Polycystic ovary syndrome as a diabetes risk factor

Keywords::

• diabetes mellitus
• impaired glucose tolerance
• insulin resistance
• PCOS
• PCOS phenotype

Polycystic ovary syndrome (PCOS) is a very common disorder affecting at least 6–8% of women in the reproductive age. Its etiology is still unknown, likely being multifactorial and consisting of both genetic and environmental components. Most women with PCOS show clinical and biochemical hyperandrogenism, menstrual irregularities and micropolycystic morphology of the ovary. Interestingly, a large body of evidence suggests PCOS is also a metabolic disease, with potential increase of cardiovascular risk.

Among the metabolic alterations found in PCOS women, insulin resistance, with the associated compensatory hyperinsulinemia, is of particular importance in terms of both frequency and clinical implications [1]. Available data indicate that up to 50–70% of PCOS subjects are insulin resistant, and it was hypothesized that they may have specific abnormalities responsible for the impairment of insulin action. However, the mechanisms involved remain largely unclear. It is noteworthy that this phenomenon seems to play a crucial role in the pathogenesis of PCOS itself.

Insulin resistance is a well-established pathogenic factor for Type 2 diabetes mellitus (DM). Beta-cell dysfunction has been also reported in these women [1]. This defect is central to the development of hyperglycemia, because the compensatory mechanism of hyperinsulinemia fails. Based on these findings, it is not surprising that abnormalities of glucose metabolism are considered a common feature of PCOS and that several scientific societies have identified PCOS as a significant nonmodifiable risk factor for DM. Nevertheless, literature on this issue is still limited.

Legro et al. reported a prevalence rate of impaired glucose tolerance (IGT) and DM, in a multiethnic sample of 254 American PCOS women, of 31.3 and 7.5%, respectively, as compared to 14 and 0% in age- and weight-matched healthy controls [2]. More recently, a meta-analysis by Moran et al., including 13 controlled studies and a whole sample of 12,105 PCOS women and 56,959 controls, concluded that DM was 4.4-times higher in PCOS women than in healthy women [3]. Moreover, in a smaller sample, frequency of IGT was 2.5-times higher in PCOS women. Interestingly, these figures were substantially unchanged (odds ratios [ORs]: 4.00 and 2.54, respectively) when only studies comparing BMI-matched PCOS women and controls were considered. Nevertheless, it should be pointed out that these data derive from cross-sectional studies, most of them with a very limited sample size and potential referral bias, and that these results were heterogeneous. Moreover, not all studies took into account differences in age or used the same criteria to define glucose abnormalities. However, the results of some retrospective studies [4–8] and of a few small prospective studies supported these conclusions.

Interestingly, in a sample of 67 Australian PCOS women with a mean BMI of 28.7 kg/m², Norman et al. reported that, after a mean follow-up of 6.2 year, conversion rate from normal glucose tolerance to IGT or to DM was 1.5 and 1.2% per year, respectively; whereas in subjects with IGT at baseline, the conversion rate to DM was 8.7% per year [9]. Similarly, in a cohort of 83 Canadian PCOS women with a mean BMI of 33.8 kg/m², at the end of a 3 year follow-up, conversion rate from normal to IGT or DM was 6.8 and 2.0% per year, respectively [10].

Consistent results also come from some long-term follow-up studies, with mean duration between 8 and 18 year, comparing incidence of DM in PCOS women and either the general population or a non-PCOS cohort [11–14]. These studies reported that diabetes incidence was 2.1–4.7-times higher in PCOS patients. Gambineri et al. [13] showed that higher baseline BMI and plasma glucose levels, at fasting and after oral glucose, increased the likelihood of developing DM. Wang et al. [12] reported that diabetes risk, adjusted for several confounders, was over three-times higher in both normal weight and overweight/obese women with PCOS compared with normal weight women without PCOS. Interestingly, this risk was especially increased in women with persistent PCOS at the follow-up visit (adjusted OR: 7.2). These data match a retrospective study from the UK, with a mean follow-up of 4.7 year [7], reporting an increase of diabetes risk also in lean PCOS women. However, in this study, weight gain during the follow-up was associated with worsening of glucose tolerance, with a 1% increase in BMI leading to a 2% increase in diabetes risk. Overall, these findings support the hypothesis that PCOS status per se is a risk factor for DM and that obesity increases this risk.

