Total and High-Molecular Weight Adiponectin in Women with the Polycystic Ovary Syndrome

Abstract

Objective

Adiponectin, an adipokine with antidiabetic properties, forms multimers, and the high molecular weight (HMW) form most closely correlates with insulin sensitivity. Therefore, we hypothesize that HMW adiponectin levels are decreased in women with polycystic ovary syndrome (PCOS), a condition characterized by insulin resistance, compared to normal controls, and that HMW adiponectin correlates with testosterone and insulin sensitivity.

Design and patients

Cross-sectional study involving 13 women with PCOS and 13 age- and BMI-matched normal controls.

Measurements

Waist-to-hip ratios (WHR), glucose, insulin, SHBG, total testosterone, total and HMW adiponectin levels were measured after an overnight fast. Free testosterone was calculated from SHBG and total testosterone, and insulin sensitivity (S_i) was determined using a frequently sampled intravenous glucose tolerance test. The study’s primary outcomes were differences in total and HMW adiponectin between women with PCOS and normal control women.

Results

Total adiponectin (p<0.01), HMW adiponectin (p<0.01), and the ratio of HMW to total adiponectin (S_A) (p=0.03), were lower in women with PCOS compared to normal women. Total and HMW adiponectin levels correlated inversely with WHR (p<0.01) and free testosterone (p<0.01) and positively with S_i (p<0.001). Using forward stepwise multivariate analysis, HMW adiponectin and WHR, but not PCOS status, were independent predictors of S_i.

Conclusions

Women with PCOS have lower total and HMW adiponectin levels compared with normal women. HMW adiponectin also comprises a smaller proportion of total circulating adiponectin in women with PCOS. Alterations in HMW adiponectin levels in women with PCOS may contribute to the insulin resistance intrinsic to the syndrome.

Introduction

The polycystic ovary syndrome (PCOS) is characterized by chronic anovulation and hyperandrogenism and affects an estimated 6–10% of women of reproductive age (1). Although much remains unknown regarding the pathophysiology of PCOS, insulin resistance appears to play a central role in the syndrome’s development. Lean women with PCOS appear to have a form of insulin resistance intrinsic to the syndrome (2). In addition to this intrinsic insulin resistance, obese women with PCOS also demonstrate the burden of insulin resistance associated with excess adiposity (2). In women with PCOS, insulin resistance is manifest clinically in increased incidence of glucose intolerance and overt type 2 diabetes (1).

Adipose tissue is an active endocrine organ, releasing a variety of bioactive peptides and adipokines that modulate the body’s metabolism at local and systemic levels (3). One specific adipokine, adiponectin, has a strong inverse relationship with obesity and insulin resistance and may have unique antidiabetic, anti-inflammatory, and antiatherogenic properties (3;4). Adiponectin levels are not only decreased in patients with impaired glucose tolerance and type 2 diabetes, but hypoadiponectinemia also independently predicts the future development of type 2 diabetes in healthy individuals (4).

In regards to alterations in adiponectin levels among women with PCOS, studies have reported conflicting results, with some groups documenting decreased levels of adiponectin in PCOS women compared with weight-and body mass index (BMI)-matched controls (5–10), but other studies showing no difference after controlling for obesity (11–13). Even among studies demonstrating lower adiponectin levels in PCOS women, the relationships among adiponectin and central obesity, insulin resistance and testosterone remain unclear (5;7;9;10;14).

Adiponectin forms multimers through disulfide bonds and exists in several oligomeric forms in serum: low molecular weight (LMW) complexes comprised of trimers, hexamers, and high molecular weight (HMW) multimers consisting of 12 to 18 subunits (3). Increasing evidence suggests that the various isoforms of adiponectin have different biologic activities, with the ratio of HMW adiponectin to total adiponectin, not the absolute serum adiponectin level, most closely correlating with measures of insulin sensitivity (3;15). Furthermore, in high-risk adults, decreased HMW adiponectin at baseline is a stronger risk factor for progression to type 2 diabetes than total adiponectin level (16).

