Insulin resistance as a determinant of fertilization efficiency in polycystic ovary syndrome patients undergoing IVF/ICSI: a retrospective cohort study.

This retrospective cohort study was conducted in compliance with the Declaration of Helsinki and received ethical approval from the Institutional Review Board of Shenzhen Zhongshan Obstetrics & Gynecology Hospital (Approval No. SZZSECHU-F-2025040). The analysis included 1,768 PCOS patients who underwent IVF/ICSI cycles at the Reproductive Medicine Center of Shenzhen Zhongshan Obstetrics & Gynecology Hospital between October 2010 and November 2024. Informed consent was waived due to the anonymized nature of retrospective data, in compliance with national ethical guidelines for secondary use of clinical records. The inclusion criteria were as follows: (1) women meeting the diagnostic criteria of the 2003 Rotterdam criterion guidelines of reproductive age; (2) women without taking any insulin-sensitizing medications or hormonal therapies for ≥ 3 months prior to ovarian stimulation; (3) women who underwent an insulin release test and glucose assessment. The exclusion criteria were (1) women lacking of FPG data or FINS measurements; (2) women with structural or functional uterine abnormalities including abnormal uterine bleeding, primary amenorrhea, hypothalamic amenorrhea, pituitary amenorrhea, and uterine amenorrhea; (3)women affecting immune homeostasis like autoimmune disease, malignant tumors, functional ovarian tumors (Fig. 1 ). Fig. 1 Diagram of the study cohort selection and distribution of the PCOS patients investigated Diagram of the study cohort selection and distribution of the PCOS patients investigated Ovarian stimulation was performed using either a GnRH antagonist or agonist protocol, with protocol selection permitted based on clinical indications. On days 2–3 of the menstrual cycle, a transvaginal ultrasound assessment was conducted to confirm adequate endometrial shedding and normal serum hormonal levels, with antral follicle diameter ≤ 7 mm. Gonadotropins (Gn) (Merck) were initially administrated and the dosage of Gn is adjusted based on follicular development and hormonal assays. Administration of a GnRH antagonist at a dose of 0.25 mg daily commenced once the follicle diameter reached ≥ 12 mm or estradiol (E2) surpassed 300 pg/ml. When the number of dominant follicles (≥ 18 mm in diameter) was ≥ 2 and the serum E2 level achieved an average of 200 pg/ml per dominant follicle, human chorionic gonadotropin (hCG) at a dose of 10,000 IU was administered to trigger ovulation. Oocyte retrieval was performed 34–36 h post-trigger under ultrasound guidance. Embryonic development was continuously monitored post-fertilization. By 72 h (Day 3), embryos underwent a critical assessment using the standard morphology criteria, wherein Grade I and II embryos were categorized as high-quality. A subset or all embryos were cultured until days 5 to 7 to observe blastocyst formation. Blastocysts were subsequently graded according to the Gardner scoring system, which was classified into three categories: Grade 1 (AA, AB, BA, BB), Grade 2 (AC, BC, CA, CB), and Grade 3 (CC). Grade 1 blastocysts were defined as high-quality, while Grade 1 and 2 were considered suitable for utilization. The best morphologically graded embryos were chosen for fresh or frozen embryo transfer (ET). All patients routinely underwent luteal phase support, flexibly adjusted based on the clinician’s experience. Two weeks after ET, pregnancy was assessed by serum β-hCG levels and confirmed by transvaginal ultrasound 4 weeks after ET. Serum β-hCG levels > 50 IU/L were regarded as biochemical pregnancy and the presence of