Phenotypic features of polycystic ovary syndrome and their covariates among Bengali young adult females assessed by three major diagnostic guidelines

Introduction

Polycystic ovary syndrome (PCOS) is one of the most common health problems among women of reproductive age,1 with widely varying global prevalence between 2.2% and 48%.2 A trend towards the increasing prevalence of PCOS since the late 1900s has been demonstrated in various studies.2 The wide variations in prevalence reported by various authors are partly explained by the use of different criteria,3 4 namely, the National Institutes of Health (NIH) criteria,4 the Modified Rotterdam criteria 20035 and the Androgen Excess Society (AES) criteria.6 The basic diagnostic features in these criteria include anovulation or oligomenorrhoea, polycystic ovarian morphology (POM) and hyperandrogenism (HA); however, the NIH criteria suggest the mandatory presence of first and third, the Rotterdam criteria suggest the presence of any two, and the AES criteria suggest the mandatory presence of the third plus optional presence of any of the first and second phenotypic features for the confirmed diagnosis of PCOS.

Apart from the issue of varying diagnostic criteria, genuine differences seem to exist among ethnic groups regarding the prevalence and metabolic characteristics of PCOS. From the ethnicity-specific guidelines, it is revealed that the prevalence of women with PCOS differs among Chinese, White and Black women.7 8 Adverse metabolic features have been found to be more common among South Asian, Asian Indian, African American and Hispanic women with PCOS. In contrast, Middle Eastern and Mediterranean women have been found to be more hirsute in general.7 A multi-country study observed a maximum predisposition to metabolic syndrome (MetS) among Indian women, followed by US and Norwegian women with PCOS.9 The influence of geographic variation on phenotypic manifestation is further highlighted by a study on Indian women from two regions of Northern India. It shows distinct phenotypes in two cities of North India—lean hyperandrogenic women from Srinagar and obese hyperinsulinaemic women from Delhi.10 According to the National Health Portal of India, the prevalence rate of PCOS in Maharashtra was noted to be 22.5%.11 Another previous report from South India, which included adolescents, showed an incidence of PCOS at 9.13%.12

Reported studies on PCOS epidemiology in Bangladesh are almost exclusively facility-based.13 Except for one,14 all the studies were single-facility based, and the subjects attended the relevant clinical departments with specific complaints, namely infertility,14–17 hirsutism18 and acne.19 The prevalence of the disease ranged from 6.11% (among the subjects visiting the gynaecology-obstetrics outpatient department (OPD)) to 92.16% (in subjects consulted for hirsutism).13 The prevalence was 29.9%–46.15% among infertile women.14–17 Except for three studies,16 18 20 which used Rotterdam criteria, the others did not mention any criteria used to diagnose the disorder. Only two studies categorised subjects with PCOS according to phenotypic features.21 22

Studies specifically targeted to young Bangladeshi women in non-clinical community settings are limited to only one,20 and even that was among students of a single medical college. The study reported a 37% prevalence of PCOS among the medical students. Further studies, using standardised diagnostic criteria, are required to get more insight into the burden and risk factors of PCOS among young Bangladeshi women. This is an optimum period in the life cycle of women to initiate preventive measures for avoiding the future development of PCOS. Thus, data on this group of women will facilitate the design of evidence-based, context-sensitive management and prevention policies for PCOS and also the planning of genotype-related studies on the disease among this subpopulation.

Under the above context, the present study has been designed to explore the phenotypic features of PCOS and MetS-related anthropometric-clinical-biochemical profiles among a young adult subgroup of Bangladeshi women. A secondary objective is to categorise the phenotypic features as per different diagnostic criteria and explore their concordance in terms of the prevalence of the disorder.