Differences in adiponectin multimers, as opposed to total adiponectin levels, may explain the apparent contradictions between adiponectin levels and anthropometric and biochemical characteristics reported in previous studies of women with PCOS. Testosterone selectively inhibits the in vitro secretion of HMW adiponectin by adipocytes (17), supporting the hypothesis that HMW adiponectin may be decreased in women with PCOS, a syndrome characterized by hyperandrogenism.

We hypothesized that women with PCOS would demonstrate lower HMW adiponectin levels than normal women after controlling for obesity and that HMW adiponectin levels would correlate inversely with insulin resistance and serum testosterone levels. To test this hypothesis, we compared HMW adiponectin levels in women with PCOS with those of age- and BMI-matched normal women. Additionally, we investigated the relationship between adiponectin multimers and insulin resistance in women with and without PCOS.

Materials and Methods

Twenty-six women (13 PCOS women and 13 normal women) enrolled in a cross-sectional study previously performed at Virginia Commonwealth University (VCU) (18) were matched for BMI and age using multivariate minimum distance matching with the nearest neighbor approach. All women were between 18–40 years of age and had BMIs ≤40 kg/m². None of the women had diabetes mellitus or received oral contraceptives or other medications known to affect insulin sensitivity for at least 2 months prior to study participation. PCOS was defined according to 1990 NICHD conference criteria: oligomenorrhea (≤8 menstrual periods in the previous year); hyperandrogenism (elevated serum total or free testosterone concentration); and exclusion of secondary causes of ovulatory dysfunction or hyperandrogenism (19). Specifically, serum prolactin, thyroid function tests and 17α-hydroxyprogesterone levels were obtained to exclude hyperprolactinemia, thyroid dysfunction, and non-classical adrenal hyperplasia respectively. Women in the PCOS group with impaired glucose tolerance (IGT) were not excluded since IGT is a common co-morbidity in PCOS. Normal women did not have clinical evidence of hyperandrogenism; had regular menstrual cycles, normal androgen levels, and normal glucose tolerance; and did not have a history of gestational diabetes or a first-degree relative with diabetes. The study was approved by the institutional review board at VCU and each subject gave written informed consent.

Subjects were admitted to the General Clinical Research Center after a 12-hour overnight fast. PCOS women were studied during the equivalent of the follicular phase of the menstrual cycle as documented by a serum progesterone ≤6 nmol/L. Normal women were studied during the mid-follicular phase of menstrual cycles (days 5–9), which most closely approximates the hormonal milieu of anovulatory women with PCOS. Height and weight were measured to the nearest 0.1 cm or 0.1 kg using a precision stadiometer or digital scale, respectively. Waist circumference (WC) and hip circumference (HC) were measured to the nearest 0.1 cm at a level midway between the lowest rib margin and the iliac crest and at the widest level over the greater trochanters, respectively. Waist-to-hip ratios (WHR) were determined from WC and HC values. Blood pressure was measured with each subject supine using an appropriately sized cuff and an automated device (Dynamap Pro 100, General Electric).

On the first day, fasting baseline laboratory tests and a 2-h oral glucose tolerance test (OGTT) with 75 g dextrose were performed. Fasting blood samples were obtained for determination of plasma insulin, glucose, sex hormone binding globulin (SHBG) and total and HMW adiponectin. During the OGTT, blood samples were collected every 15 minutes for determination of serum glucose and insulin concentrations. Impaired fasting glucose (IFG) and impaired glucose tolerance (IGT) were defined using current American Diabetes Association (ADA) recommendations (20).

On the second day, after a 12-h overnight fast, insulin sensitivity was determined by a frequently sampled intravenous glucose tolerance test (21). At time zero, 300 mg/kg dextrose was administered intravenously and 0.03 units/kg insulin was administered intravenously 20 minutes later. Twenty-seven blood samples were collected for determination of insulin and glucose during the 3-h duration of the protocol. Data were analyzed with the Minimal Model Identification Software (MINIMOD Millennium, version 6.02, 2001) (22), which yields quantitative determination of tissue insulin sensitivity (S_i).

