity (35) and illness (66). The remaining\n2745 women with a clinically suspected diagnosis of\novarian cancer were invited to participate (prior to\nsurgery, to facilitate fresh tissue collection) and, of\nthese, 2319 (85%) agreed to take part. After surgery,\npathology reports were obtained for all women and a\nfurther 608 women were excluded because their final\ndiagnosis was a benign, non-epithelial or metastatic\ntumour and not primary epithelial ovarian cancer, 25\nbecause their cancer was first diagnosed before the start\nof the study period and one women was excluded\nbecause she was not an Australian resident at the time of\nher initial diagnosis. Two researchers independently\nabstracted information on tumour site, histological\nsubtype and tumour behaviour (invasive versus border-\nline) from the diagnostic histopathology reports.\nDiscrepancies were resolved by consensus. To check\nthe quality of the abstracted data, the pathology reports\nand the full set of diagnostic slides for a sample of 200\nwomen were reviewed by a gynaecologic pathologist;\nagreement with the abstracted data was O95% for\ntumour subtype and site, and 99% for tumour behaviour.\nOf the final 1685 eligible participants, 1591 (94%)\nreturned a questionnaire.\nControls were randomly selected from the national\nelectoral roll (enrolment is compulsory) and were\nfrequency matched by age (in 5-year age bands) and\nstate of residence to the case group. Selected women\nwere mailed an invitation l etter and information\nbrochure explaining the study and then, where\npossible, followed up by telephone. Of the 3613\nwomen contacted and invited to participate, 171\nwomen were excluded due to illness (63), language\ndifficulties (97) and death (11). Of the remaining 3442\nwomen, 1613 agreed to participate and returned a\nquestionnaire (47%). Six of them reported a history of\novarian cancer and 99 reported a previous bilateral\noophorectomy and thus were excluded from the present\nstudy leaving 1508 population controls.\nThis study was approved by the Human Research\nEthics Committees at the Peter MacCallum Cancer\nCentre, Queensland Institute of Medical Research,\nUniversity of Melbourne, the Cancer Councils of New\nSouth Wales, South Australia and Victoria, the Cancer\nFoundation of Western Australia and all participating\nhospitals.\nData collection\nAfter obtaining written informed consent, information\nwas collected by a self-administered questionnaire that\nincluded questions about the demographic, medical,\nhormonal, reproductive, diet, family history and other\npotential risk factors for ovarian cancer. Women self-\nreported ever having a range of medical conditions\nincluding PCOS, severe acne as an adult, or excess\nbody hair (face, chest or abdomen). The questionnaire\nincluded detailed questions about the use of hormone\nreplacement therapy (tablets, implants, patches and\ngels/creams/pessaries) and other hormonal treatments\nC M Olsen et al.: Ovarian cancer and the ‘androgens hypothesis’\nwww.endocrinology-journals.org1062\nDownloaded from Bioscientifica.com at 06/13/2026 12:40:48AM\nvia free access\n\n\nfrom which we were able to derive ever-use of\ntestosterone or Danazol. Conditions or medication\nuse after a reference date (defined as 1 year before the\ndate of diagnosis for cases or date of first approach for\ncontrols) were excluded as they might have been\ninfluenced by the presence of preclinical disease.\nStatistical analysis\nMultivariable logistic models were used to adjust for\npotential confounders, including age at diagnosis/first\napproach, education, parity, hormonal contraceptive\nuse and BMI. Other potential confounders that were\nconsidered for all analyses but not included in the final\nmodels since they did not substantially alter risk\nestimates were: state of residence, perineal talc use,\nhistory of hysterectomy or tubal sterilization, family\nhistory of breast or ovarian cancer in a first-degree\nrelative, smoking, breastfeeding, menopausal status\nand level of recreational physical activity. For the\nanalyses of PCOS, acne, hirsutism, testosterone\nsupplements and Danazol use, the reference group\nwas defined as women with no reported history of\nPCOS, hirsutism or acne. We also created a combined\nvariable ‘any androgen-related disorder’ that included\nwomen who had a self-reported history of PCOS or\nacne or hirsutism.