Introduction

Polycystic ovary syndrome (PCOS) is the most common endocrine disorder in women of reproductive age1,2,3,4,5. Its most important features, hyperinsulinemia and insulin resistance, affect about 50% of PCOS patients1,2,3. Consequently, PCOS patients not only suffer from hypomenorrhea/amenorrhea, hyperandrogenism, and infertility, but may also develop overweight, obesity, hyperglycemia, hypertension, dyslipidemia, metabolic syndrome (MetS) or other cardiovascular diseases1,2,3. In fact, PCOS is considered an ovarian manifestation of MetS.

Metabolic syndrome (MetS) is defined as a cluster of metabolic disorders, including abdominal obesity, glucose intolerance or insulin resistance, dyslipidemia, and hypertension6,7. Current evidence demonstrates that MetS significantly impairs fertility and pregnancy outcomes in women of reproductive age6. Insulin resistance, the central pathogenesis of MetS, is thought to be a somatic feature of PCOS that further exacerbates visceral obesity and hyperandrogenemia1,8. In turn, proliferating visceral adipose tissue secretes inflammatory cytokines that worsen insulin resistance and hyperandrogenemia, creating a vicious cycle that makes PCOS and MetS mutually causative and exacerbating each other1,8. The interplay between hyperandrogenism, insulin resistance, and visceral obesity causes PCOS patients more susceptible to atherosclerosis by significantly impacting central obesity, hyperglycemia, hypertension, and dyslipidemia1,9.

Identifying predictors of PCOS with MetS in women of reproductive age is crucial for preventing and treating infertility and metabolic cardiovascular diseases. The PCOS population has been reported to be highly heterogeneous, with different PCOS patients having different metabolic risks1. Insulin resistance, hyperinsulinemia, and hormonal imbalance caused by obesity, particularly central or visceral obesity, are the common etiology and pathological basis of PCOS and MetS1,6. NC, WC, waist-to-hip ratio (WHR), and other anthropometric indicators have been used as parameters for visceral obesity and demonstrate strong associations with PCOS and MetS1,7,8. Recently, some studies have shown that PCOS were correlated with anthropometric indicators such as NC6,10 lipid-related composite indicators such as apolipoprotein A/B3,11, and compound glycolipid index such as triglyceride index11,12and some composite indexes combining anthropometry with biochemistry such as lipid accumulation product (LAP), visceral fat index (VAI), cardiometabolic index (CMI) and Chinese visceral fat Index (CVAI)2,13,14 but few studies have demonstrated the superiority or inferiority of these markers. Therefore, it is clinically important to find a simple and reliable index for early identification of central obesity to predict and prevent PCOS and MetS as early as possible.

In our previous study, both NC and TyG index could effectively predict the severity of coronary lesions in patients with MetS and acute coronary syndrome. In particular, NC, a simple time-saving anthropometric measurement, was identified as a marker of subcutaneous adipose tissue distribution in the upper body and a promising predictor of cardio-metabolic syndrome8. However, there have been no reports on the predictive value of NC versus TyG index for PCOS + MetS. In this study, a retrospective study was conducted to investigate the clinically relevant factors affecting PCOS with MetS and to find effective, simple, practical predictors of these patients.

Methods

Study population

This retrospective cohort study included 517 women (aged 18–42) from the Second Hospital of Shandong University and Qilu Hospital of Shandong University during January 2017 to January 2024. All participants underwent anthropometric, blood pressure, and biochemical assessments within 3 days before or during hospitalization.

