Insulin resistance and obesity among infertile women with different polycystic ovary syndrome phenotypes

Polycystic ovary syndrome (PCOS) is a common problem among Arab women and is the main cause of infertility due to anovulation. This study investigates insulin resistance (IR) and obesity in different PCOS phenotypes among infertile women (n = 213), of whom 159 had PCOS and 54 women without PCOS, recruited as a control group. Biometric, hormonal and clinical parameters were studied. IR was observed in 133 (83.6%) women with PCOS and in 25 (46.3%) women without PCOS (p < 0.001). IR was significantly associated with PCOS only among women with central obesity (χ2 = 35.0, p < 0.001) and not for the normal category (χ2 = 4.04, p < 0.058). The LH/FSH ratio was not significantly different among the PCOS group (n = 37, 23.3%) compared to the control group (n = 9, 16.7%) (p = 0.308). Among women with PCOS, the most common phenotype was type I (50.3%), with type III (29.6%), type II (14.5%) and type IV (5.7%). Type I had the highest values of fasting insulin (median = 12.98 mU/mL) and HOMA IR values (significant difference among the four phenotypes, p = 0.009 and 0.006, respectively) and is associated with severity of the disease. There was no difference in glucose levels.

Prevalence of different PCOS phenotypes. The most common phenotype in our study was type I (full-blown PCOS) 50.3% (n = 80/159), followed by type III (women with hyperandrogenism, polycystic ovaries
Women with PCOS have higher median rank fasting insulin, HOMA-IR, free testosterone and total testosterone than controls (p-value < 0.001). However, fasting glucose, sex hormone binding globulin (SHBG) and FSH values were not statistically significant (Table 1). Free androgen index (FAI) was significantly higher in the PCOS group (5.35 vs. 3.93, p-value = <0.001) ( Table 1).
We explored BMI and waist circumference among all study participants and found that both were significantly higher among those with IR of >2.5 compared to those with IR of <2.5. It was particularly clear among the obese, (n = 63, 39.9%) and (n = 41, 25.9%) for BMI and waist circumference, respectively (p-value < 0.001). Obesity was associated with IR in all cases and controls (Table 3). Cochran-Mantel-Haenszel statistics was chosen to assess the significance of association between the disease status (cases vs. control) and IR (>2.5 vs. <2.5) while controlling for BMI or waist circumference status (normal vs. obese) (  associated with PCOS for normal BMI (χ 2 = 6.60, p = 0.022), whereas for the obese category, it was also significant (χ 2 = 27.12, p < 0.001). Mantel-Haenszel statistics were significant with χ 2 = 29.08 and p < 0.001. IR > 2.5 was significantly associated with the PCOS only among the obese category of waist circumference (χ 2 = 35.0, p < 0.001). For the normal category, it was not significant (χ 2 = 4.04, p < 0.058). Mantel-Haenszel statistics were significant with χ 2 = 26.48 and p < 0.001 (Table 4). These findings support the role of IR as a causative mechanism for the presence of PCOS in addition to the waist circumference as a predictor of the presence of PCOS. These results also confirm that it has a greater predictability than BMI. However, the additive effect of obesity in general could not be excluded, given significant variations in IR among cases compared to controls.

Prevalence of different PCOS phenotypes.
Hormonal, metabolic and ovarian abnormalities among different PCOS phenotypes. Among women with PCOS, we found that the median of the fasting insulin values was 12.98 mU/mL for all phenotypes, with highest values among those women with type I, and there was a statistically significant difference between phenotypes (p-value = 0.009). Fasting glucose did not differ between the groups. HOMA-IR, FSH, LH and FSH/LH ratio were statistically different between the groups. Phenotype IV had the highest LH and LH/FSH ratio, whereas type I had the highest HOMA-IR values. When we compared FSH levels, using post hoc Bonferroni comparisons between different phenotypes, there were statistically significant differences in FSH levels between phenotypes I and II and between I and III (p < 0.01) ( Table 5).
When examining the values of HOMA-IR in all PCOS patients, we found that HOMA-IR values differ significantly, with high BMI and WC values with p-value < 0.001 (Table 3).
As expected, women with type I had the highest prevalence of PCO and large ovarian volume, and there was a statistically significant difference between the three phenotypes (type II excluded because there should not be features of PCO on U/S scan) with p < 0.001.
Insulin resistance (IR) among different PCOS phenotypes. There was even distribution of HOMA-IR values between obese and non-obese women with PCOS, using independent Samples Kruskal-Wallis test of the distribution of HOMA-IR with p-value < 0.005 (Table 5). In addition, HOMA-IR did significantly differ among the  four phenotype groups, using independent Samples Kruskal-Wallis test of the distribution of HOMA-IR with p-value = 0.006 (Table 5).