\nWe conducted analyses for all tumour types\n(invasive and borderline) firstly for all histological\nsubtypes combined, and then by subtype. We simul-\ntaneously compared invasive and borderline cases with\ncontrols using polytomous logistic regression. We\nexamined effect modification between the main\nvariables of interest and hormonal contraceptive use,\nmenopausal status and parity. The statistical signi-\nficance of any observed stratum-specific differences\nwas assessed by including a cross-product term in\nregression models. To assess potential biological\ninteraction between BMI and PCOS, we created a\nnew variable that re-classified women according to\ntheir combined exposure to obesity and PCOS. Risks\nfor each category of combined exposure were\nestimated relative to the reference category (normal\nweight and no PCOS) in multivariable logistic\nregression analyses. All statistical analyses were\nperformed using SAS version 9.1 (SAS Institute Inc.,\nCary, NC, USA).\nResults\nBased on the histopathology review, 1276 women had\ninvasive cancer classified as follows: serous 847\n(66%), endometrioid 142 (11%), clear cell 90 (7%),\nmucinous 42 (3%) and mixed or other histopathology\n155 (12%). A further 315 women had borderline (low\nmalignant potential) tumours classified as serous 152\n(48%), mucinous 151 (48%) and other 12 (4%). Cases\nwith mixed or other histopathology were excluded\nfrom the analyses by subtype. Descriptive statistics of\nthe study population are presented in Table 1 . Cases\nwere significantly older than controls (mean age: cases,\n57.9 years; controls, 56.4 years; PZ0.001) and were\nless likely to have continued their education beyond\nhigh school. Cases were more likely to be nulliparous\nand to report a history of breast or ovarian cancer in a\nfirst-degree relative, and were less likely to have ever\nused oral contraceptives.\nA history of PCOS was reported in 52 women; 130\nreported a history of acne and 197 of hirsutism. Of the\nTable 1 Descriptive characteristics of 1591 women with\nepithelial ovarian cancer and 1508 randomly selected popu-\nlation-based controls\nVariable\nControlsa\n(nZ1508) n (%)\nCasesa\n(nZ1591) n (%) P valueb\nAge\n!30 42 (3) 37 (2)\n30–39 112 (7) 87 (5)\n40–49 279 (18) 262 (16)\n50–59 450 (30) 485 (30)\n60–69 399 (26) 463 (29)\n70C 226 (15) 257 (16) 0.08\nHighest level of education\nHigh school 740 (49) 866 (54)\nTechnical\ncollege/trade\ncertificate\n550 (36) 509 (32)\nUniversity 218 (14) 216 (14) 0.01\nNumber of pregnancies (R6 months)\nNulliparous 180 (12) 302 (19)\n1–2 645 (43) 651 (41)\nR3 683 (45) 635 (40) !0.0001\nEver used oral contraceptives\nNo 324 (22) 511 (33)\n%5 years 365 (24) 438 (28)\nO5 years 813 (54) 623 (40) !0.0001\nHistory of breast or ovarian cancer in a first-degree relative\nYes 196 (13) 276 (17)\nNo 1312 (87) 1315 (83) 0.0008\nBody mass index one year ago\n!18.5 33 (2) 30 (2)\n!18.5–24.9 662 (44) 585 (41)\n25–29.9 453 (30) 472 (33)\nR30 341 (23) 348 (24) 0.22\nBody mass index at age 20 years\n!18.5 216 (15) 227 (15)\n!18.5–24.9 1090 (74) 1093 (73)\n25–29.9 119 (8) 113 (9)\nR30 40 (3) 43 (3) 0.84\naNumbers may not sum to total because of missing data.\nbc2 test for heterogeneity.