The criteria for inclusion and exclusion of participants are described below. Inclusion criteria: (1) Age at first visit ≥ 18 years old; (2) Confirmed PCOS diagnosis without short-acting contraceptive treatment; (3) The diagnostic criteria for PCOS were developed jointly by the European Society of Human Reproduction and Embryology and the American Society for Reproductive Medicine (ESHRE/ASRM) at the Rotterdam Conference in 20032,14,15: (3a) Anomalous ovulation: infrequent ovulation or anovulation; (3b) Hyperandrogenemia and/or clinical manifestations of Kaohsiung (such as hirsutism and acne); (3c) B-mode ultrasonography indicated a polycystic ovary (≥ 12 follicles with a diameter of 2–9 mm in one or both ovaries) and/or increased ovarian volume (unilateral ovarian volume > 10 cm³); (3d) Two of the above three need to be met, while ruling out other causes of hyperandrogenism; (4) Voluntarily accepted blood sampling for testing related laboratory indices, and read and signed the informed consent form. Exclusion criteria11,15: (1) Estradiol and progesterone, drugs that regulate blood lipid and blood sugar, and other drugs that have an impact on laboratory indices should be used in the first half year of treatment; (2) Cushing syndrome, adrenal cortical hyperplasia, androgen-secreting tumors and other diseases that lead to hyperandrogenism; (3) Other diseases that cause ovulation disorders, such as premature ovarian insufficiency (POI), premature ovarian failure (POF), hyperprolactinemia, hypothalamic/pituitary amenorrhea and abnormal thyroid function; (4) Other hormonal tumors, cardiovascular diseases, endocrine diseases, and malignant tumors.

This study was approved by and conducted with the approval of Ethical Committee of the Second Hospital of Shandong University, and the Qilu Hospital of Shandong University. All methods were performed in accordance with the relevant guidelines and regulations. It conformed to the provisions of the Declaration of Helsinki. We confirm that informed consent was obtained from all participants and/or their legal guardians.

Anthropometric and blood pressure measurement

Anthropometric parameters such as height, body weight, WC and NC were measured by trained technicians. BMI was calculated from weight and height. For NC, participants sat upright while measurements were taken at the narrowest neck point (below the laryngeal prominence). Patients with conditions affecting NC (e.g., goiter, spinal deformities, Cushing’s syndrome) were excluded. NC values were categorized into quartiles (Q1–Q4). After a rest period of at least 5 min, blood pressure (BP) was assessed using a validated standard aneroid sphygmomanometer via the auscultatory method.

Definition of metabolic syndrome

MetS was defined based on the ATP III/AHA/NHLBI criteria16. A diagnosis required at least three of the following: (1) WC ≥ 80 cm for Asian women; (2) BP ≥ 130/85 mmHg; (3) Fasting plasma glucose ≥ 5.55 mmol/L; (4) Plasma high-density lipoprotein cholesterol level (HDL-C) < 1.29 mmol/L for women; and (5) Plasma triglyceride level ≥ 1.7 mmol/L.

Calculation of composite metabolic indices

To calculate some indexes combining the biochemical and anthropometric indicators, use the following formula15,17: BMI = body weight [kg]/height [m2, LAP = (WC[cm] -58)×TG [mmol/L], TyG indexes = Ln [TG (mg/dl) × FBG (mg/dl)/2], VAI = (WC [cm] / [36.58 + 1.89 × BMI]) × (TG [mmol/L] / 0.81) × (1.52/HDL [mmol/L]), CVAI = − 187.32 + 1.71 × age + 4.23 × BMI + 1.12 ×WC (cm) + 39.76 × log10 [TG (mmol/L)] – 11.66 × HDL (mmol/L).

Statistical analysis

SPSS software 26.0 (SPSS, Inc., Chicago, IL, USA) was applied for statistical analysis. Continuous variables are presented as mean ± SD (normally distributed) or median (non-normal), compared using t-tests (two groups) or ANOVA (≥ 3 groups). Non-parametric data were analyzed using the Mann-Whitney U-test. Categorical data (%) were compared via chi-square tests. Logistic regression was chosen over Poisson regression because our case-control design with a binary outcome (PCOS + MetS presence/absence) inherently requires odds ratio. ROC analysis determined optimal cut-offs (maximizing Youden’s index: sensitivity + specificity − 1) for NC/TyG indices. AUC comparisons used DeLong’s method (Bonferroni-adjusted). P < 0.05 indicated significance.

