Effects of Endogenous Androgens and Abdominal Fat Distribution on the Interrelationship Between Insulin and Non-Insulin-Mediated Glucose Uptake in Females

Abstract

Background:

Polycystic ovary syndrome (PCOS) is associated with hyperandrogenism and insulin resistance. Glucose disposal occurs via noninsulin-mediated glucose uptake (NIMGU) and insulin-mediated glucose uptake (IMGU). It is unknown whether in PCOS NIMGU increases to compensate for declining IMGU and whether androgens and fat distribution influence this relationship.

Objectives:

The objective of the study was to compare in women with PCOS and controls the interrelationship between NIMGU [ie, glucose effectiveness (Sg)] and IMGU [ie, the insulin sensitivity index (Si)] and the role of androgens and fat distribution.

Participants:

Twenty-eight PCOS (by National Institutes of Health 1990 criteria) and 28 control (age, race, and body mass index matched) women were prospectively studied. A subset of 16 PCOS subjects and 16 matched controls also underwent abdominal computed tomography.

Main Outcome Measures:

Glucose disposal (by a frequently sampled iv glucose tolerance test), circulating androgens, and abdominal fat distribution [by waist to hip ratio and visceral (VAT) and sc (SAT) adipose tissue content] were measured.

Results:

PCOS women had lower mean Si and similar Sg and abdominal fat distribution compared with controls. PCOS women with Si below the PCOS median (more insulin resistant) had a lower mean Sg than controls with Si above the control median (more insulin sensitive). In PCOS only, body mass index, free T, modified Ferriman-Gallwey score, and waist to hip ratio independently predicted Sg, whereas Si did not. In PCOS, VAT and SAT independently and negatively predicted Si and Sg, respectively.

Conclusion:

The decreased IMGU in PCOS is not accompanied by a compensatory increase in NIMGU or associated with excessive VAT accumulation. Increased general obesity, SAT, and hyperandrogenism are primary predictors of the deterioration of NIMGU in PCOS.

The polycystic ovary syndrome (PCOS) is an important and highly prevalent hyperandrogenic disorder, affecting approximately 7% of reproductive-aged women and is frequently associated with insulin resistance (IR) and obesity. Approximately 70% demonstrate IR and hyperinsulinemia beyond that due to their degree of obesity (1). Women with PCOS are at substantial risks for developing metabolic abnormalities, including type 2 diabetes mellitus (T2DM), metabolic syndrome, and cardiovascular disease (CVD) (1).

Although insulin is recognized as crucial in regulating glucose uptake in insulin-sensitive tissues such as skeletal muscles and adipose, all tissues are capable of glucose uptake by facilitated diffusion via non-insulin-mediated glucose uptake (NIMGU). Methodologically, insulin-mediated glucose uptake (IMGU) and NIMGU are traditionally estimated by the basal (for NIMGU) and hyperinsulinemic euglycemic (for IMGU) clamps (2, 3). In addition, the insulin sensitivity index (Si) and glucose effectiveness (Sg) values, respectively, determined by the minimal model (MINMOD) based on the modified frequently sampled iv glucose tolerance test (FSIVGTT) (2–5) can also be used.

NIMGU accounts for about 75% of glucose uptake in the fasting basal state (6–8), and alterations in NIMGU play an important pathogenic role in disorders of carbohydrate metabolism, with several studies evaluating strategies that could potentially augment NIMGU (7). Some reports suggest that, in many insulin-resistant states such as T2DM (8) and adrenocortical hyperactivity (9), NIMGU is elevated as insulin sensitivity declines. This relationship has been interpreted as a vital compensatory mechanism to ensure that sufficient glucose uptake into the cell occurs during periods of impaired IMGU. In fact, in those individuals with a positive family history of T2DM, low Si and Sg appear to precede the development of T2DM (10, 11). How this relationship of NIMGU to IMGU might change in hyperandrogenic insulin resistant females, such as those with PCOS, has not been investigated.

Because pancreatic β-cell function is generally preserved in PCOS women (1), alterations in NIMGU and IMGU are likely major determinants of glucose tolerance and IR in these patients. However, a majority of previous studies on glucose metabolism in PCOS have focused on measures of IMGU (1). The few studies that have evaluated Sg (reflecting NIMGU) in PCOS women have yielded conflicting results and are limited by their small sample size (2, 12–14).

