Are expanding disease definitions unnecessarily labelling women with polycystic ovary syndrome?BMJ 2017; 358 doi: https://doi.org/10.1136/bmj.j3694 (Published 16 August 2017) Cite this as: BMJ 2017;358:j3694
- Tessa Copp, PhD student12,
- Jesse Jansen, cognitive psychologist and senior research fellow12,
- Jenny Doust, GP and professor of clinical epidemiology34,
- Ben WJ Mol, professor of obstetrics and gynaecology56,
- Anuja Dokras, professor of obstetrics and gynaecology7,
- Kirsten McCaffery, behavioural scientist and professorial research fellow12
- 1Wiser Healthcare, Sydney School of Public Health, University of Sydney, NSW 2006, Australia
- 2Centre for Medical Psychology and Evidence-Based Decision-Making (CeMPED), University of Sydney, NSW 2006, Australia
- 3Wiser Healthcare, Bond University, QLD 4229, Australia
- 4Centre for Research in Evidence-Based Practice, Bond University, QLD 4229, Australia
- 5Robinson Research Institute, School of Paediatrics and Reproductive Health, University of Adelaide, SA 5005, Australia
- 6The South Australian Health and Medical Research Institute, Adelaide, Australia
- 7Penn PCOS Centre, Department of Obstetrics and Gynaecology, University of Pennsylvania, Philadelphia, US
- Correspondence to: J Jansen
Clinical context—Polycystic ovary syndrome (PCOS) is the most commonly diagnosed endocrine disorder affecting reproductive aged women and is associated with adverse reproductive, metabolic, and cardiovascular outcomes
Diagnostic change—The Rotterdam consensus expanded the criteria from the National Institutes of Health (NIH) to include polycystic ovaries, introducing a number of different PCOS phenotypes
Rationale for change—Early identification, diagnosis, and management of PCOS could prevent the adverse long term implications by implementing lifestyle modifications and screening for comorbidities
Leap of faith—Diagnosis will provide relief to symptom burden and allow implementation of lifestyle changes to improve cardiovascular risk and increase fertility
Increase in disease—Estimated number of diagnoses in reproduction aged women increased from 4-6.6% for NIH criteria to 21% for Rotterdam criteria, depending on the population studied
Evidence of overdiagnosis—Diagnoses have rapidly increased, the criteria are of questionable use in adolescents and young women, and the non-hyperandrogenic phenotypes of PCOS do not have the same associated adverse implications as the hyperandrogenic phenotypes
Harms from overdiagnosis—In addition to taking long term medications, labelling healthy women with PCOS might negatively impact their psychological health and wellbeing, inducing fear and anxiety about future fertility and long term health
Uncertainties—Suggestion of overdiagnosis is based on the rise in diagnoses and the uncertainties that exist in the diagnosis, progression, and treatment of PCOS
Conclusion—A PCOS label might not be needed to effectively treat many symptoms of PCOS, as the label often does not change the type or intensity of the intervention. We recommend carefully weighing up the benefits and harms for each individual woman and taking a slower, stepped, or delayed approach to diagnosis to optimise benefits and reduce harm from disease labelling
Polycystic ovary syndrome (PCOS) is the most commonly diagnosed endocrine disorder affecting women of reproductive age and is associated with infertility and adverse reproductive, metabolic, cardiovascular, and psychosocial outcomes.12 Women with PCOS can experience anovulation, menstrual irregularities, ovaries appearing polycystic, and signs of androgen excess, which can vary by weight, ethnicity, and environmental factors and can change across the lifespan.34 These symptoms cluster and vary in severity.1
Diagnostic criteria have expanded over time, and the number of women diagnosed has risen,5 raising concerns about potential overdiagnosis through unnecessary disease labelling. We discuss the potential areas of overdiagnosis and consider the evidence and uncertainty surrounding diagnosis and overdiagnosis of PCOS.