the gestational sac was regarded as clinical pregnancy. If an intrauterine pregnancy was established but abortion occurred before the 12th week of gestation, it was diagnosed as an early spontaneous miscarriage. Blood samples were collected under standardized fasting conditions from patients before the initiation of ovarian stimulation. FINS and FPG levels were measured using an automated biochemical analyzer. Derived from a large-scale Chinese clinical study encompassing participants aged 25 to 74 years [ 28 ], IR was defined as the Homeostasis Model Assessment of Insulin Resistance (HOMA) index ≥ 2.69, calculated using the following formula: \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text{HOMA}=\frac{FPG*FINS}{22.5}$$\end{document} Fertilization rate denotes the proportion of oocytes that undergo successful fertilization. For conventional IVF, Fertilization rate = (Number of Fertilized Oocytes/Total Oocytes Retrieved) × 100%. For ICSI procedure, Fertilization rate = (Number of Fertilized Oocytes/Number of Metaphase II Oocytes) × 100%. 2PN rate, conventionally termed the normal fertilization rate, quantifies the proportion of fertilized oocytes exhibiting two distinct pronuclei. This parameter is calculated identically for both IVF and ICSI methodologies: Normal Fertilization Rate (2PN Rate) = (Number of 2PN Oocytes/Total Number of Fertilized Oocytes) × 100%. Data normality was assessed using the Kolmogorov–Smirnov test, with Q-Q plots and Shapiro–Wilk tests applied for nonparametric confirmation. Continuous variables with normal distributions were expressed as mean ± standard deviation (SD) and compared using Student’s t-tests, while non-normally distributed variables were reported as median (interquartile range, IQR) and analyzed via Mann–Whitney U tests. Univariate logistic regression identified candidate predictors (p < 0.10) for inclusion in multivariable modeling, adjusting for biologically plausible confounders: age, BMI infertility duration, antral follicle count, and baseline FSH. Effect estimates were reported as adjusted odds ratios (OR) with 95% confidence intervals (CI). All analyses were conducted using SPSS v23.0 (SPSS Inc., Chicago, IL), with two-tailed p < 0.05 considered statistically significant. A power analysis was conducted using PASS v20.0.1 (NCSS Inc., LLC), setting by two-sided for an alternative hypothesis and t-test for a test type.

A total of 1768 PCOS women who undergoing IVF/ICSI cycles were recruited after exclusion criteria. Participants were stratified into two groups based on the established diagnostic threshold for insulin resistance (HOMA index ≥ 2.69): 901 individuals meeting this criterion were classified as IR group, while the remaining 867 participants constituted the non-IR group. The demographic and clinical characteristics of the two cohorts were presented in Table  1 . Table 1 Baseline characteristics between the IR and non-IR groups in PCOS patients non-IR group IR group p value N  = 867 N  = 901 Age (years) 30.72 ± 3.86 30.77 ± 3.89 0.770 BMI (kg/m 2 ) 21.59 ± 3.20 25.44 ± 3.55  < 0.001* Infertility duration (years) 3.25 ± 2.43 3.74 ± 2.75  < 0.001* infertility type 0.496  Primary 531 (61.25%) 566 (62.82%)  Secondary 336 (38.75%) 335 (37.18%)  AFC (n) 25.05 ± 9.79 26.74 ± 10.74  < 0.001*  Basal AMH (ng/mL) 8.44 ± 4.48 8.21 ± 4.23 0.260  Basal FSH (mIU/mL) 10.64 ± 3.83 9.78 ± 3.25  < 0.001*  FPG (mmol/L) 5.21 ± 0.46 5.65 ± 0.84  < 0.001*  FINS (μU/mL) 7.71 ± 2.43 21.21 ± 9.22  < 0.001*  HOMA 1.79 ± 0.57 5.34 ± 2.66  < 0.001* # p  < 0.05, * p  < 0.01 BMI body mass index, AFC antral follicle counts, AMH Anti-Müllerian Hormone, FSH follicle-stimulating hormone, FPG fasting plasma