Adiposity and its body distribution may also play a role in determining abnormalities of IMGU and NIMGU. Although IR, CVD, and T2DM are associated with increasing body mass, they are more strongly correlated with the presence of abdominal or visceral obesity (15–17). In females, androgen excess has been associated with increased abdominal adiposity and IR (18, 19). However, it is unknown whether androgens and/or abdominal fat distribution influence NIMGU or its relationship to IMGU. Several studies have examined the association between IR and abdominal obesity in both the PCOS (20–23) and the general (15–17) populations, although recent MRI studies reported no differences in visceral abdominal fat content between PCOS and controls (24, 25). Overall it is still unclear whether adiposity in general or abdominal fat depot in particular, is associated with declines in NIMGU, or impacts its relationship to IMGU.

In the present study we have hypothesized that in PCOS NIMGU is inversely associated with IMGU and that androgen levels and abdominal fat distribution are associated with alterations in NIMGU and/or its relationship to IMGU.

Materials and Methods

Study population

Twenty-eight women with PCOS and 28 [age, race and body mass index (BMI) matched] healthy control women, aged 22–44 years, were prospectively studied at a tertiary care outpatient reproductive endocrinology practice from 2005 to 2009. The presence of PCOS was defined by the 1990 National Institutes of Health consensus criteria, namely the presence of oligoovulation and biochemical or clinical hyperandrogenism, excluding other known endocrinopathies as previously described (1, 26). The control women were recruited through advertisements and comprised healthy premenstrual women with eumenorrhea and no evidence of hyperandrogenism or endocrine disorders. None of the women were taking any hormonal medication (including oral contraceptives) for at least 3 months before the evaluation.

To ensure matching groups, we generally first recruited a PCOS subject and then selected a control to match (either from our pool of controls or by specifically seeking a matching control). Our overall recruitment protocol matched controls and PCOS subjects by BMI (±3 kg/m²), age (±5 years), and race. However, this in no way should imply that the results of 1 control were matched with the results of any individual PCOS subject. Rather, we used this recruitment strategy to ensure that our study groups matched to the extent possible.

All subjects underwent a physical examination with blood sampling for hormone measurements, as previously described (26, 27), and were normoglycemic by a 2-hour oral glucose tolerance testing. In addition to height, weight, and modified Ferriman-Gallwey (mFG) score, waist circumference (WC) was measured at the narrowest portion of the torso approximately midway between the lower costal margin and the iliac crest, and the hip circumference was measured over the widest portion of the gluteal and greater trochanteric region. The BMI and waist to hip ratio (WHR) were then calculated.

All subjects also underwent a modified FSIVGTT, and a subset of 32 subjects (16 PCOS and 16 matched controls) also underwent assessment of abdominal fat distribution by a single-cut abdominal computed tomography (CT) scan. The study was approved by the Institutional Review Board Cedars-Sinai Medical Center (Los Angeles, California). All subjects provided written informed consent before study entry.

Biochemical analysis

Fasting blood samples for circulating total T, free T, dehydroepiandrosterone sulfate (DHEAS), SHBG, insulin, and glucose concentrations were obtained on days 3–8 of the menstrual cycle.

Total T was measured using high-turbulence liquid chromatography tandem mass spectrometry and free T determined by equilibrium dialysis (Quest Diagnostics, San Juan Capistrano, California). The serum levels of DHEAS were measured by a competitive immunoassay (Modular E170; Roche Diagnostics, Indianapolis, Indiana). Insulin was assayed by chemiluminescence (ADVIA Centaur chemiluminescent immunoassay system; Siemens Healthcare, Deerfield, Indiana). Serum glucose levels were measured using the hexokinase/glucose-6-phosphate dehydrogenase method (Roche Applied Sciences, Indianapolis, Indiana).

Metabolic assessment

The FSIVGTT was performed as previously described (2). Briefly, after an overnight fast, 2 iv catheters were placed in each forearm between 8:00 and 9:00 am. Thereafter, iv administration of glucose (0.3 g/kg) was followed in 20 minutes by the administration of regular insulin (0.03 U/kg). Blood samples (2.0 mL) were collected 34 times from −20 minutes (relative to glucose administration) to +180 minutes. Samples drawn into prechilled tubes containing EDTA (for insulin) or sodium fluoride potassium oxalate (for glucose) and plasma were frozen at −80°C until assayed. Plasma glucose and insulin values were entered into the MINMOD computer program, and data were analyzed using MINMOD of glucose kinetics to establish glucose-insulin interactions in a single assay (3). The calculated components of the modified FSIVGTT were as follows: the acute insulin response to glucose (AIRg), which reflects the first phase endogenous insulin secretion in response to a bolus glucose injection; the Si, the insulin-mediated glucose uptake per unit of insulin evaluated by a bolus injection of known quantity of insulin; the Sg, the ability of glucose per se, independent of changes in insulin, to increase glucose uptake and suppress the endogenous output; and the disposition index (DI; Si × AIR), representing the interaction of insulin sensitivity and the compensatory ability of the β-cell to secrete insulin.