Results

Subject baseline characteristics based on PCOS/MetS

Participants were divided into four groups according to the quartiles of NCs(NC1-NC4). Except for age, significant group differences (all P < 0.05) were observed in: hemodynamic index (SBP, DBP, PPD, and HR), anthropometric index (body weight, BMI, WC, and NC), gonadal hormones (LH, FSH, PRL, E2, TE, P4, and AMH), WBC, Hcy, liver and kidney function (ALT, AST, BUN, Cr, GFR, and Cystatin C), glycolipid metabolism [FBG, HbA1c, TC, TG, LDL-C, HDL-C, Apo A, apo B, and Lp (a)], TyG indexes, and composite indexes (combined anthropometric and biochemical indicators) (LAP, VAI, and CVAI) (Table 1). MetS prevalence among participants gradually increased with the increase of NC (Table 1). Participants were categorized into three groups: normal control, PCOS + non MetS group, and PCOS + MetS group (Table 2). ANOVA results demonstrated significant differences in all metrics examined above (all P < 0.05 except age) across these groups (Table 2), which was similar to the NC quartile findings in Table 1.

Table 1 Baseline characteristics of subject data based on NC grouping.
Table 2 Baseline characteristics of subject data based on PCOS and MetS subgroups.

Potential indicators of PCOS complicated with MetS

As shown in Table 3, univariate logistic regression analysis demonstrated that except age, other non-composite indicators (similar to those in Tables 1 and 2) were related to PCOS with MetS. However, multi-factorial logistic regression analysis showed that after controlling for age, relevant anthropometric indices, blood pressure, heart rate, glucose-lipid metabolism, or gonadal hormone indices, SBP, DBP, NC, WC, LH, PRL, TE, UA, FBG, HbA1c, TG, LDL-C, and Lp(a) were positively correlated with PCOS + MetS, on the contrary, FSH, E2, P4, AMH, HDL-C, and ApoA were significantly and negatively associated with PCOS + MetS (Fig. 1).

Table 3 Regression analysis of potential dangers affecting PCOS with MetS.
Fig. 1
figure 1

Multivariate analysis of independent association between noncomplex variables and PCOS + MetS. Abbreviations as in Table 3.

Comparison of NC and composite indicators for PCOS with MetS prediction

As demonstrated in Table 4, univariate Logistic analysis exhibited that VAI, LAP, CVAI, ApoB/ApoA and LH/FSH, which are compound indicators, particularly VAI, ApoB/ ApoA, and LH/FSH, had a close correlation with PCOS or MetS. However, after adjusting for relevant confounders in multivariate analysis, most of these associations lost statistical significance (P > 0.05) (Table 4). In contrast, NC and TyG index demonstrated statistically significant differences with strong predictive value and stability both in correlation with PCOS and PCOS + MetS (all P < 0.05) (Table 4). In detail, compared with the normal control group, the adjusted OR of PCOS predicted by NC and TyG were 1.842 (95% CI1.056, 3.215) and 2.804 (95% CI1.082, 5.251), respectively (all P < 0.05). The adjusted OR of PCOS + MetS predicted by NC and TyG were3.516 (95% CI 1.910,6.472) and 3.386 (95% CI 2.013, 6.208), respectively (all P < 0.05) (Table 4; Figs. 2 and 3). Although LAP, VAI, CVAI, LH/FSH, and apoB/apoA had high OR values in predicting PCOS and PCOS + MetS, significantly statistical differences had not achieved, except for LH/FSH predictions for PCOS + MetS (Table 4).

Table 4 Comparison of the predictive ability of NC and other composite variables for PCOS or/and MetS. Abbreviations as in Table 3.
Fig. 2
figure 2

Forest plot demonstrated the comparison of NC’s ability to predict PCOS with other composite variables. Abbreviations as in Table 4.

