Use of Community-based Reference Ranges to Estimate the Prevalence of Polycystic Ovary Syndrome by the Recognised Diagnostic Criteria, a Cross-sectional Study

Marina A. Skiba; Robin J. Bell; Dilinie Herbert; Alejandra Martinez Garcia; Rakibul M. Islam; Susan R. Davis


Hum Reprod. 2021;36(6):1611-1620. 

In This Article

Abstract and Introduction


Study Question: Does the application of reference ranges for sex steroids and the modified Ferriman-Gallwey (mFG) scale established in the community from which the study sample was drawn, combined with the most conservative polycystic ovary morphology (PCOM) criteria to the recognised diagnostic criteria for polycystic ovary syndrome (PCOS) improve the certainty of diagnosis of PCOS in non-healthcare-seeking women?

Summary Answer: Despite application of the stringent definitions of the elements used to diagnose PCOS in a non-healthcare seeking community-based sample, the risk of diagnostic uncertainty remains.

What is Known Already: There is heterogeneity in prevalence estimates for PCOS due, in part, to lack of standardisation of the elements comprising the recognised National Institutes of Health (NIH), Rotterdam and Androgen Excess Society (AE-PCOS) diagnostic criteria. The AE-PCOS Society proposed refinements to the definitions of biochemical androgen excess and PCOM that can now be incorporated into these sets of diagnostic criteria to estimate PCOS prevalence.

Study Design, Size, Duration: An Australian cross-sectional study of 168 non-healthcare-seeking women.

Participants/Materials, Setting, Methods: The 168 included women were aged 18–39 years, euthyroid and normoprolactinemic, not recently pregnant, breast feeding or using systemic hormones. Each provided menstrual history and assessment of the mFG, had measurement of sex steroids by liquid chromatography, tandem mass spectrometry, and a pelvic ultrasound. The presence of PCOS was determined using modified (m) NIH, Rotterdam, and AE-PCOS criteria according to AE-PCOS Society recommendations.

Main Results and the Role of Chance: Overall, 10.1% of the included participants met the mNIH PCOS criteria, which requires the presence of menstrual dysfunction, while 18.5% met the mRotterdam and 17.5% the AE-PCOS criteria, with the latter requiring hyperandrogenism. Eight of the 27 participants with menstrual dysfunction, 10 of 31 women with PCOM, and 39 of 68 women with hyperandrogenism had no other feature of PCOS. Of the 19 participants with hyperandrogenaemia, 10 met the mNIH criteria (52.5%) and 14 met both the mRotterdam and AE-PCOS criteria (78.9%). Women who had the combination of hyperandrogenism and PCOM explained the greatest discrepancy between the mNIH and the other criteria.

Limitations, Reasons for Caution: Clinical androgenisation relied on participant self-assessment, which has been shown to be valid when compared with clinician assessment. The sample size was a function of both the strict inclusion criteria and the requirements of non-healthcare-seeking women having a blood draw and pelvic ultrasound which may have introduced a selection bias.

Wider Implications of the Findings: Despite applying stringent cut-offs for serum androgens, the mFG scale and the ovarian follicle count, these criteria remain arbitrary. Accordingly, healthy women may be captured by these criteria, and misidentified as having PCOS, while women with the condition may be missed. Consequently, PCOS remains a diagnosis to be made with care.

Study Funding/Competing Interest(S): The study was supported by the Grollo-Ruzzene Foundation. Dr S.R.D. is an NHMRC Senior Principal Research Fellow (Grant no. 1135843). S.R.D. has been paid for developing and delivering educational presentations for Besins Healthcare, BioFemme and Pfizer Australia, has been on Advisory Boards for Theramex, Abbott Laboratories, Mayne Pharmaceuticals and Roche and a consultant to Lawley Pharmaceuticals and Que Oncology and has received has received institutional grant funding for Que Oncology research; there are no other relationships or activities that could appear to have influenced the submitted work.

