Association of adiposity indices with insulin resistance among women with polycystic ovary syndrome

Article information

Korean J Fertil Steril. 2025;.cerm.2024.07388

Publication date (electronic) : 2025 August 27

doi : https://doi.org/10.5653/cerm.2024.07388

Md Shahed Morshed, Hurjahan Banu, Md Shamim Hossan, Hafsa Mahrukh, Muhammad Abul Hasanat

Department of Endocrinology, Bangabandhu Sheikh Mujib Medical University (BSMMU), Dhaka, Bangladesh

Corresponding author: Md Shahed Morshed Department of Endocrinology, Bangladesh Medical University, Dhaka 1000, Bangladesh Tel: +880-1738-842019 Fax: +880-2-55165609 E-mail: shahed.phd.m22@bsmmu.edu.bd

*Partial funding was provided by Research & Development, BSMMU, Dhaka, Bangladesh (No. BSMMU/2022/820).

Received 2024 August 15; Revised 2024 December 4; Accepted 2024 December 31.

Abstract

Introduction

Polycystic ovary syndrome (PCOS) is a prevalent reproductive endocrinopathy with an increasing global burden [1]. The reproductive symptoms of PCOS, including irregular menstruation, subfertility, and cutaneous signs of androgen excess, are widely recognized. However, the impact of this condition extends beyond these issues; it also encompasses a range of metabolic disorders, such as obesity, dyslipidemia, impaired glucose tolerance, and steatotic liver disease associated with metabolic dysfunction, as well as an elevated risk of cardiovascular disease [2].

The metabolic manifestations of PCOS are primarily mediated by excess visceral adiposity, regardless of obesity. This adiposity triggers the production of various proinflammatory adipocytokines, leading to insulin resistance (IR) [3]. In PCOS, IR selectively impacts metabolic features. Compensatory hyperinsulinemia, resulting from IR, perpetuates the vicious cycle of PCOS by increasing the production of luteinizing hormone by the pituitary gland and androgens by the ovaries. Consequently, insulin sensitizers have become a common component in the management of PCOS, particularly for patients exhibiting metabolic features. Reducing IR can also improve reproductive outcomes [4,5]. Accordingly, the prediction of IR may have a role in therapeutic intervention for PCOS, including the prevention of cardiovascular morbidity and mortality.

Currently, the measurement of IR is not recommended in clinical practice. The gold-standard method for assessing IR, the hyperinsulinemic euglycemic glucose clamp (HIEGC), is impractical; thus, an indirect mathematical estimation of IR using the homeostatic model assessment (HOMA) is widely utilized with acceptable reliability in research settings. However, this method is not used in clinical practice and lacks a unified cutoff value [6]. IR is typically inferred from various surrogate markers, particularly anthropometric measurements such as body mass index (BMI) or waist circumference (WC) [7]. These traditional measurements have several limitations, most notably inadequate performance in differentiating between visceral and subcutaneous adipose tissue [8]. Consequently, numerous novel adiposity indices that combine serum cholesterol fractions with traditional measurements have been developed. The utility of these indices has been explored in various metabolic conditions, including diabetes mellitus [9]. Nevertheless, limited data are available regarding their usefulness in predicting IR among women with PCOS. In South Asian countries, which are densely populated and experiencing a rapidly increasing burden of non-communicable diseases, patients with PCOS tend to exhibit more metabolic features than those in other regions of the world [5]. Identifying IR among Bangladeshi patients with PCOS could facilitate their improved treatment. Furthermore, most studies have identified optimal adiposity indices based on the highest area under the curve (AUC), without conducting statistical comparisons. The present research aimed to clarify the associations of various adiposity indices with IR and to evaluate their utility as markers of IR through head-to-head statistical comparisons among Bangladeshi women with PCOS.

Methods

This cross-sectional study was conducted at the PCOS clinic of the Department of Endocrinology at Bangabandhu Sheikh Mujib Medical University (BSMMU) between January 2021 and December 2023. Participants were consecutively enrolled through convenience sampling. Ethical clearance was obtained from the Institutional Review Board of BSMMU (No. BSMMU/2021/7642), and informed written consent was acquired from each participant prior to enrollment.