Discussion
Our study shows that the type I classical and full-blown severe PCOS phenotype is the most common; this accords with other studies. In addition, the prevalence of all phenotypes is similar to the findings of other studies 19,20 . Our results show that fasting insulin, fasting glucose and free testosterone were statistically different between weight categories in PCOS patients. Women with the type I PCOS phenotype were also found to have more insulin resistance and higher values of free testosterone, LH, FSH and LH/FSH ratio. On the other hand, our Jordanian study is similar in prevalence of type I PCOS phenotype to an Italian study 21 , in contrast to Iranian 22 and Chinese 23 studies, in which the type IV PCOS phenotype was the most prevalent ( Table 6). As demonstrated in our data, obesity is more prevalent in women with the classic severe PCOS phenotype. Thus, women with PCOS Type I; are at increased risk of developing insulin resistance features, which is independent and may be worsened by central adiposity. In addition, obese women have higher free testosterone levels than women with normal weight. Our data also shows that waist circumference measurement is a better predictor of central obesity than BMI. Several studies have described endocrine and metabolic differences between lean and obese women with PCOS. In addition to alteration in insulin sensitivity that was independent of obesity, these studies have demonstrated more marked hyperandrogenemia, IR and relative hyperglycemia and lower sex hormone binding globulin (SHBG) in the obese compared with lean women with PCOS 24 .
Obesity is a common problem among Jordanian females 25 . Although polycystic ovary syndrome is believed to be one of the most common endocrine disorders in women worldwide, reports are very rare regarding the clinical and biochemical features of Arabic (including Jordanian) women with PCOS 26 . The hormonal profile of PCOS among Saudi women was similar to that published in the literature, with the exception of the prevalence of metabolic syndrome, which was less than in global reports 27 . As shown above, HOMA-IR was significantly correlated with BMI. In addition, our advanced statistical analysis showed that IR is highly associated with the presence of PCOS, although the additive synergistic effect of obesity could not be excluded. The results also show that obesity would worsen the severity of insulin resistance. The prevalence of insulin resistance was not significantly different between various PCOS phenotypes and type I, the most severe phenotype had the highest values of IR, in agreement with other reports 28 . Obese women with PCOS have more severe hyperandrogenism and its related clinical features (such as hirsutism, menstrual abnormalities and anovulation) than normal-weight women with PCOS. This picture tends to be more profound 29 particularly in obese women with PCOS and central obesity.
Many patients with PCOS demonstrate other metabolic abnormalities. Especially notable is the presence of insulin resistance (IR), accompanied by compensatory hyperinsulinemia [30][31][32][33] . Actually, insulin possesses true gonadotrophic function and an increase in insulin availability within ovarian tissue may enhance excess androgen synthesis. Obesity, particularly the abdominal type, may be partly responsible for insulin resistance and   associated hyperinsulinemia in women with PCOS. Therefore, obesity-related hyperinsulinemia may play a key role in favouring hyperandrogenism 34 . Moreover, obesity is associated with IR 35 . IR is also affected by ethnicity 36 and age 37 . When we used the HOMA-IR calculation to estimate insulin sensitivity, adjusted for confounders, we observed that 133 (83.6%) of 159 PCOS patients were insulin-resistant. This is consistent with the recent findings of other reports 38,39 . Furthermore, insulin resistance is present in both obese and non-obese women with PCOS 40 . Legro et al. 41 reported a higher prevalence of insulin resistance in obese (64%) than in non-obese (20%) women with PCOS. In our study, the prevalence of insulin resistance was significantly higher in overweight patients than in patients of normal weight. As shown above, there was no difference in FSH levels between cases and controls and, in fact, there was a relative deficit in FSH levels in women with PCOS. Some reports indicate that altered secretion of inhibin B may lead to a relative deficit of FSH in PCOS 42 . The overproduction of LH and, consequently, an incorrect LH/FSH ratio is presently not considered to be a characteristic attributed to all PCOS patients. In our study, we found that LH/FSH changes were attribute of 23.3% of cases with PCOS and in 16.7% of controls, which differs from the findings of other studies, in which higher incidence was reported 29 . Moreover, our results show that obesity was not a risk factor for an abnormal LH/FSH ratio, disproving the traditional concept of the disease that the heavier the patient is, the higher the LH/FSH ratio, and our results agree with a recent report 43 .
This study has a number of strengths, including the relatively good number of patients and the richness of data, which paves the way for more studies to address this problem. In addition, our study population is homogeneous in racial and ethnic variations, which gives more credibility to the results and conclusions. Furthermore, this study included subjects with infertility with or without PCOS, in whom, unfortunately, the main focus of doctors and patients was on infertility problems; our data will encourage doctors and patients to be aware of the high prevalence and role of obesity and insulin resistance first and, second, to correct the endocrine dysfunction in those women with PCOS by reducing their weight and improving their insulin levels. The use of mFG score for the assessment of hirsutism may be considered a limitation in the study because it is a subjective issue. The other probable limitation is the use of the HOMA-IR threshold value for diagnosis of IR resistance, which may have affected the true prevalence of IR among our study subjects. However, the gold standard for establishing IR is the euglycemic hyperinsulinemic clamp, which is not suitable for large-scale clinical use; therefore, we used the HOMA-IR calculation, which correlates with the euglycemic hyperinsulinemic clamp and is often used in other reports 22 as a marker for IR.