Trial Registration Number: N/A


Polycystic ovary syndrome (PCOS) is frequently referred to as the most common endocrine disorder in women. Agreement as to what exactly constitutes this condition remains elusive as the main elements comprising the three different sets of diagnostic criteria in common usage (Zawadski and Dunaif, 1992; Rotterdam ESHRE/ASRM-Sponsored PCOS Consensus Workshop Group, 2004; Azziz et al., 2009) were developed through expert opinion and lack definition (Skiba et al., 2018). The National Institutes of Health (NIH) classification requires the presence of oligo-anovulation and clinical or biochemical androgen excess, but these elements were not defined when this diagnostic framework was developed in 1992 (Zawadski and Dunaif, 1992). The subsequent 2004 Rotterdam criteria proposed that elevated free testosterone or the free androgen index (total testosterone nmol/l/sex hormone binding globulin nmol/l × 100) were indicative of biochemical androgen excess, as well as acne and male pattern baldness as possible indicators of clinical hyperandrogenism (Rotterdam ESHRE/ASRM-Sponsored PCOS Consensus Workshop Group, 2004). The Rotterdam Workshop also added polycystic ovary morphology (PCOM) as a diagnostic element (Rotterdam ESHRE/ASRM-Sponsored PCOS Consensus Workshop Group, 2004). PCOM was defined at that time as at least 12 follicles (2–9 mm in diameter) in a single ovary or an ovarian volume >10 ml for at least one ovary (Rotterdam ESHRE/ASRM-Sponsored PCOS Consensus Workshop Group, 2004). The Androgen Excess-PCOS Society (AE-PCOS, 2009) criteria (Azziz et al., 2009) added short cycles as evidence of anovulation, but considered acne and alopecia to be unreliable indicators of androgen excess. Furthermore, the AE-PCOS Society decided that PCOS is a disorder of androgen excess, and, therefore, a woman must have either clinical androgenisation or biochemical hyperandrogenism, or both, for a diagnosis of PCOS by their criteria (Azziz et al., 2009).

We previously found that the pooled PCOS prevalence estimate using the NIH criteria (7%) to be significantly different from that identified from studies that used the Rotterdam criteria (12%), but not different from ones that used the AE-PCOS criteria (10%) (Skiba et al., 2018). The prevalence studies included in our meta-analysis had applied a variety of interpretations of the diagnostic criteria. The definition of menstrual or ovarian dysfunction was inconsistent and androgen excess was assessed in some studies by the Ferriman-Gallwey (FG) (Ferriman and Gallwey, 1961) scoring system and in others by the modified FG (mFG) system (Ferriman and Gallwey, 1961), with a range of cut-offs used to classify women as having hirsutism, or simply excess hair growth (Skiba et al., 2018). Serum androgens were mostly measured by immunoassay with an array of definitions of hyperandrogenaemia (Skiba et al., 2018). The definition of PCOM was inconsistent, as was the approach to inclusion or exclusion of hormonal contraceptive users (Skiba et al., 2018). As eloquently summarised previously (Geisthovel and Rabe, 2007), uncertainty as to the prevalence of PCOS in women in the community persists.

The AE-PCOS Society has made specific recommendations to refine the determination of PCOS. This includes establishing a reference range in a sample of unselected women from the general community and defining hyperandrogenaemia as a serum concentration above the 95th percentile of a reference range (Azziz et al., 2009). In 2014, a Taskforce of the AE-PCOS Society developed an evidenced-based, revised definition of PCOM to be used with high-resolution transvaginal ultrasound (TVU) (Dewailly et al., 2014). The Taskforce recommended that a follicle number per ovary (FNPO) of ≥25 be considered diagnostic of PCOM with high resolution TVU, and that an ovarian volume ≥10 ml in either ovary not be used in that setting, but might represent PCOM with low resolution imaging (Dewailly et al., 2014). To date, PCOS prevalence estimates using these more rigorous diagnostic elements have not been reported.

The purpose of this study was to examine the likely prevalence of PCOS in a community-based sample using reference ranges for sex steroids and the mFG established in the community from which the women in this study were drawn, and the most conservative PCOM criteria established to date. The recent International PCOS Guideline recommended the use of the Rotterdam criteria to diagnose PCOS (Teede et al., 2018). The guideline highlighted the need for further research to refine the diagnostic criteria. Therefore, we have explored each of the three recognised criteria to demonstrate how using even more conservative cut-offs for the diagnostic criteria might influence the estimated prevalence. To do so, we established reference ranges for sex steroids measured by liquid chromatography and tandem mass spectrometry (LC-MS/MS) in a sample of 588 regularly menstruating, healthy, premenopausal Australian women recruited from the community (Skiba et al., 2019a). We then documented the distribution of self-assessed mFG scores in a representative sample of 6986 Australian women aged 18–39 years and identified a mFG score cut-off that was indicative of excess hair growth (Skiba et al., 2020). Having established sex steroid reference ranges and a mFG score cut-off for Australian women we now report the prevalence of PCOS in a sample of non-healthcare seeking women using the AE-PCOS Society recommendations incorporated into each of the NIH, Rotterdam, and AE-PCOS criteria.