Diagnostic criteria for polycystic ovary syndrome

The polycystic ovary syndrome (PCOS) was first coined as a term in 1935 by Stein and Leventhal, who identified a series of patients with the combination of oligo-ovulation and hyperandrogenism [1]. Since that time, diagnosis of this syndrome has undergone multiple iterations of application of diagnostic criteria. Abnormal uterine bleeding was the most common symptom associated with the condition in the late 1800s, but over time, new and better evidence has become available to inform the diagnosis. As recognition of this entity as a complex constellation of symptoms and metabolic derangements expanded, multiple professional groups have attempted to better and more precisely characterize this common syndrome.

In the past 25 years, three major sets of diagnostic criteria have emerged to define PCOS (Table 7.1) [2–4]. The first set of criteria, outlined at the National Institutes of Health (NIH) in Bethesda, Maryland, in 1990, is the most stringent. It has largely been replaced in clinical practice by the more inclusive Rotterdam criteria. In Rotterdam, The Netherlands, in 2003, a task force sponsored by the European Society of Human Reproduction and Embryology (ESHRE) and the American Society for Reproductive Medicine (ASRM) met to review the available data and to propose a revision of the 1990 NIH diagnostic paradigm. More recently, in 2009, the Androgen Excess Society (AES) outlined its own set of criteria for PCOS. Due to the subtle heterogeneities between the various diagnostic criteria, the reported prevalence of PCOS can vary depending on which definition is used by investigators. The discriminating reader should take note of the specific criteria utilized in the reporting.

The NIH meeting in 1990 was the first international conference on PCOS. The information available to attendees at the time was based largely on expert opinion [2]. The criteria set forth included (1) chronic anovulation and (2) clinical or biochemical signs of hyperandrogenism. Both criteria must be present under NIH guidelines, and other diagnoses must be excluded. This initial step to clearly define the syndrome provided a backbone upon which the ensuing body of new analytic research could be built. A large volume of new studies thus emerged, providing additional information subsequently evaluated by the Rotterdam ESHRE/ASRM-sponsored PCOS Consensus Workshop Group to revise the original NIH proposed set of diagnostic criteria.

The Rotterdam consensus includes three diagnostic criteria, and states that any two of the three must be present in order to make the diagnosis [3]. The revised criteria included (1) oligo- or anovulation, (2) clinical or biochemical signs of hyperandrogenism, and (3) polycystic-appearing ovaries (PCO) on imaging. The diagnosis depends upon the exclusion of other causes of hirsutism and anovulatory bleeding, including nonclassic congenital adrenal hyperplasia (NC-CAH), Cushing’s syndrome, and androgen-secreting tumors, as well as some other more common entities such as thyroid dysfunction and hyperprolactinemia. By including morphological appearance of polycystic ovaries, the Rotterdam consensus added two additional phenotypes not previously included in the diagnosis: (1) women with ovulatory dysfunction and polycystic ovaries but without hyperandrogenism and (2) ovulatory women with hyperandrogenism and polycystic ovaries. By including these phenotypes, PCOS was canonized as a spectrum, with associated long-term health risk implications, such as type 2 diabetes mellitus and cardiovascular disease consequences. The Rotterdam consensus statement advocated widening the inclusion criteria to avoid missing patients with the potential for these increased health risks. Deeper explorations reveal that these subcategories within PCOS, identified based on the Rotterdam diagnostic criteria, manifest as subtle but distinct hormonal and metabolic milieus when compared with cases of PCOS identified based on the more stringent NIH criteria.