Patients were diagnosed with PCOS following the 2018 International Evidence-based Guideline, which necessitates the presence of at least two of three criteria—irregular menstrual cycles, hyperandrogenism, and polycystic ovarian morphology—after a gynecological age of 8 years. Alternatively, diagnosis can be made before that age if both irregular cycles and hyperandrogenism are present [5]. The exclusion criteria for this study encompassed endocrine disorders that can mimic PCOS clinically and biochemically, such as hypothyroidism, hyperprolactinemia, androgen-producing tumor, and congenital adrenal hyperplasia, as determined by medical history and further investigations when indicated. Pregnant or lactating mothers, individuals undergoing treatment for diabetes mellitus, and those with ischemic heart disease, chronic liver disease (indicated by serum alanine aminotransferase levels greater than twice the upper limit of normal), or renal disease (evidenced by an estimated glomerular filtration rate of less than 60 mL/min/1.73 m² of body surface area) were also excluded. Additionally, participants who had taken medications associated with weight gain or loss (such as antipsychotics, antidepressants, or anticancer drugs), oral contraceptives, steroids, or drugs affecting IR (such as metformin or pioglitazone) within the 3 months prior to enrollment were not included in the study. Upon enrollment, participants were given a date and location to visit for testing in the morning after fasting for 8 to 12 hours. Physical examinations were conducted to assess height, weight, WC, blood pressure, the presence of acanthosis nigricans, and the modified Ferriman-Gallwey (mFG) score. Blood samples were then taken to evaluate fasting glucose level, lipid profile, and insulin level. To avoid interobserver variability, a single clinician performed all anthropometric measurements. All biochemical assessments were completed on the same day as sample collection. Glucose levels were determined using the glucose oxidase method, lipids were measured with the glycerol phosphate dehydrogenase peroxidase method, and insulin concentrations were assessed using a chemiluminescence immunoassay. The following formulas were employed to calculate the adiposity indices [10]:

• BMI=weight in kg/height in meters²

• WC=measured in centimeters

• Visceral adiposity index (VAI)=[WC in cm/(36.58+[1.89×BMI])]×[triglyceride (TG) in mmol/L/0.81]×(1.52/high-density lipoprotein cholesterol in mmol/L)

• Lipid accumulation (aggregation) product (LAP)=(WC in cm–58)×TG in mmol/L

• Triglyceride-glucose (TyG) index=ln[fasting TG in mg/dL×fasting plasma glucose in mg/dL/2]

• TyG-BMI=TyG×BMI

• TyG-WC=TyG×WC

IR was assessed using HOMA-IR, calculated as [(fasting glucose in mmol/L×fasting insulin in μIU/mL)/22.5]. HOMA-IR values of 2.6 or higher were used to indicate IR [11].

SPSS ver. 25.0 (IBM Corp.) was employed for data analysis. Data were expressed as median (interquartile range) for quantitative variables or frequency (percentage) for qualitative variables. The association of each variable with IR was assessed using the Mann-Whitney U test or the Pearson chi-square test, as appropriate. A receiver operating characteristic (ROC) curve was constructed to evaluate all adiposity indices as markers of IR. The AUCs of the indices were compared using an online MedCalc calculator (MedCalc Software Ltd.) [12]. Index performance was classified based on AUC values, with 0.8 or higher deemed acceptable [13]. Binary logistic regression was applied, adjusted for total testosterone (TT), for each adiposity index based on the established cutoffs to identify any predictive relationship with IR. A two-sided p-value of less than 0.05 was considered to indicate statistical significance.

Results

Among 632 patients with PCOS, 69.1% exhibited IR. Overall, despite having comparable ages, mFG scores, and frequencies of irregular menstrual cycles, patients with IR presented with a poorer metabolic profile (regarding obesity, glycemic values, lipid profiles, etc.) than those with insulin sensitivity (IS). Although the TT level and frequency of polycystic ovarian morphology were higher in the IR group, the luteinizing hormone/follicle-stimulating hormone ratio was lower in these patients compared to those with IS (Table 1).

Table 1.

Baseline characteristics of study participants with insulin resistance (n=632)

Figure 1 presents the findings from the ROC curve analysis of various adiposity indices as markers of IR. VAI and the TyG index were identified as poor markers (AUC, 0.6 to 0.69), while the remaining indices were considered fair (AUC, 0.7 to 0.79), in predicting IR among women with PCOS. Cutoff points, presented with their corresponding sensitivities and specificities, were determined using the highest Youden index. When the study population was categorized based on a BMI threshold of 23 kg/m² to distinguish between normal weight and overweight/obesity, the performance of the indices deteriorated, with all AUCs considered poor (0.6 to 0.7; data not shown).

Figure 1.

Receiver operating characteristic (ROC) curve analysis of various adiposity indices as markers of insulin resistance (n=632). BMI, body mass index; WC, waist circumference; VAI, visceral adiposity index; LAP, lipid accumulation product; TyG, triglyceride-glucose; AUC, area under the curve; CI, confidence interval; SE, standard error.