Women with PCOS also showed increased risk for gestational diabetes. In this regard, a recent meta-analysis reported this risk is approximately three-times higher in PCOS women compared with the non-PCOS population [15]. Moreover, women with PCOS seem to have increased risk for persistence of abnormal glucose tolerance after pregnancy.

The higher prevalence of glucose intolerance in PCOS versus non-PCOS women was observed in subjects of various race/ethnicity, without significant differences between Caucasian and Afro-American subjects. However, a worse glucose tolerance was reported in Indian, Asian and Hispanic PCOS women compared with Caucasian women [16]. As in the general population, prevalence of abnormal glucose tolerance in PCOS women increases with age; however, in PCOS women this abnormality may develop at a younger age.

A very controversial issue is the possible association between glucose abnormalities and the specific clinical features of PCOS. In most studies, a worse metabolic profile was reported in hyperandrogenic versus nonhyperandrogenic women. Nevertheless, Moran et al. found that PCOS women with either the ‘classic’ or the ‘non-classic’ PCOS phenotypes (i.e., those with both hyperandrogenism and oligoanovulation or those lacking one of these two features) had a similar increase in the diabetes risk score [17]. Moreover, in a large, long-term retrospective study, evidence of hyperandrogenism was unexpectedly associated with a lower incidence of DM [8].

We have recently provided insight into metabolic differences between PCOS phenotypes, measuring insulin resistance by the gold standard hyperinsulinemic-euglycemic clamp in 137 women from the Verona PCOS Pathophysiology and Phenotype (Verona 3P) Study cohort. We demonstrated that, after adjustment for confounding factors, there is a scale of insulin resistance and associated metabolic abnormalities between the PCOS phenotypes, and that, from this point of view, the nonhyperandrogenic phenotype behaves as a separate population [18]. We also found that these women have an impaired ability to shift their substrate metabolism from predominant lipid oxidation at low (fasting) insulin levels toward predominant glucose oxidation at high (clamp) insulin levels [19]. We hypothesize that this condition of ‘metabolic inflexibility’, which was associated with both insulin resistance and hyperandrogenism, may favor glucose tolerance abnormalities. Consistently, preliminary data from this cohort indicate an increased risk for IGT among women with classic versus nonclassic PCOS phenotypes. Overall, these results also suggest that assessing PCOS phenotypes is useful in estimating the individual risk of metabolic abnormalities in these women.

At present, no prospective large-size studies including PCOS and control women are available. These studies are necessary in order to definitively identify the risk and protective factors for the development of DM in these women. Another crucial, still unanswered question is whether the different pharmacological treatments used in PCOS patients, such as insulin sensitizers, oral contraceptives or antiandrogens, may affect the risk of glucose abnormalities in these subjects. In this regard, recent retrospective data suggested that this risk could be higher in women given metformin than in those given oral contraceptive [13]. The hypothesis of a protective metabolic effect of estroprogestins is interesting. Nonetheless, the choice of treatments in these women was likely influenced by their individual characteristics. Therefore, while these intriguing findings underline the urgent need for controlled trials, they cannot influence the medical approach to these subjects.

Overall, available data support the recommendations of scientific societies that PCOS women should be screened for DM, at least in the presence of common risk factors such as overweight/obesity, metabolic syndrome, family history for DM or previous gestational diabetes. However, there is little agreement about extension of this screening. Though an ultimate answer to this question is not yet available, we think that screening should include women with the hyperandrogenic phenotypes of the syndrome, that is, all women diagnosed with PCOS according to the AE-PCOS Society criteria [20].

Lastly, it should be noted that many PCOS women with IGT or even DM may have normal fasting plasma glucose. Therefore, this measurement is of limited value in establishing which PCOS subjects should be screened for abnormalities in glucose tolerance. The American Diabetes Association recently introduced HbA1c cutoff values for the diagnosis of diabetes (≥6.5%) or a condition of increased risk for diabetes (5.7–6.4%). Consistently, prevalence of abnormal HbA1c was higher in PCOS women than in controls. Nevertheless, several studies showed little agreement between HbA1c and OGTT in identifying PCOS subjects with abnormal glucose tolerance. It would seem that sensitivity of HbA1c is especially low in young women, which means that this measurement is not suitable for screening glucose tolerance in PCOS subjects. For this purpose, OGTT remains the best choice.

Financial & competing interests disclosure

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

No writing assistance was utilized in the production of this manuscript.