Single-slice CT scan

A single slice CT (Toshiba Aquilion 16, model TSX-101A; Toshiba America Medical Systems, Tustin, California) scan of abdominal visceral and sc fat was obtained at the level of the L4-L5 lumbar vertebrae disc space, as previously described (28). Regional abdominal fat stores were identified by creating a closed region of interest surrounding the peritoneal cavity to separate sc and visceral fat stores. Connected voxels within the CT attenuation range of −190 to −30 HU were identified as fat. Fat voxels inside the drawn region of interest were classified as visceral adipose tissue (VAT) and those outside as sc adipose tissue (SAT). The cross-sectional areas of VAT and SAT in square centimeters were quantified using software-derived algorithms.

Statistical analysis

Shapiro-Wilks W-statistics was used to determine whether nominal variables were normally distributed. All continuous variables but mFG score and homeostasis model of β-cell function (HOMA-%β-cell function) reasonably follow the parametric normal distribution on the original or log scale, with 8 variables needing a log transformation [AIRg, DI, Si, total T, free T, fasting glucose, fasting insulin, and homeostasis model assessment insulin resistance index (HOMA-IR)]. Intergroup differences were evaluated using the unpaired t test for normally distributed continuous variables or the Wilcoxon rank-sum test for mFG score and HOMA-%β-cell function. A χ² test was used to compare nominal variables. Bivariate correlations between continuous variables and parameters of glucose disposal in PCOS and controls were analyzed using the Pearson correlation coefficient for all variables, except mFG score and HOMA-%β-cell function, which were analyzed using the Spearman correlation coefficient. Multiple logistic regression analysis (backward stepwise model) was used to test the independent contribution of each of the variables that seemed to be important risk factors. Log total T, log free T, log HOMA-IR, and log Si were used for the multiple regression analysis instead of their natural values. The statistical level of significance was set at P = .05 and all hypothesis tests were 2 sided.

Results

Participant characteristics

Subjects' basic demographic, endocrine, and metabolic characteristics are described in Table 1. In all, we studied 56 nondiabetic subjects (28 PCOS and 28 healthy controls). Three of the PCOS subjects (10.7%) were of normal weight (BMI < 25 kg/m²); 8 (28.6%) were overweight (BMI ≥ 25.0–29.9 kg/m²); and 17 (60.7%) were obese (BMI ≥ 30 kg/m²). Their age ranged from 22 to 43 years in PCOS and 21–44 years in controls. As noted above, the PCOS and control groups were matched for age, race, and BMI. In women with PCOS, mFG scores, baseline mean serum total T, free T, and DHEAS levels were higher than in controls. There was no difference in mean WHR, WC, fasting insulin, fasting glucose, HOMA-IR, and HOMA-%β-cell function between women with PCOS and controls. The short- and long-term β-cell functions, measured by mean AIRg and DI, respectively, did not differ between women with PCOS and controls.

Table 1.

Characteristics of Study Populations

Variable	PCOS (n = 28)		Controls (n = 28)		P Value
	Mean	SEM	Mean	SEM
Age, y	29.2	0.9	33.9	1.3	.006 3
BMI, kg/m2	32.3	1.4	29.5	1.21	.143 3
WC, cm	108.7	5.8	113.1	5.9	.596 2
WHR	0.88	0.02	0.84	0.02	.200 4
mFG scorea	8.0	1.02	1.11	0.3	.000 1
Total T, ng/dLb	38.4	3.6	20.8	1.9	.000 1
Free T, pg/mLb	4.6	0.5	1.9	0.2	.000 1
DHEAS, μg/dL	313.5	28.6	181.3	24.9	.001 8
Fasting insulin, μIU/mLb	11.0	1.9	9.2	1.6	.483 9
Fasting glucose, mg/dLb	89.8	2	88	3.5	.635 9
HOMA-IRb	2.4	0.4	2	0.4	.456 4
HOMA-%β-cell functiona	62.6	14.5	41.62	6.1	.776 1
AIRg, μU−1/mLb	439.1	65.7	359.2	47.1	.392 4
DIb	1687.7	231.3	2306.1	579.2	.088 2
Si, min−1/μU−1 · mL−1b	2.6	0.4	4.55	0.9	.037 8
Sg, min−1	0.0222	0.002	0.0251	0.0021	.310 7

P > .05 is considered significant.

