Glucose Intolerance in a Large Cohort of Mediterranean Women With Polycystic Ovary Syndrome: Phenotype and Associated Factors

This study included 121 women with PCOS, all in the reproductive age (range 14–37 years), who consecutively attended the Endocrine Unit of S. Orsola-Malpighi Hospital, Bologna, Italy, in the previous 3 years. The diagnosis of PCOS was made by the presence of chronic anovulation, hyperandrogenemia (defined by supranormal total testosterone levels according to our reference value of 1.35 ± 0.62 nmol/l, and polycystic ovarian morphology at ultrasound examination, according to previously defined criteria (17) Other causes of hyperandrogenism, such as hyperprolactinemia, Cushing syndrome, and congenital adrenal hyperplasia, were excluded by specific laboratory analysis, as previously described (18). None of the subjects included in the study had thyroid, cardiovascular, renal, or liver dysfunctions based on clinical examination and routine laboratory findings; had taken any medication known to affect glucose or sex hormone metabolism during the 3 months prior to the study; or were dieting. The study protocol was approved by the local ethics committee, and all subjects enrolled in the study signed an informed consent.

On the first day of the study, anthropometric measurements, clinical and family history, and information on dietary habits and habitual physical activity were obtained from each subject. Baseline blood samples were also obtained for androgen and SHBG determinations. An oral glucose tolerance test (OGTT) (75 g, Curvosio; Sclavo, Cinisello Balsamo, Italy) was performed, with blood samples taken at baseline and 30, 60, 90, 120, and 180 min after the glucose load for glucose measurement and at baseline and 60, 120, and 180 min for insulin and C-peptide determinations. Normal glucose tolerance (NGT), IGT, and type 2 diabetes were defined using glucose levels during the OGTT, according to the criteria proposed by the World Health Organization (WHO) (19).

On the second day, starting at 8:00–8:30 a.m., a standard 1–24 ACTH test (250 μg cosyntropin [Synacthen] i.v.) was performed; samples for cortisol, dehydroepiandrosterone (DHEA), androstenedione, and 17-hydroxyprogesterone (17-OHP) determinations were drawn at baseline and 60 min thereafter. All samples were immediately chilled on ice and centrifuged; serum or plasma aliquots were frozen at −80°C until assayed. All PCOS women were investigated after a 12-h overnight fast, provided they had followed a 3-day diet containing >250–300 g carbohydrate/day, had mild oligomenorrhea within the first 10 days after the last menstruation, or had severe oligomenorrhea or amenorrhea regardless of the menstrual cycle.

Height, weight, waist and hip circumferences, and the waist-to-hip ratio (WHR) were measured according to standardized procedures (20). BMI was calculated as weight in kilograms divided by square meters of height in centimeters and was used to classified the study population as normal weight (BMI ≥18.5 and <25 kg/m²), overweight (BMI ≥25 and <30 kg/m²), or obese (BMI ≥30 kg/m²), according to WHO classification (21). Waist circumference ≥80 cm and WHR >0.85 were also used as indexes of abdominal body fat distribution (21). Hirsutism was estimated using the Ferriman-Gallwey score (22). Menstrual status was evaluated in the 6 months before the study. Oligomenorrhea was defined as cycle lengths >35 days and amenorrhea as no cycles occurring during the previous 6 months. An anovulatory state was defined as progesterone levels ≤2 ng/ml (6.4 nmol/l), measured every 10 days throughout the last cycle (23). The presence of acne and acanthosis nigricans was evaluated by a careful clinical examination (24,25). A questionnaire regarding personal and family history was given to each subject enrolled in the study and was filled in at home under the supervision of the subjects’ parents and, finally, of the investigator. The questionnaire addressed body weight at birth, age of menarche, and familiarity of diabetes and obesity. A subject was considered to have a positive family history for type 2 diabetes and obesity if at least one first-degree relative (parent or sibling) was or had been affected. To assess the reliability of the data related to birth weight, we compared the values reported in the questionnaire with those registered in the personal pediatric record, if available at home. This comparison was done in 88 patients. No significant difference was observed between the two sources of birth weight evaluations (P < 0.001; range of difference −0.1 to 0.1 kg) and they correlated well (r = 0.998, P < 0.001), documenting a good accordance between the values described in the questionnaire and those reported in the pediatric record. Dietary intake was evaluated by the dietitian attending our unit according to a previously defined method, including quantitative information on daily energy intake and macronutrient composition of the diet. Finally, physical activity was investigated by using a self-administered questionnaire proposed by Baecke et al. (26) that made it possible to distinguish three meaningful factors (scored from 1 to 5): physical activity at work (work index), sport during leisure time (sport index), and physical activity during leisure time, excluding sport (leisure index).