\nEndocrine-Related Cancer (2008) 15 1061–1068\nwww.endocrinology-journals.org 1063\nDownloaded from Bioscientifica.com at 06/13/2026 12:40:48AM\nvia free access\n\n\nwomen with PCOS, 12 of these also reported a history\nof hirsutism and 3 of acne. There was no overall\nassociation between PCOS, hirsutism or acne and all\ninvasive cancers combined, but weak positive associa-\ntions with PCOS and hirsutism were seen for border-\nline tumours (OR 1.8, 95% CI 0.8–3.9 and OR 1.5 95%\nCI 0.9–2.3 respectively; Table 2 ). The differences\nbetween the invasive and borderline tumours were\nstatistically significant for PCOS ( PZ0.049) and\nborderline significant for hirsutism ( PZ0.05). When\nthe histological subtypes were examined separately, a\nself-reported history of PCOS was associated with\nserous borderline tumours only (OR 2.5, 95% CI\n1.0–6.1; Table 3). There was no significant association\nbetween self-reported acne or hirsutism and any of the\nsubtypes of ovarian cancer although, as for borderline\ntumours overall, serous borderline tumours were non-\nsignificantly associated with hirsutism (OR 1.5, 95%\nCI.0.8–2.7). In our combined variable, history of any\nandrogen-related disorder (PCOS or acne or hirsutism)\nwas not associated with any of the subtypes of ovarian\ncancer (data not shown).\nTable 4 considers the combined effects of PCOS and\nBMI (1 year prior to diagnosis) separately for all\ncancers, and by invasiveness and histology. There was\nno evidence of biological interaction between PCOS\nand BMI for invasive cancer. For borderline tumours,\nthe OR was 0.9 (95%CI 0.2–4.4) for non-obese women\nwith PCOS compared with non-obese women without\nPCOS, but rose to 3.0 (95% CI 1.2–7.5) for obese\nwomen with PCOS. This increased risk again appeared\nto be restricted to the single subtype of serous\nborderline tumours (OR 5.7; 95% CI 2.1–15.7 for\nobese women with PCOS). We did not observe any\nsignificant effect modification by OC use, menopausal\nstatus or parity.\nEleven cases (nine invasive and two borderline) and\nseven control women reported use of Danazol. After\nadjustment for age, education, parity, hormonal contra-\nceptive use and self-reported endometriosis, the OR for\nthe association between Danazol use and ovarian\ncancer (all cases) was 1.0 (95% CI 0.4–2.9). Eleven\ncases (all invasive) and four control women reported\never-use of testosterone (t ablets, patches, troches\nand cream). The odds ratio for the association with\ninvasive cancer (after adjustment for age, education,\nparity and hormonal contraceptive use) was 3.7 (95%\nCI 1.1–12.0).\nDiscussion\nIn this large population-based case-control study, we\nfound no evidence that self-reported histories of either\nPCOS, acne or hirsutism (all clinically associated with\nhigh circulating levels of androgens) were associated\nwith increased ovarian cancer risk overall, although\nwomen with PCOS, who were also overweight, had a\nsignificantly increased risk of serous borderline\ntumours. Women who had ever used testosterone\nsupplements had an increased risk of ovarian cancer\nwhile the use of the androgenic medication Danazol\ndid not increase risk.\nAs discussed by Risch (1998) , several lines of\nevidence suggest a possible aetiologic role for elevated\nandrogens in the pathogenesis of ovarian cancer.\nAndrogen receptors have been detected in normal\nand neoplastic ovarian epithelial cells ( Kuhnel et al .\n1987, Chadha et al . 1993 , Ilekis et al . 1997 , Cardillo\net al .1 9 9 8 , Lau et al .1 9 9 9 , Modugno 2004 ),\ngonadotropins and androgens that stimulate the\nproliferation of normal and malignant human ovarian\nepithelial cells in vitro (Syed et al . 2001 , Edmondson\net al . 2002 , Modugno 2004 , Stewart et al . 2004 ), and\novarian cancer cell growth is inhibited in vitro by anti-\nandrogens (Slotman & Rao 1989). Animal experiments\nhave demonstrated that testosterone can enhance the\ngrowth of ovarian epithelial tumours ( Sawada et al .\n1990, Tennent et al . 1993 , Silva et al . 1997 ).\nOther indirect evidence suggesting a possible\naetiologic role for elevated androgens in the initiation\nand/or progression of ovarian cancer has come from\nTable 2 Multivariable adjusteda odds ratios (OR) and 95% confidence intervals (CI) of epithelial ovarian cancer for history of\nandrogen-related disorders, by tumour invasiveness\nAll cases Invasive Low malignant potential\nAndrogen-related disorderb Controls Cases OR a (95% CI) Cases OR a (95% CI) Cases OR a (95% CI)\nPolycystic ovary syndrome 24 28 1.1 (0.6–2.0) 16 0.8 (0.4–1.6) 12 1.8 (0.8–3.9)\nHirsutism 96 101 1.0 (0.7–1.3) 72 0.9 (0.6–1.2) 29 1.5 (0.9–2.3)\nAcne 68 62 0.9 (0.6–1.3) 51 1.0 (0.7–1.5) 11 0.6 (0.3–1.2)\naAdjusted for age, education, parity, hormonal contraceptive use, BMI 1 year ago.