Laboratory assays

Blood samples were centrifuged immediately and sera were stored at −70°C until assayed. Plasma glucose levels were measured using glucose oxidase methodology (YSI 2300 Stat Plus Glucose Analyzer Yellow Springs Instruments [YSI], Yellow Springs, OH), and plasma insulin levels were measured using an ELISA for human insulin (ALPCO Diagnostics, Windham, NH). Serum total testosterone levels were was determined by radioimmunoassay (Diagnostic Products, Los Angeles, CA). SHBG concentrations were determined as previously described (23;24), and serum free testosterone was calculated by the method of Sodergard et al. (25), assuming a serum albumin concentration of 40 g/L. To avoid interassay variation, all samples were analyzed in duplicate in a single assay for each hormone. The intra-assay coefficient of variation (CV) for insulin and steroid hormone assays were 5.5% and <10%, respectively.

Quantification of total and HMW adiponectin species was performed using a double monoclonal sandwich ELISA method (Daiichi Pure Chemical, Tokyo, Japan, distributed by ALPCO Diagnostics, Windham, NH). This method utilizes pretreatment with proteases for selective measurement of human multimeric adiponectin. This assay demonstrates a sensitivity of 0.04 ng/ml, an inter-assay CV <15%, and an intra-assay CV of 5.3% and 3.3% for total and HMW adiponectin respectively (26). Adiponectin levels were described as the following: total adiponectin, HMW adiponectin (absolute serum level determined by ELISA), and the ratio of HMW to total adiponectin (S_A).

Statistical analysis

Results not normally distributed were log-transformed for all statistical analyses and reported back-transformed in their original units. All results were reported as means, or geometric means for log-transformed variables, with 95% CIs. P-values <0.05 were considered significant. All statistical analyses were performed using JMP 7.0 software (SAS Institute, Cary, NC).

Variable comparisons between PCOS and normal women were made with the Student’s unpaired two-tailed t-test. Although paired t-tests could be performed in light of the matched pairs of PCOS and normal women, analyses were performed with t-tests for independent samples to provide more conservative estimates. Univariate correlation analysis was performed using Pearson’s correlation test. A forward stepwise multiple linear regression analysis was performed to determine the best model to predict S_i considering biologically plausible predictors of S_i, i.e. PCOS status, BMI, WHR, total testosterone, free testosterone, total adiponectin, HMW adiponectin, and S_A. These independent variables were selected based on literature and/or because their correlations with S_i were significant in the univariate analyses. Biologically plausible interactions among these variables (PCOS status × BMI, PCOS status × WHR, and free testosterone × HMW adiponectin) were also entered into the model.

Results

Table 1 contains the clinical and biochemical characteristics of women with PCOS and age- and BMI-matched normal control women. Women with PCOS had significantly higher WHR compared with normal women (P=0.02). Total and free testosterone values were also significantly higher in women with PCOS compared with controls (P<0.001). Mean AUC _glucose and AUC _insulin during the OGTT were both significantly higher in PCOS women compared with normal women (P <0.01 and <0.001 respectively). Mean insulin sensitivity (S_i) was substantially lower among women with PCOS compared with normal controls (P<0.01). Four of the women with PCOS had IGT; one woman with IGT also had IFG. None of the normal women had IFG or IGT.

Table 1.

Clinical and biochemical characteristics of women with PCOS and normal control women.