glucose, FINS fasting insulin, HOMA  homeostasis model assessment of insulin resistance Values are numbers (percentages) of cases, mean ± standard deviation or median (interquartile range). Continuous variables: Independent t-test; categorical variables: Chi-square test Baseline characteristics between the IR and non-IR groups in PCOS patients # p  < 0.05, * p  < 0.01 BMI body mass index, AFC antral follicle counts, AMH Anti-Müllerian Hormone, FSH follicle-stimulating hormone, FPG fasting plasma glucose, FINS fasting insulin, HOMA  homeostasis model assessment of insulin resistance Values are numbers (percentages) of cases, mean ± standard deviation or median (interquartile range). Continuous variables: Independent t-test; categorical variables: Chi-square test Regarding clinical characteristics before the IVF/ICSI-ET procedure, statistically significant differences were observed between the two groups in terms of BMI, infertility duration and antral follicle count. Women with IR exhibited elevated BMI (25.44 ± 3.55 vs. 21.59 ± 3.20, p  < 0.001), prolonged infertility duration (3.74 ± 2.75 vs. 3.25 ± 2.43, p  < 0.001), higher AFC (26.74 ± 10.74 vs. 25.05 ± 9.79, p  < 0.001), and lower follicle-stimulating hormone (FSH) (9.78 ± 3.25 vs. 10.64 ± 3.83, p  < 0.001) levels compared to those without IR. As expected, individuals with IR phenotype also had significantly higher fasting plasma glucose (FPG) (5.65 ± 0.84 vs. 5.21 ± 0.46, p  < 0.001), fasting insulin (FINS) levels (21.21 ± 9.22 vs. 7.71 ± 2.43, p  < 0.001), and HOMA (5.34 ± 2.66 vs. 1.79 ± 0.57, p  < 0.001) levels than those without IR phenotype. The ovarian stimulation protocols, embryological characteristics and clinical outcomes of PCOS patients are systematically presented in Table  2 . The IR group required higher starting doses of Gn (2.16 ± 0.66 vs. 1.91 ± 0.57, p  < 0.001), prolonged Gn stimulation duration (10.32 ± 2.51 vs. 9.52 ± 1.97, p  < 0.001), and elevated cumulative Gn consumption (26.50 ± 11.09 vs. 20.49 ± 8.07, p  < 0.001) compared to the non-IR group. Additionally, the fertilization rate (82.02% vs.83.86%, p  = 0.005) and 2PN rate (81.07% vs.83.96%, p  < 0.001) in PCOS patients with IR were significantly lower than those in patients without IR. These findings collectively suggest that IR may negatively affect embryonic development, possibly through mechanisms involving dysregulated glucose metabolism or disrupted hormonal signaling pathways. Notably, despite these embryological impairments, no statistically significant differences were observed in pregnancy outcomes between groups, including biochemical pregnancy rates ( p  = 0.170), clinical pregnancy rates ( p  = 0.339), ongoing pregnancy rates ( p  = 0.336), and early miscarriage rates ( p  = 0.145). This discordance between suboptimal embryological parameters and preserved clinical pregnancy success suggests compensatory mechanisms—potentially involving endometrial adaptive responses or post-implantation metabolic buffering—may attenuate IR-related reproductive deficits. Further mechanistic studies are warranted to elucidate how systemic metabolic dysfunction interfaces with localized follicular and endometrial microenvironments to modulate ART outcomes in PCOS. Table 2 Comparison of COS protocols and clinical outcomes between the IR and non-IR groups in PCOS patients non-IR group IR group p value N  = 867 N  = 901 Ovarian Stimulation Parameters  Starting doses of Gn (IU) 1.91 ± 0.57 2.16 ± 0.66  < 0.001*  Days of Gn administration (days) 9.52 ± 1.97 10.32 ± 2.51  < 0.001*  Total Gn dosage (IU) 20.49 ± 8.07 26.50 ± 11.09  < 0.001*  Endometrial thickness (mm) on hCG day 9.22 ± 1.68 9.35 ± 1.76 0.136 