\nbReference group was women with no self-reported history of PCOS, acne or hirsutism.\nC M Olsen et al.: Ovarian cancer and the ‘androgens hypothesis’\nwww.endocrinology-journals.org1064\nDownloaded from Bioscientifica.com at 06/13/2026 12:40:48AM\nvia free access\n\n\nepidemiologic studies ( Risch 1998 ). Firstly, it is well\nknown that oral contraceptives, which suppress ovarian\ntestosterone production ( Gaspard et al . 1983 , Murphy\net al . 1990, Greer et al . 2005), protect against ovarian\ncancer. Secondly, there have been a number of case\nreports of ovarian cancer in female-to-male transsex-\nuals who have undergone testosterone supplementation\n(Hage et al. 2000, Dizon et al. 2006), although the true\nincidence of the disease in this population is not yet\nknown. Other potentially supportive evidence has\ncome from a small number of population-based case-\ncontrol studies. Schildkraut et al . (1996) observed a\n2.5-fold increased risk of ovarian cancer among\nwomen with PCOS, although the analysis was based\non a small number of women with PCOS (7 cases and\n24 controls). Acne and hirsutism have also been\nassociated with ovarian cancer ( Wynder et al . 1969 ),\nwhile Cottreau et al . (2003) reported that after\nadjusting for age, gravidity, OC use and family history\nof ovarian cancer, women who used Danazol ( nZ19)\nhad over three times the risk for ovarian cancer\ncompared with non-users.\nOur data do not confirm these previous findings. We\ndid not find any overall association with PCOS, acne or\nhirsutism, and our results for Danazol, based on a\nsimilar number of users ( nZ18), did not confirm the\nfindings of Cottreau et al . (2003) . We did observe a\nrelationship between PCOS and borderline serous\ntumours and can only speculate as to the reason for\nthis. Current evidence derived from molecular and\nTable 3 Multivariable adjusteda odds ratios (OR) and 95% confidence intervals (CI) of epithelial ovarian cancer for history of\nandrogen-related disorders, by tumour invasiveness and histological subtype\nInvasive cancer Borderline tumours\nAndrogen-related\nSerous\n(nZ847)\nEndometrioid\n(nZ142)\nClear cell\n(nZ90)\nMucinous\n(nZ42)\nSerous\n(nZ152)\nMucinous\n(nZ151)\ndisorderb Controls OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI)\nPolycystic ovary\nsyndrome (PCOS)\n24 1.0 (0.5–2.0) 2.5 (1.0–6.1) 0.7 (0.1–3.0)\nHirsutism 96 0.7 (0.5–1.1) 1.3 (0.7–2.4) 1.2 (0.5–2.8) 1.4 (0.5–3.9) 1.5 (0.8–2.7) 1.3 (0.7–2.4)\nAcne 68 0.9 (0.6–1.5) 1.6 (0.7–3.3) 0.7 (0.2–2.4) 1.3 (0.4–4.5) 0.6 (0.2–1.6) 0.7 (0.3–1.7)\nEstimates are not presented for the relationship between PCOS and endometrioid, clear cell and mucinous invasive cancers due to\ninsufficient numbers of exposed cases.\naAdjusted for age, education, parity, hormonal contraceptive use and BMI 1 year ago.\nbReference group was women with no self-reported history of PCOS, acne or hirsutism.\nTable 4 Multivariable adjusted a odds ratios (OR) and 95% confidence intervals (CI) for the risk of epithelial ovarian cancer\naccording to the combined effect of history of polycystic ovary syndrome (PCOS) and body mass index (BMI; reference group is\nBMI!25, no PCOS)\nNo PCOS PCOS\nControls (n) Cases ( n) OR (95% CI) Controls ( n) Cases ( n) OR (95% CI)\nAll cases\nBMI 1 year ago !25 684 600 1.0 (Ref) 10 6 0.7 (0.3–2.0)\nBMI 1 year ago R25 777 786 1.2 (1.0–1.4) 14 19 1.7 (0.8–3.4)\nInvasive cases\nBMI 1 year ago !25 684 480 1.0 (Ref) 10 4 0.6 (0.2–2.1)\nBMI 1 year ago R25 777 639 1.2 (1.0–1.4) 14 10 1.1 (0.5–2.6)\nBorderline cases\nBMI 1 year ago !25 684 120 1.0 (Ref) 10 2 0.9 (0.2–4.4)\nBMI 1 year ago R25 777 147 1.2 (0.9–1.6) 14 9 3.0 (1.2–7.5)\nSerous borderline\nBMI 1 year ago !25 684 48 1.0 (Ref) 10 1 1.1 (0.1–8.9)\nBMI 1 year ago R25 777 79 1.7 (1.1–2.4) 14 7 5.7 (2.1–15.7)\nMucinous borderline\nBMI 1 year ago !25 684 68 1.0 (Ref) 10 0 –\nBMI 1 year ago R25 777 63 0.9 (0.7–1.4) 14 2 1.1 (0.2–5.5)\naAdjusted for age, education, parity and hormonal contraceptive use.