Variable	PCOS subjects	Normal subjects	P-value
n	13	13
Age (yr)	29.9 (25.7–34.1)	30.7 (26.5–34.9)	0.77
BMI (kg/m2)	30.7 (26.5–34.9)	30.0 (25.8–34.2)	0.80
Waist circumference (cm)	90.8 (82.0–99.6)	88.7 (79.9–97.5)	0.73
Hip circumference (cm)	109.0 (101.2–116.9)	115.1 (107.3–123.0)	0.27
WHR	0.83 (0.79–0.86)	0.77 (0.73–0.80)	0.02
Systolic blood pressure (mmHg)	116 (108.1–123.3)	113 (105.7–120.9)	0.65
Diastolic blood pressure (mmHg)	70 (64.7–74.6)	67 (61.6–71.5)	0.37
Insulin (ρmol/L) †	42.4 (27.1–66.0)	22.2 (13.9–35.4)	0.05
Glucose (mmol/L)	4.51 (4.20–4.80)	4.42 (4.12–4.72)	0.69
SHBG (nmol/L) †	87.9 (54.8–141.2)	139.8 (87.1–224.6)	0.17
Total testosterone (nmol/L) †	3.23 (2.45–4.25)	1.32 (1.00–1.74)	<0.001
Free testosterone calculation (ρmol/L) †	30.9 (19.8–48.9)	9.4 (5.9–15.0)	<0.001
AUC glucose (mmol·min−1·L−1) *	838.7 (772.0–905.4)	673.6 (607.0–740.3)	<0.01
AUC insulin (nmol·min−1·L−1) † *	48.43 (34.00–68.97)	17.63 (12.38–25.11)	<0.001
Insulin sensitivity (Si) ‡	4.3 (−0.14–8.73)	13.5 (9.04–17.91)	<0.01
Total adiponectin (µg/ml)	3.5 (2.47–4.45)	6.0 (5.00–6.99)	0.001
HMW adiponectin (µg/ml)	1.3 (0.68–1.93)	2.7 (2.11–3.35)	<0.01
HMW/total adiponectin	0.35 (0.30–0.41)	0.44 (0.39–0.50)	0.03

Data are means (95% CIs) unless otherwise noted.

^†Geometric means.

^*Determined during 2-hr oral glucose tolerance test;

^‡Determined by frequently sampled intravenous glucose tolerance test. To convert values for insulin to µIU/ml, divide by 6.945; to convert values for glucose to mg/dl, divide by 0.0555; to convert values for total testosterone to ng/dl, divide by 0.0347; to convert values of free testosterone to ng/dl, divide by 34.7.

AUC - area under-the-curve; BMI - body mass index; HMW - high molecular weight; SHBG - sex hormone binding globulin; WHR - waist-hip ratio.

Total and HMW adiponectin levels

Both total and HMW adiponectin were decreased in women with PCOS compared with normal women (P<0.01). Furthermore, among women with PCOS, HMW multimers comprised a smaller proportion of total circulating adiponectin levels (S_A) than in control women (P=0.03). The observed differences in total adiponectin (P<0.01), HMW adiponectin (P=0.01), and S_A (P=0.02) between PCOS and normal women remained significant even after excluding the group pairs containing women with IGT. Furthermore, given the difference in WHR between women with PCOS and controls, multivariate regression analysis was performed to determine if group differences in adiponectin remained significant after controlling for WHR. After adjusting for observed differences in WHR, total adiponectin (partial P=0.02) and HMW adiponectin levels (partial P=0.04) remained different between groups. Similarly, after controlling for group differences in S_i by separate multivariate regression analyses, differences in total adiponectin according to PCOS status remained significant (partial P=0.04). However, differences in HMW adiponectin levels between PCOS and normal women were no longer significant after controlling for S_i.

Correlation between adiponectin levels and clinical/biochemical characteristics

Given the association between hypoadiponectinemia and increased adiposity, correlations between adiponectin and anthropometric characteristics where performed (Table 2). Considering all women (n=26), both total and HMW adiponectin levels had strong inverse linear relationships (r=−0.58 for both) with WHR. A significant negative linear relationship between S_A and WHR (r=−0.52) was also observed. However, significant relationships among other anthropometric measures (WC, HC, or BMI) and adiponectin (including total adiponectin, HMW adiponectin, or S_A) were not identified.