Materials and methods

This observational analytic study with a case–control design was conducted among participants (age 17–34 years), recruited through purposive sampling, from the primary responders to a social media call and distribution of printed leaflets among residents of the targeted sites. The origin of responders included residential halls of the universities/colleges, OPDs of tertiary care hospitals and urban families in Dhaka city. The participants were classified into the case group (PCOS) and the control group (non-PCOS). The case group comprised 158 PCOS subjects diagnosed according to the Modified Rotterdam Criteria 2003,5 which requires the presence of at least two of the following three findings: (1) menstrual abnormalities (amenorrhoea, oligomenorrhoea), (2) clinical and biochemical HA and (3) the ultrasound appearance of polycystic ovaries. In the non-PCOS group, 127 non-hirsute women (without clinical evidence of HA) had regular menstrual cycles (without anovulation); they did not show PCOS’s required hormonal and sonographic features. The sample size was calculated using the OpenEpi software,23 assuming a 38% prevalence of obesity among PCOS women, which is the minimum value reported for the Bangladeshi population in a recent review.13 The two-sided CI was 95% and the power of the study was set at 80%. Women with pregnancy, lactation, hyperprolactinaemia, Cushing syndrome, thyroid dysfunction, congenital adrenal hyperplasia and androgen-secreting tumours were excluded from the study. Use of insulin-sensitising or glucose-reducing agents, hormonal treatments, oral contraceptive pills, antihypertensives, lipid-lowering drugs or corticosteroids 3 months before the study was considered a criterion of exclusion from the study.

Using an interviewer-administered pretested Questionnaire cum Data Collection Form developed based on the Modified Rotterdam criteria 2003,5 all the relevant information was collected. Personal and medical history was taken, anthropometric measurements were done, and blood pressure (BP) was measured following standard techniques. Anthropometric measurements included body weight, height and waist circumference. Height and weight were scaled with the subjects in light clothes and without shoes. Waist circumference was evaluated using a flexible tape at the midline between the lower rib border and the curved superior border of the ilium (at the level of the umbilicus); with participants in the standing position, the measurement was done at the end of a normal exhalation. Following WHO guidelines, body mass index (BMI) was computed, and the categorisation of the study participants was done as per guidelines for Asian women.24 Using a mercury sphygmomanometer, BP was measured in the right arm after a 10 min rest period; the subjects were in a non-fasting state, wearing loose sleeves, had depleted bladders and avoided eating, drinking (except water) or smoking for at least 1 hour before the test. MetS was defined by the presence of three or more of the following five features: abdominal obesity (waist circumference, male >40”, female >35”); hypertriglyceridaemia (triglycerides (TG), ≥150 mg/dL); low levels of high-density lipoprotein (HDL) cholesterol (HDL-C) (male <40, female <50, mg/dL); elevated blood glucose level (fasting, >5.5 mmol/L) and high BP (≥130/85 mm Hg).

The participants were questioned about the regulation of their menstrual cycle and they were subjected to a clinical examination to evaluate hirsutism based on the Ferriman-Gallwey score (FG-score), with a value ≥8 being considered as clinical HA. Lack of menstrual cycle for >3 months or a cycle duration of >35 days was considered an ovulation disorder, denoted by the term Oligomenorrhoea/anovulation. A subject was preliminarily placed in the control group when both of these factors were normal; once the subsequent results of ultrasound and serum hormonal tests were also not characteristic of PCOS, she was confirmed as a non-PCOS subject in the study. Acanthosis nigricans was evaluated by the presence of black velvety patches in body folds and creases; it was used as a surrogate marker for insulin resistance. For each participant, lower abdominal ultrasonography (Voluson E6, USA), with folliculometry, was done by a trained professional; ovaries containing 12 or more follicles measuring 2–9 mm in diameter and/or enlarged ovarian volume (>10 mm³) were considered to have a positive polycystic sonographic view.6

Following an overnight fast (10–12 hours), each subject underwent a standard 75 g oral glucose tolerance test; 0-hour and 2-hour postprandial serum samples were preserved at relevant freezers for future glucose and lipid profile estimation as well as for hormonal analyses. Serum glucose was measured by Glucose Oxidase and serum lipids (total cholesterol (T-Chol), TG, HDL-C and low-density lipoprotein cholesterol (LDL-C)) were estimated by enzymatic methods using an automated chemistry analyser (Abbott, USA). The serum total testosterone (TT) levels, SHBG and C-peptide were measured by chemiluminescence immunoassay using the same autoanalyser. Free Androgen Index (FAI) was computed by the formula TT (nmol/l)/SHBG (nmol/l) × 100; a value >7.1% was considered a marker of biochemical HA following the study by Nadaraja et al25 conducted on a Malaysian population. With this cut-off value of FAI, the authors reported an AUC of 0.79 on ROC analysis; the sensitivity was found to be 62.2% and the specificity was 82.2%.

