Metabolic characteristics of women with polycystic ovaries and oligo-amenorrhoea but normal androgen levels: implications for the management of polycystic ovary syndrome
Grants and fellowships funding this work:
NovoNordisk Clinical Research Fellowship (T.M.B.)
Tom Barber, Diabetes Research Laboratories, Oxford Centre for Diabetes, Endocrinology and Metabolism, Churchill Hospital, Old Road, Headington, Oxford OX3 7LJ, UK. Tel.: 01865 857292; Fax: 01865 857299; E-mail: firstname.lastname@example.org
Objective Application of the newly introduced Rotterdam criteria for polycystic ovary syndrome (PCOS) generates four phenotypic subgroups, defined by the presence/absence of three diagnostic elements: polycystic ovarian (PCO) morphology (P); hyperandrogenism (H); and oligo-amenorrhoea (O). Whilst PCOS is associated with adverse metabolic features, the strength of the association within individual subgroups is not established. We characterized the metabolic and endocrine profiles of PCOS women who are oligomenorrhoeic but normoandrogenaemic, and compared these to other PCOS women and controls.
Design Retrospective dataset analyses.
Patients A total of 309 Europid PCOS women, all with PCO morphology, of whom 191 were also hyperandrogenaemic and oligomenorrhoeic (PHO), 76 hyperandrogenaemic with normal menses (PH) and 42 oligomenorrhoeic but normoandrogenaemic (PO); plus 76 Europid control women without PCOS.
Measurements Metabolic parameters: fasting insulin, lipids, homeostasis model assessment (HOMA) measures of insulin sensitivity; endocrine variables: LH, FSH; prevalence of metabolic syndrome.
Results Insulin sensitivity: PO women were indistinguishable from controls, and markedly less insulin-resistant than PHO women (vs. controls, P = 0·38 after adjustment for BMI and age; vs. PHO, P = 0·003). Metabolic syndrome: the prevalence in PO women (7·1%) was similar to that in controls (3·9%), and lower than in PHO women (29·3%, P < 0·0001). LH levels: PO women were intermediate between controls (vs. controls, P = 0·008) and PHO women (vs. PHO, P = 0·06).
Conclusions Normoandrogenaemic, oligomenorrhoeic women with PCOS are metabolically similar to control women with significantly fewer metabolic features than PCOS women who are also hyperandrogenaemic. However, higher than normal LH and lower sex hormone-binding globulin (SHBG) concentrations in the PO women support the view that they form part of the spectrum of PCOS.
Polycystic ovary syndrome (PCOS) is a common endocrinopathy, typically characterized by both reproductive and hyperandrogenic features. Many women with PCOS also display an adverse metabolic profile.1–10 For example, estimates of the prevalence of metabolic syndrome in white women with PCOS from the US (using the National Cholesterol Education Program Adult Treatment Panel III (NCEP ATPIII) criteria), lie between 34% and 46%.1,3,6,7 The high prevalence of adverse metabolic features in women with PCOS translates into significantly increased risks for the development of type 2 diabetes mellitus (T2D) and other indicators of susceptibility to cardiovascular disease.11
The diagnosis of PCOS rests on the presence or absence of the three cardinal features, namely polycystic ovarian (PCO) morphology on ultrasound scan, hyperandrogenic features (defined biochemically and/or clinically) and oligo-amenorrhoea (an intermenstrual interval greater than 42 days). Under the 2003 Rotterdam diagnostic criteria for PCOS,12,13 presence of any two of these features suffices to establish a diagnosis of PCOS. One of the main objectives behind the Rotterdam revised criteria was to accommodate the increasing evidence that the spectrum of clinical and biochemical features in women with polycystic ovaries was wider than that proscribed by the 1990 National Institutes of Health (NIH) criteria for diagnosis of PCOS.14 According to these earlier criteria, a diagnosis of PCOS required the presence of both hyperandrogenism (or hyperandrogenaemia) and menstrual disturbance, without reference to ovarian morphology.