The full-blown phenotype of PCOS was the most common phenotype in this cohort of Jordanian women with PCOS. The prevalence of hyperandrogenism (including hirsutism, acne and biochemical hyperandrogenemia), obesity and insulin resistance were higher in this cohort than in women from other ethnicities. This fact will encourage doctors to focus on resolving obesity and IR problems in addition to infertility problems. A focus on central adiposity evaluation by waist circumference measurement instead of BMI is more valid. LH/FSH ratio should not be part of PCOS assessment. Owing to Arabic women, including Jordanians, being at high risk developing abnormal glucose metabolism and type II diabetes, further research is required to investigate the value of the recommendation for initial and periodic screening for hyperglycemia and IR in women with PCOS, including annual evaluation for diabetes.
Incorporating information from our study, we can now characterise the phenotype of Arabic women with PCOS as being similar to other women with PCOS with regard to insulin resistance and obesity, but the key differences are that our women have more of types I and III and less of type IV phenotypes.

Methodology
Subjects and setting. This cross sectional observational study was carried out in Karak City in the southern part of Jordan from January 2012 to April 2015. A total of 219 consecutive, untreated women presenting with initial diagnosis and treatment of infertility due to different causes, including PCOS, were recruited. The diagnosis of PCOS was determined according to the Rotterdam's criteria 15 . Patients' median age was 24 (IQR: [22][23][24][25][26][27][28][29] years. Due to infertility problems, none had used oral contraceptives for at least 3 months preceding the study. Women with persistent elevations of prolactin (PRL) (>24 μg/L) or abnormal thyrotropin (TSH) values (>5.5 mIU/L or <0.35 mIU/L) were excluded. Six cases were excluded, leaving a total final number of 213 study subjects. Of them, 159 had infertility problems due to PCOS; 54 women had infertility problems due to other reasons, including tubal factor, endometriosis and pelvic inflammatory disease, and these women served as a control group. The causes of infertility were determined by clinical evaluation, laboratory investigations, hysterosalpingogram and hysteroscopy or laparoscopy.
This study was based in two gynaecological clinics, one at the Ministry of Health in Al-Karak Hospital, affiliated with the Mutah University Obstetrics & Gynaecology Department in the south of Jordan, and the other at the main author's (MA) private practice. Institutional review board approval was obtained from Mutah University Ethics Committee (N0-201217), and informed written consent was obtained from all study participants. We confirm that all methods were performed in accordance with the relevant guidelines and regulations of the ethics in Human research and all research data can be available.

Protocol.
A standardised form was used to take medical history and physical examination, with emphasis on menstrual dating and regularity, the presence of hirsutism, acne and acanthosis nigricans. Full records of medications and gynaecological and family histories were obtained. Menstrual cycles were identified as (1) regular menses: a cycle with an inter-menstrual interval of 21 to 35 days; (2) oligomenorrhea: an inter-menstrual interval of 36 days or longer; and (3) amenorrhea: an inter-menstrual interval of 6 months or longer in previously regularly menstruating women and more than 3 months in women with oligomenorrhea.
Women diagnosed with infertility were grouped into two groups: those with PCOS and those without PCOS. The diagnosis of PCOS was based on clinical, laboratory and ultrasound scan data (see ovarian assessment). All women underwent physical examination. Obesity was assessed by estimating body mass index (BMI: weight/ height 2 in kg/m 2 ) with normal range between 17 and 25, overweight between 25 and 30 and obese >30. Central obesity was assessed by waist circumference (WC) measurements according to the WHO protocol 44 , measured at the mid-point between the highest point of the iliac crest and the last palpable rib, using a flexible, inelastic measuring tape with a tension meter attached. Then all measures were classified into three categories: normal (WC 79.9 cm or less), overweight (WC 80.0 to 87.9 cm) and obese (WC 88.0 cm or more), according to age-and sex-specific cut-offs 45 .