Results

The present data show that the proportion of subjects with PCOS may differ significantly among young Bangladeshi women depending on the criteria used to diagnose the disorder. The diagnosis rate is the highest (158/285; 55%) with Modified Rotterdam 2003 criteria, and it is the lowest (124/285; 43%) with NIH criteria, while the AES criteria show an intermediate value (145/285; 51%). In line with these findings, a better agreement (k value 0.90) has been found between Rotterdam and AES criteria on the Kappa test, whereas the agreement between Rotterdam and NIH is the lowest (k value 0.76) in this group of subjects (table 1a). The NIH and AES criteria show an intermediate agreement (k value, 0.86, table 1b).

The presence of characteristic features of PCOS (anovulation/oligomenorrhoea, HA, hirsutism, POM) was significantly higher in the PCOS as compared with the non-PCOS groups in the case of using any of the three diagnostic criteria (p≤0.001 in all cases, table 2). MetS was present in around 8% of non-PCOS and 12% of PCOS subjects, irrespective of the diagnostic criteria used; the difference between the groups was significant (figure 1). FAI showed a significant correlation with fasting blood glucose (FBG) in all the criteria-based diagnostic groups (p<0.45 to <0.001) (table 3).

As expected, the proportion of acanthosis nigricans and infertility was higher among PCOS compared with non-PCOS subjects using all three criteria (table 2). Also, the median values of WHR, 2-hour-after Blood Glucose (2hABG), T-Chol, TG, LDL-C and TT were significantly higher (p<0.045 to p<0.001). The PCOS groups showed significantly higher age when using the Rotterdam and NIH criteria (p<0.05 -<0.008) and BMI showed a significantly higher value only when using the AES criteria (p=0.036). FBG was not significantly different between the PCOS and Non-PCOS groups when using any diagnostic criteria (table 2).

On binary logistic regression analysis within the PCOS group, anovulation/oligomenorrhoea was found to be independently associated with WHR only when the Rotterdam or AES criteria were followed, but not when the NIH criteria were used (table 4). No variable was found to be independently associated with POM in any criteria-based PCOS groups on binary logistic regression analysis (table 5). On multiple regression analysis (table 6), FAI was found to be significantly associated with BMI (in NIH and Rotterdam criteria-based PCOS groups), WHR (in AES-based group), FBG (in NIH and Rotterdam-based groups), 2-hour ABG (in Rotterdam- and AES-based groups)

Discussion

Many of the reported diversities in the epidemiology of PCOS are rooted in the use of different guidelines applying varying criteria for the diagnosis of the disorder.26 27 Also, it is possible that the same criteria may not be accurate in diagnosing the condition in different subgroups of women and in the same woman over the different stages of her lifespan. The diagnosis becomes especially complicated in the case of young/early adult women, as some of the challenges confronted among adolescents28 29 may remain as residues in this age group. In particular, the possibility of the presence of physiological anovulation among the younger adult women may introduce certain difficulties in the diagnosis of PCOS. In the present study, we analysed the phenotypic features of 285 young Bangalee women who were initially grouped into Non-PCOS and PCOS following the Modified Rotterdam criteria and later reassessed following NIH and AES criteria. The data suggest that, in the present young subgroup of women, an agreement between Modified Rotterdam and AES criteria is ‘Very Good’ and that between Modified Rotterdam and NIH criteria is ‘Good’.30 A ‘Very Good’ agreement is also obtained between NIH and AES; however, the exact level of agreement is slightly lower than that between the Modified Rotterdam and AES groups. Accordingly, caution is necessary to interpret evidence during clinical practice, especially when multiple providers are involved. It is important to create some consensus within a country setting and generate subgroup-specific guidelines for diagnosing, managing and preventing PCOS. The Rotterdam and AES criteria may be used interchangeably in the case of these young subjects; however, special attention to underdiagnosis should be provided when using the NIH criteria.