Table 2.

Univariate analysis of anthropometric/biochemical characteristics and total adiponectin, high-molecular weight adiponectin, and the ratio of high-molecular weight/total adiponectin when women with PCOS and normal women were analyzed together (n=26).

	Total adiponectin		HMW adiponectin		HMW/total adiponectin
Variable	Correlation	p-value	Correlatio n	p-value	Correlation	p-value
BMI	−0.18	0.38	−0.21	0.30	−0.33	0.10
WHR	−0.58	<0.01	−0.58	<0.01	−0.52	<0.01
Fasting glucose	−0.15	0.47	−0.11	0.59	−0.04	0.85
Fasting insulin †	−0.32	0.11	−0.33	0.10	−0.39	0.045
Total testosterone †	−0.41	0.04	−0.41	0.04	−0.32	0.11
SHBG †	0.46	0.02	0.43	0.03	0.24	0.23
Free testosterone calculation †	−0.57	<0.01	−0.55	<0.01	−0.36	0.07
AUC glucose*	−0.53	<0.01	−0.43	0.03	−0.24	0.25
AUC insulin † *	−0.42	0.03	−0.42	0.03	−0.45	0.02
Insulin sensitivity (Si) ‡	0.63	<0.001	0.64	<0.001	0.54	<0.01

^†Log-transformed prior to analysis;

^*Determined during a standard 2-hr oral glucose tolerance test.

^‡Determined by frequently sampled intravenous glucose tolerance test.

AUC - area under curve; BMI - body mass index; HMW - high molecular weight; SHBG - sex hormone binding globulin; WHR - waist-hip ratio.

Next, the relationships between adiponectin and androgens were considered. Although total and HMW adiponectin correlated with total testosterone, stronger inverse relationships were observed between free testosterone and both total and HMW adiponectin (r= −0.57 and −0.55, respectively). Moreover, the inverse relationships between free testosterone and both total and HMW adiponectin remained significant after controlling for BMI by multivariate regression analysis (data not shown).

Finally, the relationships between parameters of glucose homeostasis and adiponectin were explored. There were no significant linear relationships between adiponectin and fasting glucose or insulin levels. However, total adiponectin, HMW adiponectin, and S_A all positively correlated with estimates of insulin sensitivity (S_i) (r=0.63, 0.64, and 0.54, respectively). Further supporting the strong correlation between adiponectin and insulin sensitivity, total and HMW adiponectin were also found to inversely correlate with both AUC _glucose and AUC _insulin determined during an OGTT (Table 2).

Determinants of insulin sensitivity (S_i)

Univariate correlations between S_i and anthropometric or biochemical characteristics potentially affecting insulin sensitivity are outlined in Table 3. A strong inverse linear relationship was observed between S_i and WHR (r=−0.63, P<0.001). Although the relationship between total testosterone and S_i did not attain statistical significance, S_i did correlate negatively with free testosterone (r=−0.44, P=0.02). As outlined previously, significant relationships were observed between S_i and total adiponectin, HMW adiponectin, and S_A.

Table 3.

Univariate analysis of anthropometric/biochemical characteristics and insulin sensitivity in all women (n=26).

	Insulin sensitivity (Si) ‡
Variable	Correlation	p-value
BMI	−0.31	0.12
Waist circumference	−0.38	0.06
WHR	−0.63	<0.001
Total testosterone †	−0.38	0.06
Free testosterone calculation †	−0.44	0.02
Total adiponectin	0.63	<0.001
HMW adiponectin	0.64	<0.001
HMW/total adiponectin	0.54	<0.01

^†Log-transformed prior to analysis;

^‡Determined by frequently sampled intravenous glucose tolerance test.

BMI - body mass index; HMW - high molecular weight; WHR - waist-hip ratio.