On physical examination, the presence of terminal hair growth was scored using a modification of the Ferriman-Gallwey (mFG) method with a score of 8 or greater (mFG ≥ 8) was considered hirsutism 46,47 . The mFG scoring method represents the sum of the hair growth scores of nine body areas (upper lip, chin, chest, upper abdomen, lower abdomen, upper arms, thighs, upper back, and lower back), with the examiner assessing terminal hair growth in each area and rating it from 0 (no hair) to 4 (extensive hair growth) 48 . Biochemical hyperandrogenism was identified as a total T level >83.6 ng/dL (SI Unit: 2.96 nmol/L), free T level >0.81 ng/dL (0.026 nmol/L), free androgen index FAI = [total testosterone/sex hormone binding globulin (SHBG) × 100] > 5 49 or a DHEAS level >1,580 ng/mL (6.64 mol/L) 50 . We defined clinical hyperandrogenism (HA) as the presence of hirsutism and/ or acne; because there is no agreement regarding the best way to assess acne 51 , we elected to record its presence without scores in our study subjects. Hormonal assessment. Blood samples were collected from all participants after having been instructed not to eat, drink or smoke. All patients reported no consumption of alcohol ever. A 10-cc sample of blood was drawn during days 2-6 of the menstrual cycles (natural or bleeding after progestin withdrawal) in plain-top tubes for subsequent hormonal analysis. If amenorrhoea was present, this was performed after excluding a dominant follicle with ultrasound. Blood was analysed for free and total testosterone (FT and TT) (FT; nmol/L), sex hormone-binding globulin (SHBG; normal values 40-120 nmol/L), LH (units per L), FSH (units per L) and fasting blood glucose (mg/dL); fasting insulin (micro-units per mL) was detected by the glucose oxidase method (AU640 automation biochemistry analyser and its relevant reagent, Olympus Company, Hamburg, Germany), sex hormone binding globulin (SHBG), luteinising hormone (LH) and follicle-stimulating hormone (FSH). LH/FSH ratio was calculated for each subject; LH/FSH ratio >2 was considered abnormal. The LH and FSH were measured by electrochemiluminescence technique in a COBAS AutoAnalyzer using reagents manufactured by Roche Diagnostics; quality control was done by PreciControl 1 and 2, and then the mean LH/FSH ratio was obtained. Free androgen index (FAI) was calculated by [total testosterone] X 100 divided by [SHBG]. Normal insulin sensitivity was defined by fasting insulin levels <12 mU/mL. We also calculated indices of the homeostasis model assessment (HOMA-IR) 54 . HOMA−IR was calculated using the equation HOMA−IR = Fasting insulin (μU/mL) × fasting glucose (mg/dL)/405 55 with HOMA-IR > 2.5 mol μU/mL as cut-off of abnormal value 56 .
In cases of fasting glucose consistent with impaired glucose tolerance (fasting glucose [FG] 110-125 mg/dL), a 75-g oral glucose tolerance test was performed; subjects with overt diabetes mellitus and impaired glucose tolerance were referred to a diabetes specialist. In addition, serum PRL (nomograms per mL), TSH (micro-units per mL) and 17α-hydroxyprogesterone (17-αOHP; nomograms per mL) levels were examined in blood samples of women with oligomenorrhea. Normal values of 20-100 ng/dL (all blood samples were taken before ovulation on day 2 of cycle) were accepted to exclude other causes of menstrual disorders 57 . In addition, blood samples of dehydroepiandrosterone sulfate (DHEAS) ug/dL and 17-hydroxyprogesterone (OHP) were also obtained with, normal values = 380 ug/dL. If values of DHEAS were <380 ug/dL, patients with apparent adrenal problems causing hyperandrogenism were excluded.
Ovarian assessment. The revised Rotterdam criteria were used for the diagnosis of PCO on ultrasound 58 , i.e., either an ovary with ≥12 follicles measuring 2-9 mm in diameter or an ovary with increased volume (>10 cm 3 without concomitant cysts). Ultrasound examination was performed by one investigator (MA) during the first 5 days of the menstrual cycle (the early follicular phase). The ultrasound assessments were undertaken trans-vaginally using a computed sonography system with a 7.0-MHz transducer. After identification of the ovaries, the size of the ovary was measured in three orthogonal planes. Ovarian volume was calculated using the formula for a prolate ellipsoid. The total number of follicles in each ovary was counted. Measurements of the largest and smallest follicles were taken in their maximum diameters and recorded in mm. Follicles were counted on the frozen images of two non-overlapping planes in the longitudinal section of each ovary. Statistical analysis. Statistical analysis was performed using the Statistical Package for the Social Sciences (SPSS), version 20.0, for Windows (SPSS, Inc., Chicago). Data are presented as mean ± SD. When variables were significantly skewed, median and interquartile ranges were used. Categorical variables were analysed using Chi-square tests, normally distributed continuous variables using analysis of variance (ANOVA) and significantly skewed continuous variables were analysed using the Kruskal-Wallis test. Student's t-test was used for the comparison of continuous variables. Group means were compared using analysis of variance (ANOVA) with post hoc least squares means pair-wise comparisons (after log transformation of the values). Statistical significance was considered when two-tailed P was ≤ 0.05.