One inevitable consequence of the Rotterdam criteria is the generation of four phenotypic subgroups, all of which are equivalent with respect to the diagnosis of PCOS. These subgroups can be designated as follows: (1) PHO: women with PCO morphology (P), hyperandrogenic features (H) and oligo-amenorrhoea (O); (2) PH: women with PCO morphology and hyperandrogenic features but normal menses; (3) PO: women with PCO morphology and oligo-amenorrhoea, who have normal androgen levels; and (4) HO: women with hyperandrogenic features and oligo-amenorrhoea who have normal ovarian morphology on ultrasound.
The Rotterdam criteria have been controversial.12,13 This debate relates, in part, to the inclusion of two of these phenotypic subgroups (PH and PO women, the latter particularly), which, under the 1990 NIH diagnostic criteria, would not have been recognized as having PCOS at all.14 The justification for their inclusion rests on evidence from studies of the familial segregation of PCOS,15,16 improved understanding of the aetiology of the condition, and the variability in individual clinical and biochemical phenotype over time.17 The counter-argument holds that their inclusion is premature, because of uncertainty about the clinical consequences (especially the long-term health risks) associated with membership of the PH and PO subgroups. It has been shown that PH women are less insulin resistant and hyperinsulinaemic than their PHO counterparts.18 There are also recent data suggesting that the PO subgroup have the most metabolically favourable profile of all the PCOS phenotypic subgroups.19–21 If the long-term prognosis of women in the different PCOS subgroups displays substantial heterogeneity, adoption of the 2003 Rotterdam criteria (if performed without respect to clinical heterogeneity) has implications for clinical practice, research and patient insurability.22
The present study aims to establish how the metabolic features of the PO subgroup compare to those of other PCOS subgroups (all of whom, by definition, display evidence of hyperandrogenism), and thereby, to evaluate the likely implications of potential clinical heterogeneity.
Subjects and methods
White European women diagnosed with PCOS (n = 309), who were not on any confounding medication, were recruited as described previously.23 Following presentation with menstrual disturbances (oligo- or amenorrhoea) and/or hyperandrogenism, women were diagnosed with PCOS (as per the Rotterdam criteria12,13) if they also had polycystic ovarian morphology (PCO) on ultrasound scan.24 Also in accordance with the Rotterdam criteria,13 none of these women had other endocrine or neoplastic causes of hyperandrogenaemia (including congenital adrenal hyperplasia, androgen-secreting tumours and Cushing's syndrome). None of these 309 women was taking medication that could have confounded the clinical and endocrine presentation, and therefore complicated the assignment of phenotypic subgroups. (A total of 213 women from a total of 522 women with PCOS were excluded on this basis). Under the definitions outlined above, this dataset included 191 PHO, 76 PH and 42 PO women. As all women had polycystic ovarian morphology, the HO subgroup was, by design, not represented. We also included a control group comprising 76 white European women without PCOS. These women were either healthy volunteers or recruits from an infertility clinic (in whom the cause of infertility had been established to be due to male factors and/or tubal disease). All control women had ultrasound-proven normal ovarian morphology and regular menstrual cycles and none had hirsutism. Therefore, none of the control women had PCOS according to the 2003 Rotterdam criteria.12,13 None of the women with PCOS or controls had T2D.
Clinical details of all subjects are shown in Table 1. All clinical investigations were conducted in accordance with the guidelines in the Declaration of Helsinki, and the study was approved by the relevant Ethics Committee in the UK. All subjects gave fully informed consent.