The best model to predict insulin sensitivity using stepwise multivariate analysis considering PCOS status, all the variables shown in Table 3, and biologically plausible interactions was S_i = 43.9+2.83·HMW adiponectin-60.8·WHR (adjusted R²=0.46, P<0.001). Therefore, both HMW adiponectin (positive association, partial P=0.03) and WHR (negative association, partial P=0.04) significantly and independently predicted S_i.

Discussion

Our study tested the hypotheses that women with PCOS have decreased levels of HMW adiponectin compared to normal women after controlling for obesity, and that HMW adiponectin would correlate inversely with serum testosterone and insulin resistance. In our population, both HMW and total adiponectin were lower in women with PCOS compared with normal control women. Additionally, in women with PCOS, HMW adiponectin comprised a smaller fraction of total adiponectin than in normal women.

In our study, women with PCOS had higher WHR than BMI-matched normal women. Furthermore, total adiponectin, HMW adiponectin, and S_A were all found to correlate negatively with WHR. However, alterations in visceral adiposity alone could not account for the observed differences in adiponectin between groups as both total and HMW adiponectin levels remained significantly lower in women with PCOS after adjusting for WHR.

In our population, we also observed significant inverse relationships between free testosterone and both total and HMW adiponectin; and, these findings are consistent with the results of other studies (17;27;28). Both total and HMW adiponectin levels are lower in adults males compared with females (3;17;29). This sexual dimorphism in adiponectin first appears in puberty and correlates with the pubertal increase in testosterone in males (28). In vitro, testosterone appears to selectively inhibit the secretion of HMW adiponectin by adipocytes (17). Lastly, total adiponectin (multimers were not assessed) increased in women with hyperandrogenism randomized to receive flutamide, an androgen antagonist (27).

Consistent with the hypothesis that insulin resistance is central to PCOS, we demonstrated that non-diabetic women with PCOS were less insulin sensitive than age- and BMI-matched normal women. It is possible that hypoadiponectinemia plays a role in the development of insulin resistance in PCOS. However, the literature on the relationship between total adiponectin and insulin sensitivity in PCOS women is conflicting. Most studies support a positive correlation between total adiponectin and insulin sensitivity (10;12–14); however, other studies have not confirmed this relationship (5;7;30;31).

These discrepancies may be influenced by differences in adiponectin multimerization. In our study, total adiponectin, HMW adiponectin, and S_A all exhibited strong positive relationships with S_i. However, WHR and HMW adiponectin, but not total adiponectin or S_A, were the most important predictors of S_i in step-wise multivariate analysis. These findings are consistent with data suggesting that the HMW complex is the biologically active form of adiponectin (3). It is also important to highlight that, in our study, PCOS group status was not a predictor of S_i independent of differences in HMW adiponectin. This finding supports the hypothesis that the increased burden of insulin resistance observed in women with PCOS may result, in part, from alterations in HMW adiponectin.

The few previous studies investigating the relationship between alterations in adiponectin multimerization among women with PCOS have demonstrated conflicting results (32–34). Similar to our findings, Aroda and colleagues (33) demonstrated decreased total and HMW adiponectin levels among 26 women with PCOS compared with 6 normal women. Although Aroda’s group did report lower ratios of HMW to total adiponectin in PCOS women demonstrating impaired glucose tolerance compared to those with normal glucose tolerance, they did not directly correlate adiponectin multimer status and insulin sensitivity.

Glintborg and colleagues (34) reported lower total adiponectin levels in 30 obese women with PCOS compared to 14 age- and BMI-matched healthy women, but they did not find significant differences in either absolute HMW adiponectin levels or S_A between groups. Considering only women with PCOS, the authors did report significant negative linear relationships between HMW adiponectin and both WHR and insulin sensitivity. However, unlike the study by Glintborg et al., we utilized multivariate analysis to determine which anthropometric and biochemical variables independently predicted S_i.