^aAnalysis by unpaired t test, and P values are reported for all variables except Wilcoxon P value for mFG score and HOMA%β-cell function.

^bGeometric means, the antilog of the log scale mean, is reported for log-transformed data.

Because of the small number of subjects who were African Americans, Asians, or of mixed races, these subjects were classified together as “other” in this study. The vast majority of patients were either Hispanic white (HW) or non-Hispanic white (NHW). PCOS subjects comprised 10 NHW (35.7%), 11 HW (39.3%) and 7 others (25%), whereas there were 14 NHW (50%), 8 HW (28.6%), and 6 others (21.4%) in the control group. The proportion of subjects or mean Sg and Si did not differ by racial/ethnic background when PCOS women were compared with controls. Age was unlikely to confound the interrelationship between Sg and Si because no significant relationships were observed between age and Si or Sg in the study groups in both bivalent and multiple regression analysis.

Association of glucose effectiveness to insulin sensitivity

The differences in mean glucose effectiveness and insulin sensitivity between PCOS and controls are also described in Table 1 and depicted in Figure 1. As expected, mean Si was lower in women with PCOS than controls, confirming the presence of IR in the former (Table 1 and Figure 1A). Alternatively, mean Sg was similar in PCOS women compared with controls when all subjects were considered together (Table 1).

Figure 1.

Levels of Si (A) and Sg (B) in PCOS and controls showing defects in NIMGU and IMGU. A, Comparison of Si in PCOS vs controls, showing Si being significantly lower in 28 PCOS women compared with 28 control females, confirming insulin resistance in the PCOS group (P < .0378). B, Comparison of Sg in 14 most insulin-resistant PCOS vs 14 most insulin-sensitive controls, showing Sg significantly lower in PCOS (P < .0286).

To magnify the differences observed, and recognizing the significant overlap in Si between PCOS and control women (Figure 1A), we reanalyzed the data including only those PCOS women with significant IR (ie, those with an Si value less than the median for the entire PCOS group, an Si < 2.57 min⁻¹/μU⁻¹ · mL⁻¹) and comparing them with the 14 most insulin-sensitive controls (ie, those with an Si level above the median for the entire control group, an Si > 4.39 min⁻¹/μU⁻¹ · mL⁻¹). In this analysis, mean (±SEM) Sg levels were lower in PCOS than controls [0.0187 ± 0.002 vs 0.0280 ± 0.003 (minute −1), P < .0281; Figure 1B]. Meanwhile, no difference in Sg was observed between IR and insulin-sensitive PCOS or controls.

Bivariate correlation coefficients between Sg and Si in PCOS and controls are presented in Figure 2. In women with PCOS, the association of Sg to Si (r = 0.36, P = .0593) (Figure 2) and other measures of IR such as HOMA-IR (P = .0550) did not reach the level of statistical significance based on threshold level of P < .05. Similarly, Sg demonstrated no association with fasting insulin levels (P = .0790), fasting glucose levels, and HOMA%β-cell function (r = −0.33; P = .0864). In controls, there was no significant bivariate relationship between Sg and Si (Figure 2) or other measures of IR. Multiple regression analysis also showed no independent association of Sg with Si or other measures of IR (Table 2).

Figure 2.

Correlation between Si with Sg in PCOS (A) vs controls (B).

Table 2.

Multiple Regression Analysis in PCOS and Matched Controls, With Sg as the Dependent Variable^a

Variable	PCOS		Control
	Partial r	P Value	Partial r	P Value
Log Si	0.266	.231	0.0553	.8386
Log free T	−0.488	.0213	−0.445	.0844
Log total T		NS		NS
mFG score	−0.475	.0254		NS
WHR	0.469	.0275		NS
BMI	−0.73	.0001		NS
Log HOMA-IR		NS		NS
r2, %	74.2		26.05

Abbreviation: NS, nonsignificant. Bold indicates significant P values.

^aBackward stepwise model using all 12 candidates (log Si, log free T, log total T, mFG score, WHR, BMI, log HOMA-IR, DHEAS, log insulin, log glucose, WC); log Si is forced in.