\nEndocrine-Related Cancer (2008) 15 1061–1068\nwww.endocrinology-journals.org 1065\nDownloaded from Bioscientifica.com at 06/13/2026 12:40:48AM\nvia free access\n\n\ngenetic studies suggests that the borderline and\ninvasive serous tumours develop through independent\npathways (Singer et al. 2003, Shih Ie & Kurman 2004 ,\nBell 2005 ), and thus the endocrine consequences of\nPCOS may have differential effects on the patho-\ngenesis of these different tumour types. In addition,\nPCOS is a complex disorder associated with alterations\nin endogenous sex hormone levels, suppression of\novulation, infertility and a number of metabolic\ndisorders including insulin resistance (Solomon\n1999). This complex array of conditions make it\ndifficult to assess the effect of increased androgen\nlevels on ovarian cancer risk in isolation, and may also\nexplain why we did not observe an association for\nsome subtypes. We observed a threefold increased risk\nof ovarian cancer for women who had ever used\ntestosterone supplements, and although our analysis\nwas based on a very small number of women, this\nfinding is potentially interesting and warrants further\ninvestigation.\nFour prospective studies have examined the associ-\nation between circulating androgens and the risk of\novarian cancer (Helzlsouer et al. 1995, Lukanova et al.\n2003, Rinaldi et al . 2007 , Tworoger et al . 2007 ).\nAlthough the first study, based on only 31 exposed\ncases, found an increasing risk of ovarian cancer with\nincreasing levels of androstenedione and dehydroe-\npiandrosterone ( Helzlsouer et al . 1995 ), other larger\nand more recent studies have not observed an\nassociation between prediagnostic androgens and\novarian cancer risk ( Lukanova et al . 2003 , Tworoger\net al .2 0 0 7). Rinaldi et al . (2007) observed an inverse\nassociation between free testosterone concentrations\nand ovarian cancer risk in postmenopausal women only\n(192 cases); however, other studies have not confirmed\nthis finding, and our data do not suggest a differential\neffect by menopausal status. The results from\nprospective studies of circulating androgen levels are\nthus generally null, possi bly because circulating\nandrogens may not reflect androgen exposure at the\ntissue level. The ovarian epithelium is not vascular and\nthus paracrine hormonal influences may be more\nimportant than endocrine sources ( Lukanova &\nKaaks 2005 ).\nStrengths of our study include the population-based\ndesign, large number of cases and detailed infor-\nmation on multiple exposures. A limitation was the\nrelatively low participation rate among controls\n(47%), which could have resulted in selection bias;\nhowever, a comparison with the data from the\nAustralian National Health Survey (NHS) conducted\nin 2004 (a representative survey of the Australian\nadult population; ABS 2006 )r e v e a l e dt h a tt h e\ndistributions of education level, parity and BMI\namong our control women were almost identical to\nthose from the NHS ( Jordan et al . 2007 ), and it is\ntherefore unlikely that non-response could have\nresulted in appreciable bias. Another limitation was\nreliance upon self-reported medical history of PCOS,\nhirsutism and acne, and medication use. PCOS is a\ncomplex condition that may not be accurately reported\nby women. There may also be asymptomatic women\nwith PCOS in our study population ( Polson et al .\n1988, Azziz et al . 2004 ); however, such under-\nreporting of PCOS would most likely have been\nrandom, and probably would have resulted in bias\ntowards the null. It is possible that the cases were\nmore likely to recall a history of PCOS and also use of\nmedications/hormonal pre parations than controls;\nhowever, this cannot explain the observed association\nbetween PCOS and the minority of borderline cases\nbut not invasive cases. Women are unlikely to\nassociate hirsutism and acne with ovarian cancer and\ntherefore any misclassification of these conditions is\nexpected to be non-differential.