Table 2 presents comparisons of pairs of adiposity indices by AUC and standard error. VAI underperformed all other adiposity indices except the TyG index. Furthermore, the TyG index was outperformed by TyG-BMI and TyG-WC. However, the AUCs of TyG-BMI and TyG-WC did not differ significantly from those of BMI and WC.

Table 2.

Comparison of AUCs between pairs of adiposity indices (n=632)

Even after adjustment for TT, all adiposity indices demonstrated significant associations with the odds of IR among women with PCOS (Figure 2).

Figure 2.

Individual predictive associations of adiposity indices with insulin resistance among women with polycystic ovary syndrome (n=632). Binary logistic regression analysis was conducted with insulin resistance as the dependent variable, adjusted for total testosterone levels. BMI, body mass index; WC, waist circumference; VAI, visceral adiposity index; LAP, lipid accumulation product; TyG, triglyceride-glucose; OR, odds ratio; CI, confidence interval.

Discussion

The present study revealed significant relationships between all examined adiposity indices and IR in women with PCOS. VAI and the TyG index were identified as poor indicators of IR, whereas BMI, WC, LAP, TyG-BMI, and TyG-WC were found to be good markers. Both TyG-BMI and TyG-WC outperformed VAI and the TyG index. Furthermore, VAI was an inferior marker relative to BMI, WC, and LAP.

In a 2024 report, Feng et al. [14] demonstrated positive correlations between HOMA-IR and BMI, WC, VAI, and LAP in a study of 140 Chinese women aged 18 to 44 years with PCOS, aligning with our findings. Another study involving 114 Chinese women revealed a high correlation between HOMA-IR and both the TyG and TyG-BMI indices [15]. Among 180 Korean women with PCOS, aged 16 to 41 years, VAI was independently associated with IR after adjusting for the effects of age, glucose, and testosterone levels [16]. Similarly, a strong association between the TyG index and HOMA-IR was observed in Indonesian women with PCOS [17].

Our previous research indicated that BMI, WC, and LAP were moderate discriminators of IR in women with PCOS who had a normal BMI (<23 kg/m²), using a HOMA-IR cutoff of 2.3 [18]. In contrast, Huang et al. [19] found in 2019 that in normal-weight women with PCOS (<24 kg/m²), VAI and LAP were fair discriminators, while BMI and WC were poor. In patients with overweight/obesity, all four indices were poor discriminators of IR (HOMA-IR >2.77) [19]. The cutoffs for HOMA-IR and BMI varied across these studies, and no adiposity index was excellent at diagnosing IR. As hyperandrogenemia may confound the relationship between these indices and IR, incorporating androgen levels into the adiposity indices could improve their AUCs. This prospect warrants further investigation [18].

In this study, VAI and the TyG index were poor discriminators of IR in women with PCOS. This aligns with findings from a study of 305 Iranian women with PCOS, in which the TyG index exhibited an AUC of 0.639 [20]. In 2009, Wiltgen et al. [21] demonstrated that LAP was a superior marker to BMI and WC in a cohort of 51 Brazilian women with PCOS, aged between 14 and 35 years. An Argentinian study also reported that both VAI and LAP were good markers of IR, defined as HOMA-IR greater than 2.3 [22]. Among 40 Turkish women with PCOS, using a HOMA-IR criterion of ≥2.7, BMI, WC, and LAP were considered good predictors of IR based on their AUCs, while VAI was deemed fair [23]. Studies including a higher proportion of participants with obesity may have exhibited lower predictive power, and ethnic variations are also possible.

We evaluated the AUCs of various adiposity indices and found that TyG-BMI and TyG-WC exhibited the highest values. However, no significant difference was observed between the AUCs of these indices and those of classic anthropometric measurements such as BMI and WC. Although most studies have not employed statistical analysis to compare these indices, some research suggests that the TyG-BMI index has the highest AUC [22]. In previous work, we demonstrated that WC had a higher AUC than BMI, VAI, and LAP for IR among women with lean PCOS (BMI <23 kg/m²) [24]. Thus, the utility of novel adiposity indices may be limited among Bangladeshi women with PCOS relative to traditional measures. Factors such as ethnic variation, inconsistent HOMA-IR cutoffs, analyses across various age groups, and the inclusion of a larger number of women with obesity in PCOS populations may contribute to the differences in these adiposity indices in diverse groups. Limitations of this study include its reliance on data collected from a single tertiary care center and the use of HOMA-IR instead of the HIEGC method.

Compared to VAI and the TyG index, the TyG-BMI and TyG-WC indices demonstrated stronger correlations and superior predictive performance in relation to IR. However, they cannot be recommended as surrogate markers of IR among women with PCOS. Additionally, their diagnostic utility is comparable to that of traditional anthropometric measurements, specifically BMI and WC.

Article information Continued