Fig. 3
figure 3

Forest plot demonstrated the comparison of NC’s ability to predict PCOS + MetS with other composite variables. Abbreviations as in Table 4.

Moreover, the predictive values of NC and TyG index for PCOS and/or MetS were further confirmed by ROC curves (Fig. 4). For NC, the AUC values were 0.909 (PCOS) and 0.911 (PCOS + MetS), with optimal cutoff points at 34.6 cm (Youden index 0.859) and 35.9 cm (Youden index 0.668), respectively. These cutoffs demonstrated high sensitivity (88.6% for PCOS, 84.8% for PCOS + MetS) while maintaining good specificity (73% and 72%) (Fig. 4). TyG index showed predictive values for PCOS (AUC 0.942, cutoff 8.92) with 82.8% sensitivity and 53.0% specificity, and maintained good performance for PCOS + MetS (AUC 0.857, cutoff 9.04) with 89.6% sensitivity and 65.5% specificity (Fig. 4).

Fig. 4
figure 4

ROC curve of PCOS/MetS assessed by NC and TyG index. ROC curve, receiver operating characteristic curve; Other abbreviations as in Table 5.

Table 5 ROC curve analysis of NC and TyG index predicting pcos/mets. Abbreviations as in Table 3.

Discussion

Our findings establish NC and the TyG index as clinically significant predictors of PCOS with MetS in women. Our findings align with prior studies2,8,18 showing positive correlations between NC and traditional obesity markers (body weight, BMI, WC) in univariate analysis (Table 1). These parameters were significantly higher in PCOS and PCOS + MetS groups versus controls (Table 2). However, multivariate analysis revealed only NC maintained independent predictive value for PCOS + MetS (Table 3; Fig. 1), consistent with reports by Daghestani and He et al.6,9. Although Pan X et al. believe that blood pressure (SBP and DBP) is not related to PCOS patients19Meta analysis demonstrates that women of reproductive age with PCOS are closely associated with SBP and DBP, independent of obesity20. This study confirmed that blood pressure was an independent indicators for PCOS + MetS (Table 3; Fig. 1).

Our study confirms the well-documented metabolic disturbances characteristic of PCOS, with lipid metabolism disorders alongside impaired glucose tolerance and hyperuricemia21,22,23,24. As NC increased, we observed progressive elevations in TC, TG, LDL-C, Lp(a), and ApoB, accompanied by decreasing HDL-C and ApoA (Table 1). While these abnormalities persisted regardless of MetS comorbidity (Table 2), multivariate analysis revealed that only TG, LDL-C, HDL-C, Lp(a), and ApoA in the lipid profile showed independent association with PCOS + MetS (Table 3; Fig. 1). The glucose metabolism assessment revealed HbA1c was more closely correlated with PCOS + MetS than FBG [2.240 (95%CI, 1.144, 3.929) vs. 2.116 (95%CI, 1.449, 3.090), all p < 0.05] (Table 3; Fig. 1), despite its exclusion from standard MetS criteria. This study also demonstrated that with the increase of NC, the UA levels of subjects increased, and multivariate analysis showed that UA was independently associated with PCOS + MetS (Table 3; Fig. 1). Hyperuricemia in PCOS forms a vicious cycle with insulin resistance through androgen-mediated URAT1 activation and visceral fat-driven purine metabolism24.

PCOS involves complex endocrine-metabolic disturbances1 particularly elevated LH/FSH ratios and hyperandrogenism, which impair follicular development and exacerbate insulin resistance25,26,27. In agreement with Ebrahimi-Mamaghani M, et al.26 our study demonstrated that these hormonal imbalances correlate with worsening metabolic profiles as NC increases (Tables 1, 2 and 3; Fig. 1). Both univariate and multivariate analyses in our study identified significant changes in gonadal hormone levels in patients with PCOS with MetS as compared to the control group: significant increase in LH, PRL, and TE, and marked reduction in FSH, E2, P4, and AMH (Tables 1, 2 and 3; Fig. 1).