Endometrium preparation 0.081  HRT cycle (%) 440 (50.75%) 481 (53.39%)  Natural ovulation cycle (%) 100 (11.53%) 68 (7.55%)  Downregulation cycle (%) 52 (6.00%) 53 (5.88%)  Stimulation cycle (%) 236 (27.22%) 255 (28.30%) Embryological Outcomes  Fertilization protocol 0.083 IVF (%) 638 (73.59%) 695 (77.14%) ICSI (%) 229 (26.41%) 206 (22.86%) Number of embryo transfer (n) 1.29 ± 0.45 1.26 ± 0.44 0.178 Embryo transfer 0.666  D3 (n) 110 (12.69%) 119 (13.21%)  D5 (n) 757 (87.31%) 782 (86.79%) Embryonic outcomes  MII rate (%) 87.77% 88.03% 0.651  Fertilization rate (%) 83.86% 82.02% 0.005*  2PN rate (%) 83.96% 81.07%  < 0.001*  Embryo formation rate (%) 58.87% 57.87% 0.315  Available blastocysts rate (%) 51.03% 49.59% 0.144  High-quality blastocysts rate (%) 25.90% 24.96% 0.303 Pregnancy outcomes  Biochemical pregnancy rate (%) 690/867 (79.58%) 692/900 (76.89%) 0.170  Clinical pregnancy rate (%) 591/865 (68.32%) 595/899 (66.18%) 0.339  Early miscarriage rate (%) 110/591 (18.61%) 128/595 (21.51%) 0.145  Ongoing pregnancy rate (%) 468/867 (53.98%) 456/893 (51.06%) 0.366 Live birth outcomes  Live birth rate (%) 408/867 (47.06%) 387/900 (43.00%) 0.083  Newborns (n) 1.08 ± 0.40 1.04 ± 0.37 0.090  Sex ratio (%, male/total) 237/408 (58.09%) 222/387 (57.36%) 0.836  Birth weight (g) 2442.08 ± 389.44 2266.45 ± 529.01 0.071 # p  < 0.05, * p  < 0.01   Gn gonadotropins, HRT hormone replacement therapy, IVF in vitro fertilization, ICSI  intracytoplasmic sperm injection Values are numbers (percentages) of cases, mean ± standard deviation or median (interquartile range) Continuous variables: Independent t-test; categorical variables: Chi-square test Comparison of COS protocols and clinical outcomes between the IR and non-IR groups in PCOS patients # p  < 0.05, * p  < 0.01 Gn gonadotropins, HRT hormone replacement therapy, IVF in vitro fertilization, ICSI  intracytoplasmic sperm injection Values are numbers (percentages) of cases, mean ± standard deviation or median (interquartile range) Continuous variables: Independent t-test; categorical variables: Chi-square test As shown in Table  2 , both the fertilization rate and 2PN rate were significantly lower in the IR group compared to the non-IR group. To thoroughly investigate the factors influencing fertilization rate and 2PN rate, a linear regression analysis was implemented with sequential variable inclusion. Variables exhibiting statistically significant differences in baseline characteristics (Table  1 and 2 ) were initially subjected to univariate logistic regression analysis to identify potential predictors of embryological outcomes. Subsequently, all significant variables ( p  < 0.05) from the univariate analysis were incorporated into a multivariate logistic regression model to adjust for confounding effects and evaluate independent associations (Table  3 ). Table 3 Linear logistic regression and interaction analysis of clinical parameters Outcomes Variables B (95% CI) p value Adjusted B (95% CI) p value Fertilization rate Infertility duration (years) −0.104 (−0.353, 0.145) 0.411 / / BMI (kg/m 2 ) −0.353(−0.519, −0.187)  < 0.001* −0.355 (−0.562, −0.149)  < 0.001* AFC (n) −0.103 (−0.167, −0.039) 0.002* −0.091 (−0.155, −0.127) 0.005* Basal FSH (mIU/mL) 0.172 (−0.010, 0.353) 0.063 / / Starting doses of Gn (IU) −0.594 (−1.623, 0.435) 0.258 / / Days of Gn (days) 0.367 (0.085, 0.649) 0.011# 0.612 (0.315, 0.910)  < 0.001* Total Gn dosage (IU) 0.035 (−0.029, 0.099) 0.282 / / IR or not −1.837 (−3.131, −0.542) 0.005* −0.664 (−2.551, 1.222) 0.490 FPG (mmol/L) −0.645 (−1.549, 0.259) 0.162 / / FINS (μU/mL) −0.093 (−0.161, −0.026) 0.007* −0.110 (−0.380, 0.160) 0.424 HOMA −0.332 (−0.557, −0.086) 0.008* −1.215 (−2.891, 