Plasma glucose levels were determined by the glucose oxidase technique immediately after blood was drawn. Levels for insulin, total testosterone, DHEA and its sulfate (DHEA-S), androstenedione, 17-OHP, SHBG, and cortisol were determined as previously described (18,27,28). The intra-assay coefficients of variation in our laboratory were 3.0% for insulin, 7.0% for total testosterone, 9.0% for DHEA, 5.9% for DHEA-S, 6.0% for androstenedione, 13.0% for 17-OHP, 6.5% for SHBG, and <8.0% for cortisol. Free testosterone was calculated as the ratio between total testosterone and SHBG, according to the method of Vermeulen et al. (29). To investigate insulin sensitivity in the basal condition and after the glucose challenge, the Quantitative Insulin-Sensitivity Check Index (QUICKI) and the homeostasis model assessment (HOMA) applied to the OGTT (HOMA_OGTT) were calculated, as previously proposed (30,31).

All data are expressed as means ± SD and frequencies (percent). Glucose, insulin, and C-peptide response to the OGTT are expressed as the area under the curve (AUC), calculated by the trapezoidal method, whereas hormonal response to the ACTH test is expressed as Δ values, calculated as the difference (Δ) between peak values and basal values of each hormone measured.

Continuous data were compared among the three groups by ANCOVA, using age as the covariate. Category data were analyzed using χ² testing. ANOVA and linear regression were used to compare birth records with self-reported birth weights. Two-tailed P values <0.05 were considered statistically significant. Statistical analysis was performed by running the SPSS/PC+ (SPSS, Chicago, IL) software package (32).

According to the glucose tolerance status, 2.5% (3 of 121) of the PCOS women had type 2 diabetes, 15.7% (19 of 121) had IGT, and 81.8% (99 of 121) had NGT. All women with IGT and type 2 diabetes (the GI group) were overweight or obese, whereas among NGT women, 80.8% (80 of 99) were overweight or obese (the OB-NGT group) and 19.2% (19 of 99) were normal weight (the NW-NGT group). The three women with type 2 diabetes were included in the same group as those with IGT (the GI group), as they had fasting glucose values <126 mg/dl (114, 110, and 114 mg/dl) and glucose levels at 120 min of the OGTT (243, 209, and 204 mg/dl) slightly exceeding the WHO threshold of 200 mg/dl (19). In addition, the overlapping age and similar abdominal obesity phenotype were also reasons to pool the type 2 diabetes and IGT subjects.

Subject characteristics are given in Table 1. PCOS women included in the GI group were significantly older compared with the other two groups (P < 0.05); therefore, statistical analyses among the groups were performed after adjusting for age. No significant difference in anthropometric parameters was found between the GI and OB-NGT groups; as expected, however, both groups had significantly higher values of body weight (P < 0.001), WHR (P < 0.001), and waist circumference (P < 0.001) with respect to the NW-NGT group. The hirsutism score was similar in all the groups. No difference was present in the frequency of oligomenorrhea between PCOS women with GI and those of the OB-NGT group, but in both groups oligomenorrhea was significantly more frequent than in the NW-NGT group (P < 0.05). Conversely, the frequency of acne was negligible (5%) in the GI group, whereas it was present in 25% of the OB-NGT group (P < 0.05 vs. GI) and 58% of the NW-NGT group (P < 0.001 vs. GI and P < 0.01 vs. OB-NGT). Finally, about half of the women in the GI and the OB-NGT groups had acanthosis nigricans, whereas none of the NW-NGT women were affected (P < 0.001).