The most recent set of diagnostic criteria released was from the AES in 2009 [4]. This consensus group reexamined the key features of PCOS, including menstrual dysfunction, hyperandrogenemia, clinical signs of hyperandrogenism, and polycystic ovarian morphology. A thorough review of existing literature was used to examine each feature for its appropriateness for inclusion as a defining criterion. A slightly modified version of the criteria for the diagnosis of PCOS emerged in this process: (1) hyperandrogenism, including hirsutism and/or hyperandrogenemia, (2) ovarian dysfunction, including oligo-anovulation and/or polycystic appearing ovaries, and (3) exclusion of other androgen excess or related disorders. Under the AES’s PCOS criteria, related disorders of hyperandrogenism or anovulation also must be excluded, such as androgen-secreting neoplasms, Cushing’s syndrome, congenital adrenal hyperplasia, thyroid disorders, hyperprolactinemia, and premature ovarian failure. The clinician may take into account the prevalence of these differential diagnoses when deciding what test to order. Similar to the NIH criteria, androgen excess is an essential component of the diagnosis by AES criteria. Therefore, the phenotype of ovulatory dysfunction and polycystic ovaries (PCO) alone – permissible under Rotterdam – does not qualify for a diagnosis of the syndrome by AES criteria. The combination of menstrual dysfunction and polycystic ovaries, in the absence of features of hyperandrogenism or evidence of hyperandrogenemia has, in fact, been shown to have the most similar anthropometric, hormonal, and metabolic risk factors to control subjects without PCOS matched for age and body mass index. The AES consensus criteria for defining PCOS are thus more inclusive than the NIH version, but less inclusive than the Rotterdam criteria.

Anti-müllerian hormone (AMH) has recently been proposed as a biochemical marker to replace ultrasonographic assessment of PCO morphology. Specificity and sensitivity of 97.1% and 94.6% when using the Rotterdam criteria, or 97.2% and 95.5% using the NIH criteria, have been reported [5]. AMH levels correlate independently with both PCO morphology and androgenic profile [6]. Ovarian stromal volume, measured as a ratio of the stromal area to total area of the ovary (S/A ratio), has been proposed as another ultrasonographic parameter of the ovarian morphology highly associated with hyperandrogenemia. Although this S/A ratio performs well when discriminating between women with and without PCOS, and correlates well with androgen levels, it has not been adopted as part of any of the existing diagnostic criteria [7,8].

Both the clinical and biochemical determination of hyperandrogenism in females can be challenging. Laboratory assays for androgens were initially designed for detection in males, and have been calibrated accordingly. For example, total testosterone assays are typically calibrated for normal male levels, the lower end of which is 250 ng/dL. The upper end of normal female total testosterone ranges between 55 ng/dL and 80 ng/dL (inter-laboratory differences exist and clinicians should familiarize themselves with the assay range for the laboratories serving their patient population). The female normal range occurs well below the fifth percentile for the assay detection range commonly utilized in men. At these lower levels, results may be less reliable; notably in most commercial assays. An additional diagnostic dilemma is that the reporting of clinical hyperandrogenism is examiner-dependent and can be quite subjective. While a standard tool such as the Ferriman–Gallwey score is frequently applied in an attempt to create an objective evaluation, this method has been shown to have better intra-observer reliability than inter-observer reliability [9]. Furthermore, a universal application of such tools across all ethnic groups may discount the normal ethnic variation in the appearance of body hair.

As evidenced by the variation among diagnostic criteria, inclusion of ultrasonographic evidence of PCO morphology into the definition of PCOS is controversial. The NIH criteria do not address ovarian morphology. The Rotterdam criteria in 2003 included polycystic ovaries as a phenomenon distinct from menstrual irregularities. The AES groups ovarian morphology into an “ovarian dysfunction” category along with oligo-anovulation, and they require only one or the other to suffice as a diagnostic criterion. It is important to appreciate that PCO morphology is not specific to PCOS and can be found in 20–30% of the general population of women 20–25 years of age. Isolated PCO, therefore, should not be considered an indication of the syndrome in the absence of menstrual irregularities, infertility, or complaints of hirsutism [10].

In some ways, efforts to agree on diagnostic criteria are artificial. There continues to be controversy and lack of complete agreement for what elements constitute optimal criteria for PCOS diagnosis, in part because of the natural clinical desire to move to discrete categorical criteria for the ease of diagnosis. In truth, there is a continuum of presentation from those persons minimally affected, with regular menses and only mild excess of androgens, to those who have a unilateral PCO, to those who manifest more severe grades of androgen excess. Efforts to include hyperandrogenemia as diagnostic criteria will remain inadequate until the sensitivity of androgen assays is better refined because of our current inability to accurately quantify circulating androgens in women. It is useful to remember that androgen production is affected by metabolic clearance and concentrations, and often people with normal androgen levels in the circulation have increased androgen production. It is simply not practical to measure androgen production, so we are limited to imperfect representation of hyperandrogenemia using current outpatient widely available laboratory measurements in the blood.