PCOS was first described in 1935 in a case series of seven women with amenorrhoea and infertility associated with multiple cysts in the ovaries.6 Today, three different diagnostic criteria for PCOS are in use (table 1⇓).789
The National Institutes of Health (NIH) criteria were established in 1990.17 In 2003 at a meeting of experts in Rotterdam, sonographic presence of polycystic ovaries was added (table 1⇑).18 In 2006 a taskforce of the Androgen Excess and PCOS Society said that clinical or biochemical evidence of hyperandrogenism was essential for diagnosis, as the non-hyperandrogenic phenotypes do not have the same long term risks as the hyperandrogenic phenotypes.9
Both the NIH and Rotterdam criteria are based on expert majority opinion and have been criticised owing to insufficient high quality evidence on long term follow-up or therapeutic benefit to derive an evidence based definition of the syndrome.58 Studies that report associations between PCOS and long term complications or treatment benefit often do not distinguish between the different phenotypes, so which women will benefit from or might be at greater need of treatment is unclear. Most studies have small samples of women recruited from specialist clinics, who typically have more severe symptoms and the classic NIH phenotypes A-F.12 This might overestimate long term morbidities if generalised to all women with PCOS.513
Including polycystic ovaries as one of the three key diagnostic criteria has been criticised14 because they are present in many women without PCOS.114 Features of polycystic ovaries on ultrasonography have been found in 62-84% of women aged 18-30 in the general population and in 7% of women aged 41-45 years.1415
Rationale for change
The criteria were expanded to reflect the broad clinical expression of PCOS,1 but this assumes that a PCOS diagnosis, no matter how mild, reduces the risk of associated comorbidities and improves fertility.
PCOS is associated with endometrial carcinoma, obesity, type 2 diabetes mellitus, hypertension, complications of pregnancy (gestational diabetes mellitus, pre-eclampsia), and possibly cardiovascular disease .916217 For some women, such as those with severe hyperandrogenism, early diagnosis and management of PCOS could arguably slow progression of comorbidities and prevent long term harms by prompting early intervention. A PCOS label might, for example, help women undergo regular screening for comorbidities and make dietary and lifestyle changes, such as losing weight and reducing risk of diabetes or cardiovascular disease (table 3⇓).2122
Diagnosis might also enable the use of drugs, such as oral contraceptive pills to regulate the menstrual cycle, reduce hirsutism, and lower the risk of endometrial cancer or metformin to improve insulin sensitivity and potentially lower risk of impaired glucose tolerance and diabetes.23 Women with PCOS who are seeking pregnancy might benefit from knowledge of their ovulatory dysfunction to aid family planning and treatment, including ovulation induction drugs such as clomiphene citrate, letrozole, and metformin.424 Interventions aimed at weight loss improve pregnancy and live birth rates in women with PCOS.25
Impact on prevalence
The widening of the diagnostic criteria by the Rotterdam definition contributed to a rise in estimated PCOS prevalence in reproductive aged women from 4-6.6% using NIH criteria2627 to 21% using Rotterdam criteria,528 depending on the population studied (table 4⇓).