Women with PCOS exhibited more visceral fat accumulation regardless of body weight14. While traditional measures (BMI, WC) poorly assess central obesity, composite indices (VAI, CVAI, LAP) combining anthropometric and metabolic parameters have been proposed for better prediction28. However, in this study, multivariate analysis revealed that VAI, CVAI and LAP in PCOS patients of childbearing age failed to achieve the ideal prediction value, possibly due to limited age variation (Table 4; Figs. 2 and 3). On the contrary, both NC and TyG exhibited good and stable predictive value for both PCOS and PCOS + MetS (Table 4; Figs. 2 and 3). ROC curve analysis further confirmed that NC and TyG index had better predictive value for PCOS + MetS than other composite indexes including LH/FSH and Apo B/Apo A (Table 5; Fig. 4).

Cut-offs for NC (≥ 35.9 cm) and TyG (≥ 9.04) were derived from Youden index in our cohort. Hooman Ebrahim et al. found cut-off for NC ≥ 36 cm in Iran MetS cohorts29 which align with our results. However, Kamrul-Hasan, A B M et al. identified that NC ≥ 34.25 cm was the best cutoff value for MetS in the women with PCOS30. Therefore, external validation of different medical institutions and ethnic populations is required before widespread clinical application. Early identification of MetS risk via NC/TyG could prompt targeted interventions (e.g., lifestyle modification, dieting and exercise, and metformin), potentially mitigating long-term cardiometabolic morbidity and improving fertility outcomes in PCOS. NC’s superiority lies in its practicality: it requires no blood sampling, is cost-free, and unaffected by fasting status, breathing, body position or laboratory variability. Unlike TyG (which reflects insulin resistance), NC directly captures upper-body adiposity, a key driver of metabolic dysfunction in PCOS. This makes NC particularly valuable in resource-limited settings. Therefore, NC for predicting PCOS + MetS might be more suitable for clinical operation and wide application. These finding highlight the particular value of NC and TyG for risk stratification.

Conclusion

To the best of our knowledge, this is the first study comparing the predictive value of anthropometric indicators with the composite indicators of somatometry and glycolipid biochemistry for PCOS + MetS reproductive-aged women. The results show that NC and TyG are not inferior to VAI, CVAI and LAP for predicting PCOS + MetS, and compared to TyG, NC is more convenient, intuitive, economical, and practical regardless of dietary status.

Limitations

Firstly, the different diagnostic criteria for MetS defined by different organizations may have influenced the statistical results of this study. Secondly, while we adjusted for key covariates (age, BMI, medication use) in multivariate models, unmeasured confounders, such as diet, physical activity, and socioeconomic factors, may influence results. Future prospective studies should incorporate standardized lifestyle assessments. Thirdly, the retrospective design of our study may introduce documentation bias and limits causal inference. Due to the retrospective nature of the study, subjects such as serum insulin concentration, hip circumference and its correlation: body adiposity index (BAI), were not examined at that time or the data were missing, so this study failed to include these statistics, which may affect the persuasiveness of insulin resistance and anthropometry. Fourthly, our cohort’s homogeneity (Chinese, reproductive-age) limits extrapolation to other populations. The recruitment of participants exclusively from two tertiary academic hospitals may have further skewed our sample toward more severe PCOS phenotypes and urban, medically insured populations, which may not fully represent the broader population. Additionally, a significant limitation is the potential confounding effect of medications used by PCOS patients, such as insulin-sensitizing agents, oral contraceptives, melatonin39, et al., which could alter metabolic and hormonal biomarkers and potentially biasing our results. Future studies should prospectively document all medications and their durations to isolate their effects on metabolic parameters.