Lastly, Barber and colleagues (32) compared adiponectin multimer distribution between 50 PCOS cases and 28 female controls, including 22 BMI- and fat mass-matched pairs, and found that total and HMW adiponectin levels were lower in PCOS women. However, after adjusting for fat mass and age, these differences were no longer significant. In light of these findings, the authors concluded that the observed differences in adiponectin multimers between PCOS and normal women were attributable to differences in fat mass, and rejected the hypothesis that adiponectin plays a primary role in the development of PCOS independent of adiposity. Of note, many (n=21) of the PCOS women in the study by Barber et al. (32) where not metformin-niave. Although metformin was discontinued in these women 1 week prior to testing and significant differences in total and HMW adiponectin between metformin-niave and metformin-exposed women were not observed in their study, adiponectin levels may be altered by treatment with metformin (9).

Various methods for quantifying HMW adiponectin have been described in the literature and each of the outlined studies investigating alterations in adiponectin multimers in women with PCOS have utilized different techniques. Hence, differences in methodology make direct comparisons among studies difficult. Aroda and colleagues determined adiponectin multimerization status using gel electrophoresis under non-reducing conditions followed by Western blotting (33), whereas Glintborg et al. utilized fast protein liquid chromatography to determine the distribution of adiponectin multimers (34). Absolute values for HMW adiponectin were not determined directly by either of these methods. Instead, the absolute HMW adiponectin level was calculated by multiplying total serum adiponectin levels (determined by immunofluorometric or radioimmunoassay) by the estimated HMW fraction. Barber et al. (32) measured HMW adiponectin levels with an immunoassay using monoclonal antibodies.

In our study, HMW adiponectin levels were measured directly by a commercially-available ELISA that utilized sample pre-treatment with a protease that selectively digests non-HMW forms of adiponectin. Quantification of HMW adiponectin using this specific ELISA method has previously been shown to correlate closely with results from quantitative western blot analysis (26) and has been subsequently used by several investigators (35–37). Although direct measurement of HMW adiponectin using a standardized ELISA may represent a strength of our study, Bluher and colleagues reported that measurement of HMW adiponectin using the same ELISA technique was not superior to total adiponectin in assessing insulin sensitivity at baseline or in response to physical training in men and women with normal glucose tolerance, impaired glucose tolerance, and overt type 2 diabetes (37). At least two other ELISA systems have also been developed that directly measure HMW adiponectin (38;39); however, the precision and accuracy of each of the ELISA methods have not yet been directly compared. Consequently, further studies are needed to determine the most accurate and reliable method for measuring adiponectin multimers, particularly in large populations.

Another potential limitation of our study is the relatively small sample size. In this study, the differences in total and HMW adiponectin observed between groups were highly significant. Although it is unlikely that a larger sample size would change these relationships, a more robust sample may allow for the identification of other significant differences between groups. In particular, the sample size may limit the ability of multivariate regression analysis to identify independent variables or interactions that would be significant if a larger population were studied. Additionally, it is important to emphasize our cross-sectional study design can be used to identify alterations in adiponectin among PCOS women but cannot establish causality.

In summary, after controlling for age, weight status and central adiposity, we found that women with PCOS have lower HMW adiponectin levels compared with normal women. Furthermore, compared with normal women, women with PCOS appear to have alterations in adiponectin multimerization, with HMW adiponectin comprising a smaller proportion of total circulating adiponectin levels. Levels of HMW adiponectin also correlated inversely with WHR and free testosterone. Although total adiponectin and the ratio of HMW to total adiponectin were found to correlate with insulin sensitivity, absolute HMW adiponectin levels and WHR were identified as the strongest independent predictors of insulin sensitivity. After controlling for differences in insulin sensitivity, PCOS group status was no longer identified as an independent predictor of differences in HMW adiponectin levels. While the limited number of previous studies assessing adiponectin multimers in women with PCOS have yielded conflicting results, our findings support the hypothesis that alterations in HMW adiponectin among PCOS women may contribute to the insulin resistance intrinsic to the syndrome.