Table 1. Clinical characteristics, endocrine and metabolic variables across the phenotypic subgroups of women with polycystic ovary syndrome (PCOS) compared with control women
|BMI (kg/m2)||29·2 (22·3, 38·3)||23·7 (20·1, 28·0)||23·9 (20·0, 28·6)||< 0·0001|| –||24·0 (19·9, 28·9)||0·99||–|
|Age§||32·9 (6·3)||34·8 (5·8)||32·1 (5·5)||0·04||–||36·9 (5·0)||< 0·0001||–|
|Waist circumference (m)||0·89 (0·73, 1·09)||0·76 (0·68, 0·84)||0·77 (0·67, 0·87)||< 0·0001||–||0·78 (0·68, 0·89)||0·55||–|
(2005 IDF criteria)
|56 (29·3%)||5 (6·6%)||3 (7·1%)||< 0·0001**||–||3 (3·9%)||0·45††||–|
|LH (U/l)||7·3 (3·7, 14·4)||6·0 (3·8, 9·5)||5·9 (3·6, 9·7)||0·03||0·06||4·1 (2·5, 6·8)||0·001||0·008|
|FSH (U/l)||5·3 (3·6, 7·9)||6·4 (4·6, 8·9)||6·0 (4·0, 8·9)||0·001||0·03||5·9 (3·9, 8·9)||0·86||0·83|
|Testosterone (nmol/l)||2·46 (1·73, 3·47)||1·94 (1·35, 2·81)||1·59 (1·11, 2·32)||< 0·0001||< 0·0001||1·42 (0·94, 2·18)||0·16||0·21|
|Androstenedione (nmol/l)||7·80 (5·27, 11·51)||5·79 (3·78, 8·88)||5·30 (3·50, 8·08)||< 0·0001||< 0·0001||4·75 (3·09, 7·28)||0·19||0·48|
|SHBG (nmol/l)||33·0 (18·8, 58·1)||50·8 (31·4, 82·2)||51·9 (34·3, 78·6)||< 0·0001||0·08||60·6 (40·0, 91·7)||0·07||0·02|
|HOMA2-IR||0·61 (0·16, 2·35)||0·31 (0·08, 1·19)||0·24 (0·06, 0·95)||< 0·0001||0·003||0·31 (0·11, 0·82)||0·55||0·38|
|Fasting insulin (pmol/l)||28·1 (7·1, 111·5)||15·0 (3·9, 57·6)||11·4 (2·9, 44·5)||< 0·0001||0·004||9·9 (2·9, 34·2)||0·70||0·77|
|HDL cholesterol (mmol/l)||1·2 (0·9, 1·6)||1·3 (1·0, 1·7)||1·3 (1·0, 1·7)||0·005||0·66||1·4 (1·1, 1·7)||0·66||0·82|
|Triglycerides (mmol/l)||1·2 (0·7, 2·0)||0·8 (0·6, 1·3)||0·8 (0·5, 1·3)||< 0·0001|| 0·18||0·8 (0·5, 1·2)||0·74||0·93|
Serum testosterone and SHBG concentrations were measured as previously described.23 Biochemical hyperandrogenism was defined as a serum testosterone concentration 2·84 nmol/l, androstenedione 9·26 nmol/l or free androgen index 6·94. Free androgen index was calculated as the total testosterone (nmol/l) divided by SHBG (nmol/l) and multiplied by 100. These cut-off values were derived, as previously described,25 from the distribution (mean + 2SD) for the corresponding variables measured (with the same assays) in the 76 control women. Clinical hyperandrogenism was defined as the presence of acne, hirsutism (Ferriman-Gallwey score 8) or alopecia. Other endocrine variables including LH and FSH were measured as previously described.26 Whenever possible, blood samples were taken from women during the follicular phase of the menstrual cycle (Days 2–5), but this was not possible in many of the women with PCOS who had absent or irregular menses.