Association of glucose effectiveness to hyperandrogenism and adiposity

To investigate whether the relation between Si and Sg was modulated by hyperandrogenism and adiposity, multiple linear regression analyses were performed, using a backward stepwise model, with Sg as the dependent variable, and maintaining log Si as an independent variable in the model (Table 2). Other potential independent variables included variables related to androgenicity (mFG score, log, free T, total T, and DHEAS), adiposity (WC, WHR, and BMI), other measures of IR (log glucose, log fasting insulin, log HOMA-IR), and age. BMI, free T, and mFG score were found to be negative and WHR positive independent predictors of NIMGU in PCOS. In women with PCOS about 74.2% of the variation in Sg can be explained by the variables measured in this model (Table 2; r² = 74.2%, P < .0001). In contrast, these variables demonstrated no independent association with Sg in the controls.

To determine which fat depot maintains independent association with measures of glucose uptake, we studied the correlation between Sg and more precise measurements of body fat distribution in a subgroup of 16 PCOS subjects and 16 controls who also underwent an abdominal single-slice CT scan. Mean (±SEM) SAT (463.0 ± 50.6 vs 397.7 ± 45.2 cm², respectively) and VAT (134.1 ± 20.2 vs 117.0 ± 14.2 cm², respectively) did not differ between women with PCOS and controls. A multiple linear regression model was constructed using Sg or log Si as the dependent variable, and SAT and VAT as the independent variables, in the 16 women with PCOS and 16 female controls matched for age, race, and BMI (Table 3). We observed that the association between Sg and VAT in PCOS was lost after adjusting for SAT, leaving SAT as the stronger independent and negative predictor of Sg. Alternatively, VAT was found to be the stronger independent determinant of Si in PCOS. In controls SAT, but not VAT, correlated negatively with Si, and neither VAT nor SAT was an independent predictor of Sg in these subjects.

Table 3.

Multiple Regression Analysis of Association Between VAT and SAT and Between Sg and Si in Women With PCOS and Matched Controls

Variables	PCOS (n = 16)				Control (n = 16)
	Sg		Log Si		Sg		Log Si
	Partial r	P Value	Partial r	P Value	Partial r	P Value	Partial r	P Value
VAT, cm2	−0.335	.2218	−0.602	.0175	−0.498	.059	0.459	.0851
SAT, cm2	−0.547	.0347	−0.102	.7178	0.367	.1779	−0.558	.0247
r2, %	51.51		44.92		25.60		33.50

Bold indicates significant associations.

Discussion

Our data indicate first, and contrary to what we had hypothesized, that declining Si (IMGU) was not associated with increasing Sg (NIMGU) in PCOS. Although Si, as expected, was lower in women with PCOS than controls, Sg was overall similar between the 2 groups when all subjects were considered together. These data suggest that in general, and contrary to what happens with IMGU, NIMGU is relatively preserved in PCOS. This notion is reinforced by our observation that Sg was not independently related to Si. However, the preservation of NIMGU is not complete because Sg values were lower in our most insulin-resistant PCOS subjects compared with our most insulin-sensitive controls. Our data did indicate that in PCOS, but not controls, deterioration in NIMGU is independently related to increasing general obesity, sc adiposity, and hyperandrogenism.

We expected for NIMGU to increase in patients with PCOS as Si (IMGU) declined. And although we did not observe this inverse relationship, we did observe that NIMGU was significantly preserved in PCOS, albeit incompletely. The reason(s) for the failure of NIMGU to compensate for declining IMGU in our study is unknown but may relate to several factors including disturbances in the functional relationships between glucose transporter (GLUT)-1 and GLUT-4 transporter mechanisms (29, 30), reduced ability of glucose to enhance its own uptake through its own mass action, or a decrease in the availability of insulin-independent glucose transporters (eg, GLUT-1) or sodium glucose cotransporter (6, 7) or hexokinase isomer (HK; HK-1 and HK-2) (31).

Our observation of preserved Sg when all PCOS subjects were considered together, but Sg decline in the most insulin-resistant subgroup reflects the significant overlap in insulin sensitivity between PCOS and controls because some women with PCOS are not insulin resistant at all, and even normal-appearing controls may demonstrate IR. Indeed, previous studies suggest that ∼30% of women with PCOS do not demonstrate overt IR (1). It is also possible that there may be a threshold level of IR above which defects in NIMGU become detectable. Further studies are warranted to elucidate the underlying etiologies responsible for the declining NIMGU in PCOS.