This rise coincides with the inclusion of more phenotypes, along with increases in obesity,36 disease awareness, and sensitivity of tests. Overdiagnosis in other conditions tends to result in a rise in milder cases, but in PCOS the phenotypes that are increasing are unknown. We have found no quality longitudinal data published since introduction of the Rotterdam criteria. Consequently, the association between rising prevalence and benefit:harm in modern cohorts of women with PCOS is unknown. The proportion of each phenotype has been estimated,5 but these estimates differ between referred clinical samples and unselected PCOS populations.13
Evidence of overdiagnosis
Early diagnosis in adolescence
Symptoms suggestive of PCOS tend to emerge in adolescence. But diagnosis is challenging at this time3738 because the symptoms overlap with common features of pubertal development (such as acne and oligomenorrhoea).1838 None of the diagnostic criteria currently used takes the adolescent population into account.91837 Cut-off points for abnormal androgen concentrations in adolescence have not been established, and using adult cut-off points is likely to be inappropriate.39
Several guidelines exist for diagnosing adolescents, and stricter criteria for adolescents (such as requiring all three Rotterdam criteria) have been proposed by specialists to limit inappropriate diagnoses,3839 but not enough high quality evidence exists to support these models in clinical practice.53740 There is also limited evidence indicating any long term benefit of identifying and managing PCOS in this age group.373841 As a result, some researchers have suggested that adolescents be considered “at risk” and for doctors to refrain from establishing the diagnosis.37
Diagnosis in young adulthood
Diagnosis in early adulthood might also be problematic. A substantial proportion of younger women, such as those with menstrual irregularity and raised free testosterone at one assessment (phenotype F, table 2⇑) might meet the strict NIH definition but can experience spontaneous resolution over time (fig 1)⇓.3 Three studies in different populations found that prevalence of PCOS rapidly fell after 25 years of age, indicating that the symptoms and signs of PCOS might be transitory for many women.33442
Diagnosis of non-hyperandrogenic phenotypes
The Rotterdam criteria widened the NIH definition by including non-hyperandrogenic phenotypes (such as women with menstrual irregularity and ovaries that appear polycystic). Observational data show that these women lack associated long term consequences,1511434445 which might not be communicated to women. Along with the readily available information about PCOS on social media (often of low quality), women might be unaware of their less severe phenotype and think that they have an increased risk for metabolic and cardiovascular outcomes, despite evidence demonstrating otherwise.44 Some researchers have suggested that these women do not represent the syndrome and might overlap with women with hypothalamic amenorrhoea.511
Limitations in the evidence of treatment benefit
Evidence that lifestyle modification and metformin are beneficial has been extrapolated from the effects of large scale randomised controlled trials of lifestyle and treatment interventions in populations with pre-diabetes.46 Existing studies in PCOS populations are often of low quality (open label or small sample sizes) and lack long term follow-up.21 The women tend to be recruited from fertility, gynaecology, or endocrinology clinics, and therefore may have more severe symptoms.1213 Whether these interventions provide long term benefit and how the existing evidence of treatment benefit applies to the different PCOS phenotypes and women with milder symptoms remain uncertain.547
Label may not alter treatment
Women who are overweight with menstrual irregularity and signs of hyperandrogenism do not need a diagnosis of PCOS to be recommended the oral contraceptive pill, fertility counselling, or healthy lifestyle changes.
Although a PCOS diagnosis is thought to motivate compliance with screening recommendations and implementation of lifestyle changes, the data to support this are limited. The uptake of regular screening for associated diseases is not always seen in practice.48 Whether communicating a PCOS diagnosis results in a healthier lifestyle is unknown. A recent synthesis of systematic reviews found that communicating personalised risk information of various conditions does not improve health related behaviour.49
Groups at risk of underdiagnosis
Some women with PCOS who are at high risk of long term harms might not have a diagnosis. Populations with limited access to healthcare or at high risk of obesity and diabetes, such as indigenous populations and Hispanic Americans, are less likely to receive a diagnosis.2850 A lower threshold for screening for PCOS has been suggested for these groups in order to provide access to healthcare and implement lifestyle interventions.4 However, overall, evidence of increased prevalence in these populations is scant.
PCOS is associated with psychological burden
Women diagnosed as having PCOS are more likely to have depression and anxiety, poor self esteem, negative body image, disordered eating, and decreased sexual satisfaction in various clinical and community samples.51525354 Qualitative research also emphasises that PCOS can be a stigmatising condition.55
The association between timing of diagnosis and psychological symptoms is unclear,565758 as is whether these findings vary between phenotypes. Neither do we know whether these negative consequences are due to the condition, its symptoms, or the label of PCOS and its perceived long term implications.12 A recent meta-analysis found that reduced psychological wellbeing in women with PCOS is weakly associated with symptoms that are particularly distressing, such as hirsutism.59 Unnecessarily labelling healthy people as patients, however, can also negatively affect their health and wellbeing.6061
The PCOS label might cause fear and anxiety as it has no cure and is associated with fertility implications,58 an obligation for screening for comorbidities,11 implementing dietary and lifestyle changes, and taking long term medications.38 A recent study found that young women given the PCOS disease label in a hypothetical scenario had higher intentions of undergoing ultrasonography, perceived their condition to be more severe, and had lower self esteem than women not given the label.62 For each individual woman these potential harms need to be weighed against the benefits of a PCOS diagnosis and early intervention, taking into account the limitations of the evidence.