The metabolic variables assessed included serum fasting insulin, HDL cholesterol and triglyceride concentrations. These were measured as previously described.26 HOMA2-IR values were calculated using the Oxford Diabetes Trials Unit calculator (http://www.dtu.ox.ac.uk; University of Oxford, UK). Women were also assessed for the presence of metabolic syndrome, as defined by the recent 2005 international diagnostic consensus (International Diabetes Federation, IDF27). In women, an IDF-proscribed diagnosis of metabolic syndrome requires the presence of central obesity (waist circumference 80 cm in European women), in addition to at least two of the following criteria: (1) elevated triglycerides ( 1·7 mmol/l); (2) reduced HDL cholesterol (< 1·29 mmol/l in women) or specific treatment for lipid abnormalities; (3) elevated blood pressure (systolic BP 130 mmHg or diastolic BP 85 mmHg) or specific treatment of previously diagnosed hypertension; and (4) impaired fasting plasma glucose 5·6 mmol/l or previously diagnosed T2D.
All statistical analyses were conducted in SPSS (version 12·0 for Windows; SPSS Inc., Chicago, IL, USA). All variables were skewed and underwent logarithmic transformation prior to statistical analysis, by one-way anova. Due to the known influences of body mass index (BMI) and age on the outcome variables (including fasting insulin, HDL cholesterol, testosterone and LH), analyses were optionally adjusted for BMI and age. A P-value < 0·05 was considered significant.
Clinical and biochemical data for the three main phenotypic subgroups of women with PCOS (PHO, PH and PO) and the control group are shown in Table 1. One-way anova of the data from the three PCOS phenotypic subgroups reveals significant between-group heterogeneity for all the endocrine and metabolic traits assessed, including HOMA2-IR (P < 0·0001), fasting insulin (P < 0·0001), HDL cholesterol (P = 0·005) and triglyceride (P < 0·0001) concentrations. The PHO subgroup clearly emerges as the most insulin-resistant (HOMA2-IR geometric means [SD ranges]: 0·61 [0·16, 2·35], 0·31 [0·08, 1·19] and 0·24 [0·06, 0·95] for the PHO, PH and PO subgroups, respectively; PHO vs. PO, P < 0·0001; PHO vs. PH, P = 0·002). In support of these findings, the PHO subgroup also has the most abnormal lipid profiles (Table 1).
The fact that the PHO subgroup also has the greatest BMI (Table 1) raises the question as to whether these differences are purely related to adiposity. Whilst adjustment for BMI (and age) generally reduces the significance of the between-group comparisons, for most of the metabolic variables (with the exception of SHBG and the lipid measures), statistically significant between-group heterogeneity remains (for example, fasting insulin, P = 0·004; HOMA2-IR, P = 0·003; Table 1). As with the unadjusted analyses, the PHO subgroup is the most abnormal (HOMA2-IR: PHO vs. PO, P = 0·003; PHO vs. PH, P = 0·021). Interpretation of between-group differences in androgens needs to take into account the fact that the subgroups were defined, at least in part, on the basis of androgen levels.
Across the range of clinical and biochemical data, the PO subgroup were the most similar to the controls. However, serum LH concentrations were significantly higher in the PO subgroup than in the controls (P = 0·001; Table 1). Following adjustment for BMI and age, significant differences in LH values between the PO and control groups persisted (P = 0·008; Table 1) and differences in SHBG appeared (P = 0·02; Table 1). All other PO vs. control group comparisons remained nonsignificant following adjustment (Table 1). Equivalent comparisons involving the PH subgroup (PH vs. controls) revealed that LH (P < 0·0001) and, as expected, both testosterone and androstenedione (P < 0·0001 and P = 0·05, respectively) were higher and SHBG was lower (P = 0·04) than in the control group. Endocrine and metabolic indices were similar in the PO and PH subgroups with the exception of higher serum testosterone in the PH subgroup (P = 0·007), again an expected result of the subgroup definitions. Thus, whilst endocrine abnormalities were found in both PO and PH subgroups, there was no evidence of metabolic disturbance in either of these two groups.
Prevalence rates of metabolic syndrome (as defined by the 2005 IDF consensus27) in the different subgroups were tracked with the measures of insulin sensitivity (HOMA2-IR, fasting insulin, triglycerides) described above. Notably, 29·3% of women in the PHO subgroup met these criteria (Table 1), as compared to 6·6% of PH women, 7·1% of PO women and 3·9% of controls (χ2 comparisons of the three PCOS subgroups, P < 0·0001; PO vs. controls, P = 0·45).
