Overall, 228 Chinese women of matching age were enrolled in this study at Shengjing Hospital of China Medical University, comprising 89 patients with PCOS and 139 healthy women (control group). Blood samples were collected in the morning, after overnight fasting, and between the third and fifth days of spontaneous menstrual bleeding. The diagnosis of PCOS was made as per the Rotterdam standards (two out of three):1. oligo‐ and/or anovulation 2. clinical and/or biochemical signs of hyperandrogenism 3. polycystic ovaries (15). The control group consisted of women with tubal infertility or male-partner infertility. Patients who were pregnant or lactating, on hormone medication and medications (corticosteroids, oral contraceptives, aspirin, insulin-sensitizing drugs, nicotinic acid, anti-androgens, statins) within 6 months, and with endocrine diseases such as androgen-secreting tumor, diabetes, Cushing’s syndrome, congenital adrenal hyperplasia, and hyperprolactinemia, were excluded. The full list of exclusion criteria is the same as that specified in our previous articles (5, 6).

To measure the leptin levels in patients with different types of PCOS, the control group and PCOS patients in our cohort were subdivided into the normal-fasting serum insulin (FSI) and hyper-FSI groups. This subdivision was chosen because it has been proposed in recent literature that the FSI level can be regarded as a precise alternative predictor of insulin resistance in PCOS patients, with an accuracy similar to fasting plasma glucose (FPG)/FSI and the homeostatic model assessment of insulin resistance (HOMA-IR) values (16). The reference ranges of the FSI levels were determined at Shengjing Hospital of China Medical University. The diagnosis of biochemical hyper-FSI was made when the FSI was ≥ 11.8 mIU/L. The characteristics of each FSI subgroup are shown in Table 1 .

To illustrate the relationship between leptin and androgens, participants in the PCOS group were further divided into the normal-androgenic and hyperandrogenic groups according to serum androgen levels. Total testosterone > 0.68 ng/mL, free testosterone > 0.040 nmol/L, and/or dehydroepiandrosterone sulfate (DHEAS) > 7025.11 nmol/L were used as the diagnostic criterion for biochemical hyperandrogenism. These parameters represented the 95th percentile of basal serum androgens in 139 control participants without PCOS (17). The characteristics of each androgenic subgroup are shown in Table 2 .

To analyze associations between leptin and obesity in PCOS, the control group and PCOS patients were each analyzed as lean and overweight/obese subgroups. Body mass index (BMI) ≥ 23 kg/m² of female is diagnostic for overweightness/obesity in Asians (18). Characteristics of each subgroup are shown in Supplementary Table 2 .

The concentration of human leptin in each sample was measured using the Human Leptin ELISA Kit (CSB-E04649h, Cusabio Biotech, Wuhan, China) with a quantitative sandwich-based enzyme immunoassay technique. Each serum sample was diluted 1:5 with the sample diluent before the assay was performed, in accordance with the standard procedure. The detection range was between 0.156 ng/mL–10 ng/mL, and the intra- and inter-assay precision values were less than 8% and 10%, respectively. The sample concentrations were determined by comparison with a standard curve at 450 nm optical density.

The serum concentrations of free testosterone (CSB-E05096h, Cusabio Biotech, Wuhan, China) and DHEAS (CSB-E05105h, Cusabio Biotech, Wuhan, China) were determined using commercial enzyme-linked immunosorbent assay kits, according to the manufacturers’ standard protocols. The intra-assay coefficients of variation for free testosterone and DHEAS were 6.8% and 5.5%, respectively, and the inter-assay coefficients of variation were 10.2% and 8.3%, respectively. The serum FSI levels were measured using the Architect Insulin Reagent Kit (8K41, Abbott Laboratories, IL, USA). FPG and the lipid profiles, including total cholesterol, triglycerides, high- and low-density lipoprotein cholesterol, were detected using enzymatic methods. Total testosterone, luteinizing hormone (LH), anti-Müllerian hormone (AMH), follicle-stimulating hormone (FSH), estradiol, prolactin, thyroid-stimulating hormone, and progestin were measured using chemiluminescence analysis.