0.460) 0.155 Interaction BMI*IR −0.079 (−0.129, −0.029) 0.002* 0.098 (−0.302, 0.497) 0.631 2PN rate Infertility duration (years) 0.209 (−0.051, 0.468) 0.115 / / BMI (kg/m 2 ) −0.159 (−0.333, 0.015) 0.074 / / AFC (n) −0.058 (−0.125, 0.008) 0.086 / / Basal FSH (mIU/mL) 0.314 (0.125, 0.504) 0.001* 0.197 (−0.007, 0.400) 0.058 Starting doses of Gn (IU) 1.303 (0.227, 2.379) 0.018# 1.338 (0.153, 2.522) 0.027# Days of Gn (days) −0.421 (−0.716, −0.126) 0.005* −0.277 (−0.581, 0.027) 0.074 Total Gn dosage (IU) −0.037 (−0.104, 0.030) 0.275 / / IR or not −2.889 (−4.240, −1.539)  < 0.001* −2.540 (−4.456, −0.624) 0.009* FPG (mmol/L) −0.788 (−1.733, 0.158) 0.102 / / FINS (μU/mL) −0.117 (−0.187, −0.046) 0.001* −0.042 (−0.317, 0.234) 0.767 HOMA −0.403 (−0.659, −0.147) 0.002* 0.076 (−0.887, 1.040) 0.876 Interaction BMI*IR −0.110 (−0.162, −0.058)  < 0.001* −0.093 (−0.179, −0.007) 0.033# # p  < 0.05, * p  < 0.01 Interaction analysis: Adjusted for infertility, BMI, AFC, basal FSH, starting of Gn, days of Gn, total Gn dosage, FPG, FINS BMI  body mass index, AFC  antral follicle counts, FSH  follicle-stimulating hormone, Gn  gonadotropins, FPG  fasting plasma glucose, FINS  fasting insulin Linear logistic regression and interaction analysis of clinical parameters # p  < 0.05, * p  < 0.01 Interaction analysis: Adjusted for infertility, BMI, AFC, basal FSH, starting of Gn, days of Gn, total Gn dosage, FPG, FINS BMI  body mass index, AFC  antral follicle counts, FSH  follicle-stimulating hormone, Gn  gonadotropins, FPG  fasting plasma glucose, FINS  fasting insulin Multivariable regression analysis demonstrated no significant association between the IR-associated index and fertilization rate. Conversely, IR exhibited a negative correlation with 2PN formation ( B  = −2.540, 95% CI: −4.456 to −0.624, p  = 0.009). Next, we conducted an interaction analysis to clarify the independent role of IR from BMI on the 2PN rate. Surprisingly, the 2PN rate was influenced by both BMI and IR ( B  = −0.093, 95% CI: −0.179 to −0.007, p  = 0.033)(Table  3 ). This specificity highlights the importance of distinguishing between systemic metabolic dysregulation and localized ovarian insulin signaling abnormalities when assessing IR-related embryological issues. These findings support targeted studies on follicular microenvironment-specific insulin signaling to address differences between systemic and gamete-level effects of IR. Given the well-documented association between IR and obesity, we stratified patients into BMI categories: normal weight (BMI < 24 kg/m 2 ), overweight (24 kg/m 2  ≤ BMI < 28 kg/m 2 ), and obese (BMI ≥ 28 kg/m 2 ). To further assess the independent impact of IR on oocyte fertilization, subgroup analysis revealed a significantly reduced 2PN rate both in normal weight (80.78% vs. 83.79%, p  = 0.002) and overweight (80.77% vs. 84.51%, p  = 0.008) groups (see Supplementary Material 2, Supplementary Table 1). Furthermore, IR was identified as a risk factor for impaired fertilization specifically in normal-weight PCOS patients ( B : −2.694 (−4.768, −0.620), p  = 0.011) in an adjusted linear regression model (Table  4 ). This pattern underscores IR as an independent determinant that impairs the oocyte fertilization competence of normal-weight PCOS patients. Table 4 Association between BMI and insulin resistance affecting fertilization and 2PN rate Outcomes Variables B (95% CI) p value Adjusted B (95% CI) p value Fertilization rate  Normal weight BMI −0.729 (−1.153, −0.304)  < 0.001* −0.852 (−1.343, −0.360)  < 0.001* IR −0.458 (−2.280, 1.364) 0.622 0.826 (−1.184, 2.835) 0.420  Overweight BMI −0.348 (−1.422, 0.7726) 0.525 −0.018 (−1.201, 1.165) 0.976 IR −3.449 (−5.993, −0.904) 0.008* −3.417 (−6.146, −0.688) 0.014#  Obese BMI 0.004 (−0.853, 0.860) 0.993 −0.096 (−1.012, 0.821) 0.837 IR 2.407 (−2.825, 7.639) 0.366 1.365 (−4.062, 6.792) 0.621 2PN rate  Normal weight BMI −0.278 (−0.718, 0.163) 0.217 −0.034 (−0.542, 0.473) 0.895 IR −3.013 (−4.885, −1.141) 0.002* −2.694 (−4.768, −0.620) 0.011#  Overweight BMI −1.865 (−3.020, −0.710) 0.002* −1.444 (−2.700, −0.188) 0.024# IR −3.749 (−6.508, −0.989) 0.008* −2.679 (−5.576, 0.217) 0.070  Obese BMI −0.209 (−1.084, 0.666) 0.638 −0.028 (−1.010, 0.955) 0.956 IR −2.295 (−7.645, 3.055) 0.399 −1.744 (−7.560, 4.073) 0.555 Adjusted for infertility, BMI, AFC, basal FSH, starting of Gn, days of Gn, total Gn dosage, FPG, FINS # p  < 0.05, * p  < 0.01 BMI  body mass index, AFC  antral follicle counts, FSH  follicle-stimulating hormone, Gn  gonadotropins, IR  insulin resistance, FPG  fasting plasma glucose, FINS  fasting insulin Association between BMI and insulin resistance affecting fertilization and 2PN rate Adjusted for infertility, BMI, AFC, basal FSH, starting of Gn, days of Gn, total Gn dosage, FPG, FINS # p  < 0.05, * p  < 0.01 BMI  body mass index, AFC  antral follicle counts, FSH  follicle-stimulating hormone, Gn  gonadotropins, IR  insulin resistance, FPG  fasting plasma glucose, FINS  fasting insulin Consistently, variables demonstrating statistically significant differences in baseline characteristics (Tables  1 and 2 ) were first analyzed via univariate logistic regression to identify potential predictors of embryological outcomes. Subsequently, all variables were incorporated into a multivariate logistic regression model to adjust for potential confounding biases, with backward elimination retaining covariates at p  < 0.05 significance (see Supplementary Material 1, Table 5). In the univariable model, maternal age predicted reduced clinical pregnancy rate (OR: 0.952, 95% CI: 0.928 to 0.977, p  = 0.001). Interestingly, the IR-associated index including FPG, FINS and HOMA failed to attain statistically significant associations with pregnancy rate in PCOS patients. In the aspect of ovarian stimulation protocol and embryological parameters, the number of transferred embryos exerted a modest yet significant positive influence on clinical pregnancy success (OR: 1.265, 95% CI: 1.008 to 1.584, p  = 0.043). Multivariate logistic regression incorporating covariates further corroborated age as a robust negative predictor (OR: 0.952, 95% CI: 0.924 to 0.980, p  = 0.001), while the number of transferred embryos (OR: 1.297, 95% CI: 1.022 to 1.646, p  = 0.032), and endometrial thickness (OR: 1.070, 95% CI: 1.007 to 1.137, p  = 0.029) emerged as favorable influencing factors. At this stage, FPG, FINS, and HOMA remained nonsignificant, underscoring IR’s limited explanatory power in PCOS pregnancy outcomes. Subsequent sensitivity analyses evaluating ongoing pregnancy outcomes reinforced these patterns. Similarly, age persisted as the dominant negative factor in both unadjusted and adjusted models. AFC demonstrated a marginal positive association (OR: 1.015, 95% CI: 1.004 to 1.026, p  = 0.008), while the use of a hormone-replacement therapy (HRT) cycle correlated with reduced ongoing pregnancy rate (OR: 0.731, 95% CI: 0.579 to 0.924, p  = 0.009). Unfortunately, IR-associated indices remained nonsignificant ( p  > 0.05). A power analysis confirmed the study was adequately powered for its primary outcomes (see Supplementary Material 2, Supplementary Table 2). However, definitively establishing the influence of IR on pregnancy outcomes in PCOS requires further prospective studies designed to control for confounding factors like varying treatment protocols and hormonal profiles.