We have shown that oligomenorrhoeic women with PCOS who are normoandrogenaemic have a metabolic profile indistinguishable from control women without PCOS, and that evidence of severe metabolic and endocrine dysfunction in PCOS is restricted to the PHO subgroup. These findings persist even after allowing for marked differences in BMI between some of the subgroups. These data show that significant clinical heterogeneity exists between PCOS subgroups, and that this extends beyond the parameters used for their definition. Interestingly, however, SHBG levels were significantly lower in the PO (and PHO) women than in the controls. Low SHBG has been proposed as a surrogate marker of metabolic syndrome in women with PCOS.28 The persistence of significant between-group differences in fasting insulin and indices of insulin sensitivity, following adjustment for differences in BMI and age, points towards a direct causal relationship between insulin sensitivity (or insulin concentrations) and androgen levels, though the direction of causality cannot be identified. Similar relationships between hyperinsulinaemia and hyperandrogenaemia have been reported previously in a number of settings.29–31
Our findings confirm and extend those of recent studies in this area. In a study of 406 French women with PCOS, Dewailly et al. showed that PCOS women with features similar to our PO subgroup had lower fasting insulin concentrations than other PCOS subgroups, but this study did not include any measurement of insulin sensitivity.20 Broekmans et al. restricted their analysis to obese anovulatory women, finding that obese normoandrogenaemic women with PCOS also had a milder metabolic phenotype than obese women with PCOS diagnosed according to the 1990 NIH criteria (effectively a comparison of PO vs. PHO + HO in our terminology).19 More recently, Welt et al. studied 418 women with PCOS recruited in Iceland and Boston, USA. They showed that BMI and fasting insulin concentrations were highest in the PCOS women with features similar to our PHO subgroup, compared with women in the other phenotypic subgroups.21
In combination, the results of these studies are fairly clear. There is marked clinical heterogeneity within PCOS subgroups with respect to metabolic risk profiles such that women who have polycystic ovarian morphology but lack the full-blown syndrome (i.e. PO and PH subgroups) have much less extreme departures from normality.
One limitation of our study is that we were obliged to focus on those women with PCOS not on confounding medication, as to do otherwise would have made it difficult to accurately define phenotypic subgroups. A second limitation of all studies to date is that they are cross-sectional. Given the variability of the PCOS phenotype within individuals over time, prospective studies will be required before the implications of these findings can be extrapolated into predictions of lifelong morbidity and mortality. Nevertheless, these findings (and particularly their reproducibility across the four studies described), do suggest that stratification of metabolic and cardiovascular risk within PCOS, should take account of the clinical subphenotype rather than be based solely on a diagnosis of PCOS. Such stratification becomes increasingly relevant as enthusiasm grows for therapeutic intervention in subgroups considered at high risk of metabolic or cardiovascular disease (those with ‘prediabetes’ or ‘metabolic syndrome’, for example), especially as some of those interventions may be incompatible with the desire for future fertility.3,32,33 These data therefore support a more flexible approach to the management of the metabolic features of PCOS, which is tailored to the individual patient and influenced by the phenotypic subgroup of PCOS ascertained at each assessment.
Given these findings of clinical heterogeneity, should PO women even be considered to have PCOS at all? Our endocrine data help to address this question. As a group, the PO women have LH levels which are significantly greater than controls although somewhat lower than those seen in women with the full-blown syndrome (PHO subgroup). Dewailly et al.20 also found higher than normal LH levels in the equivalent subgroup. Although not selected on this basis, the PO women also have modestly (nonsignificantly) higher androgen levels and significantly lower SHBG concentrations than the controls. This establishes that PO women are indeed part of the PCOS spectrum, and is consistent with the fact that they share the same abnormalities of ovarian morphology.