How to move forward
More adequately powered, carefully phenotyped, and multiethnic longitudinal research among women outside of hospital and outpatient settings is needed to validate associations with long term adverse effects. Additionally, studies examining prevalence of the diagnostic criteria at different ages are needed.
Better research is needed to characterise the benefits and harms of treatment for women with severe and milder symptoms. Double blinded randomised controlled trials of long duration with placebo arms in community settings are needed to investigate the value of early treatment, using patient centred outcomes (such as menstrual regularity or hirsutism), not just biochemical markers (such as androgen assays or insulin levels).
Age specific cut-off points are needed for follicle count using modern ultrasonography equipment, as 12 or more follicles (Rotterdam criteria) might be normal in early adulthood.14 In 2014 an expert panel recommended increasing the threshold for defining polycystic ovaries to >24 follicles per ovary owing to advancements in imaging technology,16364 but this recommendation has not yet been adopted in PCOS guidelines.19
Conversations with women must be more transparent, so that they understand the limitations in current evidence on this condition (box 1) and the problems of applying one-size-fits-all diagnostic criteria to heterogeneous presentations of symptoms. Although some women, such as those with classic NIH phenotypes or those with more severe symptoms might benefit from the disease label, others may not need the label for effective and quick treatment of presenting symptoms.
Box 1: PCOS facts for more transparent conversations with women
Many of the signs and symptoms of PCOS can occur in normal women
Diagnosis is harder in younger women
Although the label might seem permanent, we now think that several women with signs and symptoms might “grow out” of the label by the time they are in their 30s
PCOS is linked to lower fertility, diabetes, and dyslipidaemia, but much of our understanding of these risks comes from those with severe signs and symptoms. We do not accurately understand to what extent these risks apply to women with milder forms
Women with normal androgen levels might not have an increased risk of diabetes and metabolic syndrome
Current recommendations for managing the condition surmount to maintaining a healthy lifestyle; for example, keeping weight neither too high nor too low, being fit and active, and avoiding smoking
Although some women with PCOS have difficulty ovulating, they have increased ovarian reserve compared with women without PCOS, which is an indicator of high fertility potential, and is an important positive message
The problem of widening disease definitions is a concern for many other conditions where people with lower risk of consequences receive permanent disease labels and lifelong treatments that might cause more harm than benefit, such as osteoporosis, gestational diabetes, and hypertension.6566 Until we have adequate data, we should provide treatment for symptoms but avoid making a diagnosis, particularly for adolescents and women with the non-hyperandrogenic phenotypes, as they might experience prolonged distress and psychological anguish as a result of the diagnosis. At a minimum, we suggest carefully weighing up the benefits and harms for each individual woman.41 A slower, stepped or delayed approach to diagnosis could be a way of optimising benefits and reducing harm from disease labelling.
We thank Jolyn Hersch and Brooke Nickel for helpful discussion and input regarding the manuscript.
Contributors and sources: TC is a PhD student at the School of Public Health, University of Sydney. JJ is a cognitive psychologist at the School of Public Health, University of Sydney. JD is a general practitioner and professor of clinical epidemiology at the Centre for Research in Evidence-Based Practice, Bond University. BM is a professor of obstetrics and gynaecology at Robinson Research Institute, School of Paediatrics and Reproductive Health, University of Adelaide. AD is a professor of obstetrics and gynaecology at the Department of Obstetrics and Gynecology, University of Pennsylvania. KM is a behavioural scientist at the School of Public Health, University of Sydney. All authors contributed to the concepts and structure of this manuscript. JJ is the guarantor.
Competing interests: We have read and understood BMJ policy on declaration of interests and have no relevant interested to declare.
This article is part of a series on overdiagnosis looking at the risks and harms to patients of expanding definitions of disease and increasing use of new diagnostic technologies.