PCOS women of the GI group had significantly lower birth weight than the other two groups, regardless of body weight (P < 0.01 vs. OB-NGT group and P < 0.05 vs. NW-NGT group). Birth weight of the three women with type 2 diabetes was 1.8, 2.5, and 2.8 kg. Moreover, the GI group had a significantly earlier menarche age (P < 0.05). A family history of diabetes was very high in all groups, with no significant difference among them, whereas the occurrence of obesity was not significantly different between the GI (60%) and OB-NGT (70%) groups but was significantly lower in the NW-NGT group (12%) (P < 0.001). Dietary intake and habitual physical activity did not significantly differ among the groups.

Table 2 shows subjects’ values related to glucose tolerance and insulin resistance. The GI group had significantly higher fasting glucose and glucose_AUC values with respect to the other two groups (P < 0.001). In addition, compared with the two NGT groups, the GI group had significantly higher fasting insulin (P < 0.05 vs. OB-NGT group and P < 0.001 vs. NW-NGT group), insulin_AUC (P < 0.001), and C-peptide_AUC (P < 0.01 vs. OB-NGT group and P < 0.001 vs. NW-NGT group) values. However, fasting C-peptide values were similar between the GI and the OB-NGT groups and significantly higher than in the NW-NGT group (P < 0.01). Finally, compared with the other two groups, the GI women had significantly lower QUICKI (P < 0.01 vs. OB-NGT group and P < 0.001 vs. NW-NGT group) and HOMA_OGTT (P < 0.001) values. However, among PCOS women with NGT, those included in the OB-NGT group had significantly higher fasting insulin (P < 0.01), C-peptide (P < 0.01), and C-peptide_AUC (P < 0.05) values and significantly lower QUICKI (P < 0.001) and HOMA_OGTT (P < 0.001) indexes.

Subjects’ steroids and SHBG levels are shown in Table 3. Baseline DHEA, DHEA-S, 17-OHP, androstenedione, and total testosterone did not differ among the groups. Compared with the NGT groups, the GI group had significantly higher free testosterone (P < 0.05 vs. OB-NGT group and P < 0.01 vs. NW-NGT group) and lower SHBG concentrations (P < 0.05 vs. OB-NGT group and P < 0.01 vs. NW-NGT group). These two parameters were also higher in the OB-NGT than in the NW-NGT group (P < 0.05). Fasting cortisol was significantly lower in the GI (P < 0.05) and OB-NGT (P < 0.05) groups than in the NW-NGT group.

The responses of DHEA and 17-OHP (Fig. 1) were similar among the groups; on the contrary, the response of cortisol (P < 0.001) and androstenedione (P < 0.05) was progressively and significantly higher in the OB-NGT versus the NW-NGT women and in the GI versus the other two groups (cortisol: P < 0.05 vs. OB-NGT group and P < 0.001 vs. NW-NGT group; androstenedione: P < 0.01 vs. OB-NGT group and P < 0.001 vs. NW-NGT group).

In this study, the first of its kind performed in the Mediterranean area, we found that 2.5% of 121 PCOS women had type 2 diabetes and 15.7% had IGT. This prevalence rate was significantly higher than that described in the general population of similar age (33), but somewhat lower than that reported in previous studies performed in the U.S. (4,5) and Asia (6). Some differences among these studies in the selection criteria of the patients enrolled cannot be ignored. In particular, our PCOS cohort was younger than those investigated in the aforementioned studies (4–6), which is supported by the concept that glucose tolerance tends to worsen with increasing age (33). Moreover, body weight also differed; in particular, it was lower in our cohort than in the others. Finally, the impact of environmental factors, particularly dietary habits, should also be taken into account, as the study populations belonged to different ethnic groups. The young age of our cohort reinforced the concept that GI states in PCOS women tend to appear earlier in life than expected (4–6). Moreover, the finding that our PCOS women with GI were slightly but significantly older than those with NGT supports the concept that the conversion rate from normal to altered glucose tolerance may be accelerated in these women (34).