The present data indicate a special importance of POM in the diagnosis of PCOS in these young subjects. Using the Modified Rotterdam criteria, clinical or biochemical HA has been found among 69% PCOS subjects (table 7). On the other hand, POM has been detected among 84% PCOS subjects. The corresponding proportion is higher in the case of menstrual abnormality (88%); however, considering the possibility of physiological anovulation in the younger subjects, this phenotypic feature may have some limitations, and the more objective finding of POM may be given central importance in clinical practice involving such women. Clinical HA was detected in this study by hirsutism using FG scoring, which has been shown to have ethnic variations.26 High sensitivity of ultrasound POM criteria in young women has been reported before.31

Covariates of PCOS have been reported to differ substantially among various populations.7–10 The present data suggest that some of these diversities may be due to the use of different criteria. When the Non-PCOS and PCOS groups are compared, the typical anthropometric and clinical features (like BMI, WHR, acne, alopecia, acanthosis nigricans and infertility) are found to be associated with PCOS irrespective of the guideline used. Also, hyperglycaemia, dyslipidaemia and MetS show similar types of association when using different guidelines. A noteworthy feature is the highly significant correlation of BMI with FAI in young non-PCOS women diagnosed by all three criteria. This may reinforce the importance of weight control in preventing further HA in these women. Another notable feature is the consistent negative correlation (p<0.014 to 0.008) of HDL with FAI in the PCOS groups on using all three diagnostic criteria. Accordingly, hypo-HDL-cholesterolaemia seems to be an important covariate of HA in these young women. The importance of HDL has been demonstrated in a recent large-scale population-based study in Iran,32 where only this lipid subtype has been found to be inversely associated with PCOS when the effects of age and BMI were adjusted.

Differences emerge among diagnostic criteria, especially when the individual phenotypic features are analysed for their association with individual covariates after adjusting for the confounding variables. Generalised obesity does not seem to be a significant factor associated with anovulation/oligomenorrhoea in these young women; however, central obesity shows an association with these phenotypic abnormalities when the Rotterdam or AES criteria are used, but not when using the NIH criteria. In contrast, generalised obesity is associated with HA only when using NIH and Rotterdam criteria, and central obesity is associated with HA only when using the AES criteria (table 3). The role of generalised obesity vs fat distribution patterns is still debated and a suggestion has been given that, regarding metabolic impact in women with PCOS, body weight alone or BMI may not be as crucial as fat distribution, especially the adverse effects of centripetal obesity.32 It also appears that if a patient has oligomenorrhoea and HA in adolescence, there is an increased risk of developing obesity by the age of 24 years. This suggests a temporal association of PCOS with obesity, even if a primary predisposition does not exist.25 33 Further studies are required to explore these issues in relation to the criteria–specific differences between generalised and central obesity in the present study.

Fasting hyperglycaemia shows a higher degree of association with HA when using NIH and Rotterdam criteria; however, postprandial hyperglycaemia is more associated with HA in the case of AES and Rotterdam. The significance of these findings still needs to be better understood through further large-scale and preferably prospective studies. This, and many other clarifications on population and subpopulation-specific PCOS issues, require international collaborative efforts to generate and synthesise evidence as suggested in a recent thematic review.34 In the Bangladesh context, the diagnostic criteria of PCOS are not even mentioned in most of the publications.13 It raises the possibility that enough attention is not paid to the application of standardised diagnostic criteria for the disorder in the clinical settings of the country. This should be a matter of major clinical concern and efforts should be taken to implement standardised criteria (preferably the modified Rotterdam criteria, since it is now mostly regarded as the global standard) uniformly in all settings. It may, however, be generally inferred that individual covariates of PCOS, irrespective of the criteria used, need proper attention when designing disease prevention and management strategies.

The limitations of the cross-sectional design and relatively small sample size do not allow for a deeper analysis of the causalities and implications of the discrepancies found in the present study, between various diagnostic criteria of PCOS. Analysis based on individual phenotypes of PCOS was also not possible due to the limited sample size. The purposive sampling technique may also introduce some bias in the study, especially in favour of the inclusion of pre-existing PCOS subjects. However, the present data offer a starting point for understanding PCOS among young Bangladeshi women and raise the necessity of further research for robust conclusions on relevant issues.

Conclusions

In conclusion, the present study demonstrates better concordance between Rotterdam and AES criteria in the diagnosis of PCOS among young Bangalee women and it also draws attention to the differential association of body weight and hyperglycaemia with anovulation/oligomenorrhoea and HA when one or other diagnostic criteria are used. PCOS may remain underdiagnosed among young Bangalee women if NIH criteria are used. In general, interpreting epidemiological data on PCOS, even in the same ethnic group, needs careful consideration of the criteria used to diagnose the disease as well as individual covariates of the disease in the specific person.