\nIn summary, we found no consistent evidence for a\nrole of androgens in the aetiology of ovarian cancer,\noverall or by subtype, and thus our findings do not\nsupport the hypothesis that androgen-related disorders\nincrease the risk of ovarian cancer. Although labora-\ntory studies have suggested a role for androgens in the\ndevelopment of ovarian cancer, there is very little\nepidemiological evidence to support an association.\nStudies are also constrained by small sample sizes, and\nthere is heterogeneity both in the types of exposure\nmeasures reported and the research findings. The\nresults from prospective studies of circulating andro-\ngen levels are generally null. Large collaborative\nanalyses are required to examine the associations\nbetween markers of high androgen levels and risk of\novarian cancer subtypes, and future research should\ntarget the relative roles of endocrine versus paracrine\nandrogen sources.\nDeclaration of interest\nThe authors declare that there is no conflict of interest that\ncould be perceived as prejudicing the impartiality of the\nresearch reported.\nFunding\nThe Australian Ovarian Cancer Study was supported by the\nUS Army Medical Research and Materiel Command under\nDAMD17-01-1-0729, The Cancer Council Tasmania and\nThe Cancer Foundation of Western Australia; The Australian\nCancer Study was supported by the National Health and\nC M Olsen et al.: Ovarian cancer and the ‘androgens hypothesis’\nwww.endocrinology-journals.org1066\nDownloaded from Bioscientifica.com at 06/13/2026 12:40:48AM\nvia free access\n\n\nMedical Research Council of Australia (199600). David\nWhiteman and Penelope Webb are supported by Senior\nResearch Fellowships from the National Health and Medical\nResearch Council of Australia. Christina Nagle and Susan\nJordan are supported by Postdoctoral Public Health Training\nFellowships from the National Health and Medical Research\nCouncil of Australia. Catherine Olsen is supported by a\nUniversity of Queensland Postdoctoral Fellowship.\nAcknowledgements\nWe gratefully acknowledge the cooperation of the following\ninstitutions: New South Wales: John Hunter Hospital, North\nShore Private Hospital, Royal Hospital for Women, Royal\nNorth Shore Hospital, Royal Prince Alfred Hospital,\nWestmead Hospital, New South Wales Cancer Registry;\nQueensland: Mater Misericordiae Hospital, Royal Brisbane\nand Women’s Hospital, To wnsville Hospital, Wesley\nHospital, Queensland Cancer Registry; South Australia:\nFlinders Medical Centre, Queen Elizabeth II, Royal Adelaide\nHospital, South Australian Cancer Registry; Tasmania:\nRoyal Hobart Hospital; Victoria: Freemasons Hospital,\nMercy Hospital For Women, Monash Medical Centre,\nRoyal Women’s Hospital, Victorian Cancer Registry;\nWestern Australia: King Edward Memorial Hospital, St\nJohn of God Hospitals Subiaco, Sir Charles Gairdner\nHospital, Western Australia Research Tissue Network\n(WARTN), Western Australia Cancer Registry. We also\nacknowledge the contribution of the study nurses and\nresearch assistants and would like to thank all of the\nwomen who participated in the study. Full membership of the\nAustralian Ovarian Cancer Study Group is listed at http://\nwww.aocstudy.org/; the Australian Cancer Study Investi-\ngators are A Green, P Parsons, N Hayward, P Webb and D\nWhiteman.\nReferences\nABS 2006 2004-05 National Health Survey: summary of\nresults, vol. Cat. No. 4364, pp 92. Australian Bureau of\nStatistics.\nAzziz R 2003 Androgen excess is the key element in polycystic\novary syndrome. Fertility and Sterility 80 252–254.\nAzziz R, Woods KS, Reyna R, Key TJ, Knochenhauer ES &\nYildiz BO 2004 The prevalence and features of the\npolycystic ovary syndrome in an unselected population.\nJournal of Clinical Endocrinology and Metabolism 89\n2745–2749.\nBell DA 2005 Origins and molecular pathology of ovarian\ncancer. Modern Pathology 18 (Supplement 2) S19–S32.