Our results also demonstrated that abdominal fat content and distribution, as determined by WC, WHR, SAT, and VAT, did not differ substantially between PCOS women and controls, consistent with previous studies (27, 28), suggesting that any defects in IMGU and NIMGU in PCOS are not merely the result of increased abdominal obesity. In addition, in PCOS the degree of sc adiposity appeared to be the more important determinant of Sg, whereas visceral fat content was a stronger and negative correlate of Si. In controls, both VAT and SAT were not associated with Sg. These data suggest that sc fat in women with PCOS, in contrast to controls, may differ in its ability to modulate glucose uptake, a finding consistent with recent molecular findings in our laboratory (32).

Few studies have investigated the relationship between androgens and Sg in PCOS (12–14, 33). The association between androgenicity and Sg observed in our study contrasts with some previous reports that suggest that androgen excess of ovarian origin does not affect Sg, although these are small studies including between 6 and 12 PCOS patients (12–14). We previously observed a positive correlation between Sg and adrenal androgen response to ACTH stimulation, including basal DHEAS in PCOS subjects (9 PCOS and 9 control women) (2). The underlying reasons for lack of correlation between DHEAS and Sg in the present study are unexplained, with the exception that we now include a greater number of subjects, which carries with it a greater degree of intersubject variation. However, we should note that in adipose tissue, androgen synthesis and interconversion can take place (34) and androstenedione perhaps testosterone may be derived from DHEA, reducing the distinction between the effects of DHEAS (a dehydroepiandrosterone prohormone and storage pool) and that of other androgens. In future studies, it will be interesting to determine the relationship between androstenedione levels and measures of glucose uptake.

The prevailing view until recently has been that hyperandrogenism favors accumulation of visceral fat, thereby promoting IR (18–22). This view has been challenged by our current results and by recent MRI data from other investigators demonstrating no significant differences in VAT or SAT accumulation between PCOS- and BMI-matched controls, despite significant differences in IR measured by HOMA-IR (24) or euglycemic hyperinsulinemic clamp (25). In fact, although possibly the consequence of too few subjects, examining lean women with nonhyperandrogenic PCOS, Dolfing et al (35) actually found smaller visceral fat volume as measured by MRI in 10 lean PCOS women compared with BMI-matched controls. PCOS in these 3 studies was diagnosed using the broader European Society of Human Reproduction and Embryology/American Society for Reproductive Medicine criteria, which is less frequently associated with metabolic dysfunction (1).

Taken together, our data support emerging evidence demonstrating that the metabolic abnormalities of PCOS is not primarily the result of increased abdominal fat content (18–22) but may be linked to abnormal adipocyte function or morphology as we and others have demonstrated (25, 32). Longitudinal studies will be required to test the hypothesis that preferential accumulation in the VAT or SAT compartments over time leads to impairment of IR in PCOS or vice versa.

Prior studies have demonstrated depot-specific functional differences, including variations in gene expression, hexokinase isomer (HK-1 and HK-2) and GLUT expressions (16, 17, 31). The stronger negative association of Si with VAT vs SAT content in our PCOS women affirms previous observations that demonstrated that VAT burden correlates better with IR, and the risks of T2DM and CVD, than does SAT (24, 29). However, because IR is determined by not only IMGU but also NIMGU, and increasing SAT content is associated with good health in some studies (19, 20), the strong negative association of Sg with SAT content requires further investigation. We have recently demonstrated evidence of adipogenic dysfunction involving impaired glucose transport, exaggerated inflammatory markers, and impaired adiponectin secretion in PCOS adipocytes from SAT (32), suggesting that SAT-related adipocyte dysfunction may contribute to the abnormal glucose uptake in PCOS.

Overall, our results suggest that perturbations in insulin-dependent glucose use (IMGU) are not accompanied by a compensatory increase in NIMGU or reflected a preferential increase in visceral adiposity accumulation. In fact, NIMGU was generally preserved in PCOS, albeit lower in the most insulin-resistant PCOS subjects. Furthermore, increases in general obesity, sc adiposity, and hyperandrogenism are important determinants of deteriorating NIMGU, independent of declining insulin sensitivity, in PCOS. As abdominal adiposity and pathways of glucose uptake become targets for therapeutic interventions (7, 20), further studies are required to confirm the relationship between Sg and sc fat in longitudinal studies using a larger sample size and to fully understand the mechanisms by which androgens, general adiposity, and sc fat interact with NIMGU and IMGU to produce the IR usually observed in PCOS.