PCOS is a prevalent infertility disorder influenced by multiple etiological factors. In this study, we classified PCOS patients based on HOMA and systematically investigated the impact of IR on clinical outcomes especially embryonic outcomes in individuals undergoing IVF/ICSI treatment. Our findings illuminate critical intersections between metabolic dysregulation and reproductive pathophysiology in PCOS. IR-positive patients exhibited distinct baseline characteristics, including a higher BMI, prolonged duration of infertility, and lower FSH levels. These observations align with established hypothalamic-pituitary-ovarian (HPO) axis dysfunction in PCOS, characterized by attenuated FSH secretion and elevated luteinizing hormone (LH)/FSH ratio—biomarkers correlated with FPG, FINS and HOMA [ 29 , 30 ]. Reduced FSH impaired granulosa cell aromatase activity, significantly suppressing cyclic follicular recruitment and preventing dominant follicle formation, with most follicles arrested at the antral follicle stage [ 31 ]. While basal FSH primarily reflects ovarian reserve, its suppression in IR may exacerbate ooplasmic energy insufficiency, potentially undermining fertilization competence and early embryogenesis [ 32 ]. During controlled ovarian stimulation, patients with IR required higher start doses of Gn, prolonged Gn stimulation, and greater total Gn consumption compared to non-IR individuals. This pharmacodynamic divergence may stem from BMI-driven alterations in recombinant FSH pharmacokinetics, as demonstrated by reduced bioavailability in overweight/obese PCOS cohorts [ 33 ]. Attenuated estrogenic negative feedback at the hypothalamic level exacerbates endogenous FSH deficiency [ 34 , 35 ], necessitating exogenous Gn escalation. Concurrently, obesity-associated chronic inflammation may perturb Gn receptor sensitivity via cytokine-mediated pathways, further modulating follicular response thresholds [ 36 ]. The increased Gn consumption led to LH hypersecretion, which induced ovulatory dysfunction and hyperandrogenism [ 37 ]. Additionally, IR also results in hyperinsulinemia stimulating GnRH secretion and ovarian theca cell androgen production, and decreases SHBG production, further contributing to hyperandrogenism [ 9 ]. Excess androgen production in ovarian theca cell results in follicular arrest and anovulation [ 38 ]. Mechanistically, IR impaired oocyte mitochondrial function by inducing oxidative stress, characterized by a reduced glutathione (GSH)/oxidized disulfide (GSSG) ratio, decreased ATP content, and reduced mitochondrial DNA copy number. These impairments occur in both germinal vesicle (GV)-stage and MII-stage oocytes. Consequently, this mitochondrial dysfunction increases spindle abnormalities and chromosomal misalignment in MII oocytes. In a PCOS-IR mouse model, this damage cascade elevates the incidence of fragmented embryo development [ 39 ]. Single-cell RNA sequencing analysis also revealed significant differences in blastocyst formation between IR and non-IR mouse models, linked to altered biological processes and signaling pathways including JAK-STAT signaling pathway, fatty acid metabolism, steroid biosynthesis, and carbon metabolism [ 40 ]. The transcriptome of granulosa cells from PCOS women exhibited impaired metabolic pathways, including downregulated fatty acid and cholesterol biosynthesis, as evidenced by GO and KEGG analysis [ 41 ]. These metabolic perturbations impair lipid-dependent ooplasmic maturation, compromising oocyte developmental competence [ 41 ]. Paralleling this, elevated levels of free fatty acids in follicular fluid and serum lipid levels—independent of IR—have been identified as critical determinants of oocyte quality in PCOS [ 42 , 43 ]. Furthermore, hormonal levels, particularly FSH, were found to exhibit a linear relationship with serum lipid levels [ 31 ], suggesting hormonal-metabolic crosstalk. Our results corroborate these findings: PCOS patients with dyslipidemia had significantly lower blastocyst rates compared to those with normal lipid levels ( p  = 0.040) (see Supplementary Material 2, Supplementary Table 3), with multivariable regression identifying dyslipidemia as a key predictor of embryological compromise (see Supplementary Material 2, Supplementary Table 4). These results underscore a synergistic interplay between IR and dyslipidemia in mediating metabolic embryotoxicity, highlighting dual pathways through which systemic dysfunction impairs IVF/ICSI outcomes in PCOS. Our findings revealed a statistically