\nCardillo MR, Petrangeli E, Aliotta N, Salvatori L, Ravenna L,\nChang C & Castagna G 1998 Androgen receptors in\novarian tumors: correlation with oestrogen and pro-\ngesterone receptors in an immunohistochemical and\nsemiquantitative image analysis study. Journal of Experi-\nmental & Clinical Cancer Research 17 231–237.\nChadha S, Rao BR, Slotman BJ, van Vroonhoven CC & van\nder Kwast TH 1993 An immunohistochemical evaluation\nof androgen and progesterone receptors in ovarian tumors.\nHuman Pathology 24 90–95.\nCottreau CM, Ness RB, Modugno F, Allen GO & Goodman\nMT 2003 Endometriosis and its treatment with danazol or\nlupron in relation to ovarian cancer. Clinical Cancer\nResearch 9 5142–5144.\nCoulam CB, Annegers JF & Kranz JS 1983 Chronic\nanovulation syndrome and associated neoplasia. Obste-\ntrics and Gynecology 61 403–407.\nDizon DS, Tejada-Berges T, Koelliker S, Steinhoff M &\nGranai CO 2006 Ovarian cancer associated with\ntestosterone supplementation in a female-to-male trans-\nsexual patient. Gynecologic and Obstetric Investigation\n62 226–228.\nEdmondson RJ, Monaghan JM & Davies BR 2002 The\nhuman ovarian surface epithelium is an androgen\nresponsive tissue. British Journal of Cancer 86 879–885.\nGambineri A, Pelusi C, Vicennati V, Pagotto U & Pasquali R\n2002 Obesity and the polycystic ovary syndrome.\nInternational Journal of Obesity and Related Metabolic\nDisorders 26 883–896.\nGaspard UJ, Romus MA, Gillain D, Duvivier J, Demey-\nPonsart E & Franchimont P 1983 Plasma hormone levels\nin women receiving new oral contraceptives containing\nethinyl estradiol plus levonorgestrel or desogestrel.\nContraception 27 577–590.\nGreer JB, Modugno F, Allen GO & Ness RB 2005 Androgenic\nprogestins in oral contraceptives and the risk of epithelial\novarian cancer. Obstetrics and Gynecology 105 731–740.\nHage JJ, Dekker JJ, Karim RB, Verheijen RH & Bloemena E\n2000 Ovarian cancer in female-to-male transsexuals:\nreport of two cases. Gynecologic Oncology 76 413–415.\nHelzlsouer KJ, Alberg AJ, Gordon GB, Longcope C, Bush\nTL, Hoffman SC & Comstock GW 1995 Serum\ngonadotropins and steroid hormones and the development\nof ovarian cancer. Journal of the American Medical\nAssociation 274 1926–1930.\nIlekis JV, Connor JP, Prins GS, Ferrer K, Niederberger C &\nScoccia B 1997 Expression of epidermal growth factor\nand androgen receptors in ovarian cancer. Gynecologic\nOncology 66 250–254.\nJordan SJ, Green AC, Whiteman DC & Webb PM 2007 Risk\nfactors for benign, borderline and invasive mucinous\novarian tumors: epidemiological evidence of a neoplastic\ncontinuum? Gynecologic Oncology 107 223–230.\nKuhnel R, de Graaff J, Rao BR & Stolk JG 1987 Androgen\nreceptor predominance in human ovarian carcinoma.\nJournal of Steroid Biochemistry 26 393–397.\nLau KM, Mok SC & Ho SM 1999 Expression of human\nestrogen receptor-alpha and -beta, progesterone receptor,\nand androgen receptor mRNA in normal and malignant\novarian epithelial cells. PNAS 96 5722–5727.\nLucky AW 1995 Hormonal correlates of acne and hirsutism.\nAmerican Journal of Medicine 98 89S–94S.\nEndocrine-Related Cancer (2008) 15 1061–1068\nwww.endocrinology-journals.org 1067\nDownloaded from Bioscientifica.com at 06/13/2026 12:40:48AM\nvia free access\n\n\nLukanova A & Kaaks R 2005 Endogenous hormones and\novarian cancer: epidemiology and current hypotheses.\nCancer Epidemiology, Biomarkers and Prevention 14\n98–107.\nLukanova A, Lundin E, Akhmedkhanov A, Micheli A,\nRinaldi S, Zeleniuch-Jacquotte A, Lenner P, Muti P,\nBiessy C, Krogh V et al. 2003 Circulating levels of sex\nsteroid hormones and risk of ovarian cancer. International\nJournal of Cancer 104 636–642.\nMerritt MA, Green AC, Nagle CM & Webb PM 2008 Talcum\npowder, chronic pelvic inflammation and NSAIDs in\nrelation to risk of epithelial ovarian cancer. International\nJournal of Cancer 122 170–176.\nModugno F 2004 Ovarian cancer and polymorphisms in\nthe androgen and progesterone receptor genes: a HuGE\nreview. American Journal of Epidemiology 159\n319–335.\nMurphy A, Cropp CS, Smith BS, Burkman RT & Zacur HA\n1990 Effect of low-dose oral contraceptive on gonado-\ntropins, androgens, and sex hormone binding globulin in\nnonhirsute women. Fertility and Sterility 53 35–39.