All the data were analyzed using SPSS version 22 (IBM Corp., Armonk, NY). The normality of the continuous variable distribution was measured using the Kolmogorov–Smirnov test, and variables with a normal distribution were represented by means ± standard deviations. Student’s t test was applied to ascertain the differences of the means. The Mann–Whitney U test was conducted for continuous variables that did not conform to a normal distribution, and the data are expressed as medians (interquartile ranges). Spearman’s rank correlation coefficient was used to determine the correlation intensity between the pairs of variables, and the relationship between the clinical and metabolic variables and PCOS was analyzed using a multiple logistic regression model. The validity of leptin toward the diagnosis of PCOS, or leptin combined with total testosterone, DHEAS, free testosterone and AMH was estimated by receiver operating characteristic (ROC) curves. The area under the ROC curve (AUC) with specificity, sensitivity, negative predictive value (NPV), and positive predictive value (PPV) and 95% confidence interval (CI) were computed in order to analyze the diagnostic efficacy of the various indexes for PCOS. The optimal cutoff value was determined by calculating the Youden index. All the tests were bilateral, and statistical significance was defined as P < 0.05.

The serum leptin levels of the overall PCOS group were significantly higher than that of the control group ( Supplementary Table 1 ).

To systematically analyze the association between serum leptin levels and PCOS in the context of FSI, subjects in each of the study groups, PCOS and control, were categorized into normal-FSI or hyper-FSI subgroups. In both the normal-FSI and hyper-FSI subgroups, the serum leptin levels in PCOS patients were markedly higher than those in the control participants. In the hyper-FSI subgroup, the serum leptin levels in the PCOS patients were also significantly higher than those in the normal-FSI subgroup (P < 0.001). Moreover, PCOS patients in the hyper-FSI subgroup had the highest leptin levels compared to the other subgroups. The clinical and biochemical characteristics of the subgroups are presented in Figure 1A and Table 1 .

The PCOS group was divided into the normal-androgenic and hyperandrogenic subgroups. As shown in Figure 1B and Table 2 , PCOS patients in both normal-androgenic and hyperandrogenic subgroups had higher serum leptin levels than the control participants. Additionally, PCOS patients in the hyperandrogenic group had higher serum leptin levels than those in the normal-androgenic group.

As shown in Supplementary Table 2 , control and PCOS groups were divided into the lean and overweight/obese subgroups. Serum leptin levels were significantly increased in the overweight/obese group relative to the lean group in PCOS patients (P < 0.001). Moreover, serum leptin levels were significantly higher in PCOS patients than control subjects in the overweight/obese subgroup.

ROC curves were used to estimate the diagnostic performance of leptin, which was then compared and combined with the levels of total testosterone, DHEAS, free testosterone and AMH ( Figures 2A, F ). The AUC of leptin was 73.7%, with a cutoff value of 11.58 ng/mL, and sensitivity and specificity of PCOS prediction of 77.5% and 62.6%, respectively. The PPV and NPV were 57.0% and 81.3%, respectively. However, the combination of leptin and total testosterone resulted in a higher AUC of 81.8%, with poor sensitivity (55.1%) and excellent specificity (93.5%) for the prediction of PCOS ( Figures 2B, F ). The PPV and NPV were 84.5% and 76.5%, respectively. When leptin and free testosterone were combined, the AUC was 76.2%, with excellent diagnostic sensitivity (88.8%) and poor specificity (52.5%) for predicting PCOS ( Figures 2C, F ). The AUC of PCOS was predicted to be 77.5% for the combination of leptin and DHEAS, with an intermediate diagnostic sensitivity (70.8%) and a moderate diagnostic specificity (72.7%) ( Figures 2D, F ). Notably, the combination of leptin and AMH resulted in the highest AUC of 92.3%, with excellent sensitivity (93.3%) and moderate specificity (78.3%) for predicting PCOS. The PPV and NPV were 73.5% and 94.8%, respectively ( Figures 2E, F ). Moreover, the multivariate logistic regression revealed that an elevation of serum leptin level was correlated with a high risk of PCOS (P = 0.015; odds ratio, 1.128; 95% CI, 1.024–1.244; Figure 2G ).

Correlation analyses were used between leptin levels and the other parameters in PCOS women ( Supplementary Table 3 ). The serum leptin levels exhibited an association with FSI, FPG, HOMA-IR, BMI and total testosterone. There was a moderate and significant correlation between leptin and BMI (r = 0.695) and HOMA-IR (r = 0.415), whereas the correlation was weak between leptin and FSI (r = 0.383), FPG (r = 0.286), and total testosterone (r = 0.308).