Although all PCOS women with GI investigated in this study were obese, it is noteworthy that nearly 80% of obese PCOS subjects had NGT. Surprisingly, there was no difference in the frequency of the abdominal phenotype among these groups. Therefore, although obesity appears to be an important prerequisite for the development of GI in PCOS (4–6), additional pathogenetic factors need to be taken into consideration. Notably, we found that PCOS women presenting with GI were significantly more insulin resistant and had higher insulin blood levels than those with NGT, regardless of body weight or fat distribution. These data confirm and extend the results of the only long-term follow-up study performed by our group in PCOS women, demonstrating that insulin sensitivity, measured by means of insulin and glucose values during an OGTT, tends to worsen markedly over time, particularly in those with obesity, and is associated with the appearance of GI in many of these women (27). Taken together, these findings strongly support the concept that insulin resistance, worsening in time, may play a major role in the development of GI in obese PCOS women (35). On the other hand, in the presence of insulin resistance, pancreatic β-cell insulin secretion increases in a compensatory fashion and type 2 diabetes develops when this compensation is no longer sufficient to maintain euglycemia (36). Under normal circumstances, the relation between β-cell function and insulin sensitivity is constant (36). Previous studies have shown that several obese PCOS women may be characterized by impaired early-phase insulin secretion (15), which may play a contributory role in the development of GI. In this study, we found that PCOS women with GI had higher fasting and glucose-stimulated insulin and C-peptide concentrations. Although we cannot rule out the possibility that subtle defects in insulin secretion may be present in these women, our data nevertheless emphasize that more severe hyperinsulinemia in obese PCOS women is associated with a worsened insulin resistance state, even in the presence of GI.

This study also attempted to identify early markers of the development of GI in PCOS women, which can be important in clinical practice to plan preventive and therapeutic strategies. Two new interesting factors emerged from our investigation. The first was represented by the presence of lower birth weight in the GI group with respect to the two NGT groups, confirming the close asso-ciation of this feature with insulin resistance and susceptibility to developing type 2 diabetes (37). The mechanisms responsible for this association are still speculative; it has been hypothesized that although undernourished fetuses make a metabolic adaptation from which they benefit in the short-term by increasing fuel availability, in time this condition becomes damaging, leading to insulin resistance (37). The second factor was the earlier menarche age found in PCOS women with GI in comparison with those with NGT. These data are in line with those observed in a different group of patients where a precocious pubarche was related to hyperinsulinemia and low birth weight (38). At variance with this finding, a positive family history for diabetes was similar in all groups, regardless of the glucose tolerance state. Therefore, these data do not seem to support the concept that a family history positive for diabetes may predict the development of IGT or type 2 diabetes. However, this observation could be attributable to the fact that all our GI subjects were included in the same group. Indeed, when considered separately, a positive family history for diabetes was present in all diabetic women and in approximately half of those presenting with IGT, making these results similar to those reported in other studies (5,6).