\nOlive DL & Pritts EA 2001 Treatment of endometriosis. New\nEngland Journal of Medicine 345 266–275.\nPierpoint T, McKeigue PM, Isaacs AJ, Wild SH & Jacobs HS\n1998 Mortality of women with polycystic ovary syndrome\nat long-term follow-up. Journal of Clinical Epidemiology\n51 581–586.\nPolson DW, Adams J, Wadsworth J & Franks S 1988\nPolycystic ovaries – a common finding in normal women.\nLancet 1 870–872.\nPurdie DM, Webb PM, Siskind V, Bain CJ & Green AC 2003\nThe different etiologies of mucinous and nonmucinous\nepithelial ovarian cancers. Gynecologic Oncology 88\nS145–S148.\nRinaldi S, Dossus L, Lukanova A, Peeters PH, Allen NE,\nKey T, Bingham S, Khaw KT, Trichopoulos D,\nTrichopoulou A et al. 2007 Endogenous androgens and\nrisk of epithelial ovarian cancer: results from the\nEuropean Prospective Investigation into Cancer and\nNutrition (EPIC). Cancer Epidemiology, Biomarkers and\nPrevention 16 23–29.\nRisch HA 1998 Hormonal etiology of epithelial ovarian\ncancer, with a hypothesis concerning the role of\nandrogens and progesterone. Journal of the National\nCancer Institute 90 1774–1786.\nRisch HA, Marrett LD, Jain M & Howe GR 1996 Differences\nin risk factors for epithelial ovarian cancer by histologic\ntype. Results of a case-control study. American Journal of\nEpidemiology 144 363–372.\nSawada M, Terada N, Wada A, Mori Y, Yamasaki M, Saga T\n& Endo K 1990 Estrogen- and androgen-responsive\ngrowth of human ovarian adenocarcinoma heterotrans-\nplanted into nude mice. International Journal of Cancer\n45 359–363.\nSchildkraut JM, Schwingl PJ, Bastos E, Evanoff A & Hughes\nC 1996 Epithelial ovarian cancer risk among women with\npolycystic ovary syndrome. Obstetrics and Gynecology\n88 554–559.\nShih Ie M & Kurman RJ 2004 Ovarian tumorigenesis: a\nproposed model based on morphological and molecular\ngenetic analysis. American Journal of Pathology 164\n1511–1518.\nSilva EG, Tornos C, Fritsche HA Jr, el-Naggar A, Gray K,\nOrdonez NG, Luna M & Gershenson D 1997 The\ninduction of benign epithelial neoplasms of the ovaries\nof guinea pigs by testosterone stimulation: a\npotential animal model. Modern Pathology 10 879–883.\nSinger G, Oldt R III, Cohen Y, Wang BG, Sidransky D,\nKurman RJ & Shih Ie M 2003 Mutations in BRAF and\nKRAS characterize the development of low-grade ovarian\nserous carcinoma. Journal of the National Cancer\nInstitute 95 484–486.\nSlotman BJ & Rao BR 1989 Response to inhibition of\nandrogen action of human ovarian cancer cells in vitro .\nCancer Letters 45 213–220.\nSolomon CG 1999 The epidemiology of polycystic ovary\nsyndrome. Prevalence and associated disease risks.\nEndocrinology and Metabolism Clinics of North America\n28 247–263.\nStewart SL, Querec TD, Gruver BN, O’Hare B, Babb JS\n& Patriotis C 2004 Gonadotropin and steroid hormones\nstimulate proliferation of the rat ovarian surface\nepithelium. Journal of Cellular Physiology 198\n119–124.\nSyed V, Ulinski G, Mok SC, Yiu GK & Ho SM 2001\nExpression of gonadotropin receptor and growth\nresponses to key reproductive hormones in normal and\nmalignant human ovarian surface epithelial cells. Cancer\nResearch 61 6768–6776.\nTennent BJ, Shultz KL & Beamer WG 1993 Genetic\nsusceptibility for C19 androgen induction of ovarian\ngranulosa cell tumorigenesis in SWXJ strains of mice.\nCancer Research 53 1059–1063.\nTitus-Ernstoff L, Perez K, Cramer DW, Harlow BL, Baron\nJA & Greenberg ER 2001 Menstrual and reproductive\nfactors in relation to ovarian cancer risk. British\nJournal of Cancer 84 714–721.\nTworoger SS, Lee IM, Buring JE & Hankinson SE 2007\nPlasma androgen concentrations and risk of incident\novarian cancer. American Journal of Epidemiology 167\n211–218.\nWynder EL, Dodo H & Barber HR 1969 Epidemiology of\ncancer of the ovary. Cancer 23 352–370.\nC M Olsen et al.: Ovarian cancer and the ‘androgens hypothesis’\nwww.endocrinology-journals.org1068\nDownloaded from Bioscientifica.com at 06/13/2026 12:40:48AM\nvia free access","source_license":"CC0","license_restricted":false}