After having defined the factors associated with the development of GI in PCOS, we tried to characterize the phenotype of our cohort. We found that hyperandrogenemia was more severe in PCOS women with GI than in the other groups due to the significantly lower SHBG concentrations and, consequently, the increased free testosterone fraction. It is likely that hyperinsulinemia represents the major contributing factor, as hepatic SHBG synthesis is negatively regulated by insulin (39). In addition, PCOS women with GI were characterized by a higher response of cortisol and androstenedione to ACTH stimulation in comparison with the OB-NGT group and, particularly, the NW-NGT group, suggesting an adrenal hyperresponsiveness in obese PCOS subjects, particularly those with coexisting GI (40). Baseline cortisol levels in both GI and OB-NGT groups were, however, significantly lower than in these groups’ normal-weight counterparts. The lower fasting cortisol levels suggest increased cortisol clearance in obese PCOS women and, therefore, adrenal hyperresponsiveness to ACTH may be viewed as a compensatory mechanism used to maintain normal cortisol levels (41,42). Increased cortisol clearance may be secondary to an increased activity of the 5α-reductase (42) and/or of the 11 β-hydroxysteroid dehydrogenase (41), which convert cortisol to inactive α-tetrahydroderivatives and cortisone, respectively, in both the adipose tissue and the liver. As suggested by Rodin et al. (41), increased androgen production, due to compensatory adrenal hyperfunction, may be partly involved in the hyperandrogenic state in PCOS, particularly when obesity is present. Hyperinsulinemia may be partly responsible for this (43). Alternatively, insulin might exert a direct modulation of adrenal steroidogenesis, an idea supported by in vitro studies where insulin administration reinforced the responsiveness of the enzymes involved in the adrenal synthesis to ACTH stimulation (3), and also by in vivo studies performed in PCOS patients showing that the amelioration of insulin resistance and related hyperinsulinemia by insulin sensitizer administration was associated with a decrease of adrenal androgen blood levels (44).

Despite different blood androgen levels among the groups, the hirsutism score was similar, suggesting that hirsutism may depend on the excess androgen per se, rather than on the degree of hyperandrogenemia or on some differences in the peripheral biological action of androgens, which may vary among PCOS women. Surprisingly, however, the frequency of acne was very high in the normal-weight PCOS women but significantly lower in the OB-NGT group and negligible in the GI group, where acne was found in only one patient. This suggests that acne in PCOS women may not be related to obesity or to the severity of insulin resistance and may be relatively independent of circulating androgen levels, as suggested by other studies (45). Moreover, the important difference in the frequency of acne despite a low difference in age among the three groups, together with the evidence that the two NGT groups with a similar age had a significant difference in the frequency of acne, make it unlikely that age is a confounding factor in the evaluation of acne in our study.

Conversely, acanthosis nigricans was clearly related to the presence of a mild-to-moderate insulin resistant state, as otherwise expected (46). The fact that oligomenorrhea was significantly more frequent in obese PCOS women, regardless of the subjects’ glucose tolerance state, than in their normal weight counterparts further confirmed the well-known negative impact of obesity and associated more severe insulin resistance and hyperadrogenemia on menstrual cycles in PCOS.

Finally, the finding of a lack of significant difference in energy and macronutrient intake and in physical activity between PCOS subjects with GI versus those with NGT in this study was intriguing. Although the limitations of the low number of patients belonging to each group, particularly the GI and NW-NGT groups, and therefore the low analytical power (P < 0.42) of these parameters must be considered, our data suggest that lifestyle habits are probably not relevant for defining the categories of subjects susceptible to developing GI, at least in those with PCOS. On the other hand, our findings do not exclude the possibility that women with PCOS may have specific alterations of dietary intake, as several other studies seem to suggest (rev. in 47). This latter possibility has not been adequately investigated in well-done epidemiological studies.

In conclusion, this study showed, in a large cohort from the Mediterranean region, that GI states are common in relatively young PCOS women and are associated with the presence of obesity. Moreover, our results clearly indicate that PCOS women with GI are characterized by more severe insulin resistance, insulin hypersecretion, and hyperandrogenemia. Other than a positive history for diabetes, lower birth weight and early menarche appear to be important factors associated with the development of GI later in life. This may be relevant from the clinical point of view while planning both preventive and therapeutic strategies.

The study was supported by a grant, “Ricerca Fondamentale Orientata ex quota 60%,” from the University of Bologna.

We thank Susan West for reviewing the manuscript for English usage.