significant decline in both fertilization rate and 2PN rate in the normal-weight-PCOS-IR group independent of BMI. While a multicenter trial has demonstrated comparable risks of aneuploidy and mosaic embryo abnormalities between PCOS women and non-PCOS women, the specific impact of IR on embryonic developmental competence in PCOS populations remains underexplored [ 44 ]. To disentangle metabolic contributors, multivariable linear regression—restricted to FPG and FINS to mitigate multicollinearity—revealed FINS as the sole significant predictor of diminished 2PN rates, with no association observed for fertilization rates. These findings align with broader evidence of IR’s embryotoxic effects. In non-PCOS populations, IR correlates with reduced oocyte maturity [ 12 ], while PCOS-specific studies demonstrate negative associations between glucose metabolism indices (FPG, FINS, HOMA-IR) and MII oocyte yield [ 42 ]. A meta-analysis similarly demonstrated that the IR group had a significantly lower number of MII oocytes (−1.07, 95% CI: −1.54 to −0.59, p  < 0.001), which aligns with our findings [ 45 ]. One potential mechanism is that insulin can stimulate androgen production in theca cells, and elevated androgen levels may disrupt oocyte fertilization, thereby reducing the number of mature oocytes [ 46 ]. Concurrently, obesity-associated chronic inflammation exacerbates oxidative stress via the accumulation of toxic reactive oxygen species (ROS) [ 47 ], disrupting mitochondrial function and cytoskeletal organization critical to oocyte meiosis [ 48 ]. Additionally, Adiposity further induces epigenetic modifications that impair DNA integrity, which may compromise embryo development [ 49 ]. However, the precise mechanisms through which IR directly influences oocyte fertilization remain to be fully elucidated and warrant further investigation. Multivariable logistic regression analyses adjusting for baseline confounders (age, BMI, and duration of infertility, AFC, basal FSH, ovarian stimulation protocol, fertilization and endometrium preparation) revealed no statistically significant associations between FINS, FPG, and clinical or ongoing pregnancy rates. Although the influence of IR on implantation and pregnancy outcomes has been investigated, the null results observed in our study may be attributable to the strategic selection of the highest-grade embryos undergoing ART cycles. Due to their elevated number of antral follicles, PCOS patients typically yield more embryos during the oocyte retrieval process. The selection bias effectively filters out metabolically compromised embryos that might otherwise manifest IR-induced mitochondrial dysfunction or oxidative stress damage. Furthermore, hormonal therapy optimizes endometrial receptivity, creating an environment conducive to implantation, thereby potentially masking the underlying detrimental effects of IR on pregnancy success. Nonetheless, several limitations were encountered in the present analysis. Retrospective design precluded causal inference, while single-center recruitment limits generalizability. Prospective multicenter cohorts incorporating granular metabolic profiling, including follicular fluid oxidative stress markers and embryonic mitochondrial DNA quantification, are critical to elucidate the role of IR in pregnancy success. Additionally, integrating biomarkers of metabolic embryo competence with traditional morphological grading could refine ART protocols for PCOS populations.

In conclusion, our data underscore the importance of IR screening and metabolic optimization to enhance ART efficacy in normal-weight PCOS population. Our large-scale retrospective analysis demonstrates that IR adversely affects the fertilization process in normal-weight PCOS patients undergoing IVF/ICSI, potentially further impairing embryo quality. These findings highlight the need for mechanistic investigations to delineate IR-driven perturbations in folliculogenesis, oocyte maturation, and zygotic genome activation. Prospective trials integrating multi-omics approaches are warranted to elucidate how hyperinsulinemia and systemic metabolic dysfunction converge to diminish reproductive potential. Such insights will guide the development of optimal ovarian stimulation protocols and embryo selection strategies for PCOS patients.

Polycystic ovary syndrome (PCOS) is acknowledged as a prevalent infertility and endocrine disease affecting up to 6–21% of reproductive-age women based on racial diversity and various diagnostic criteria [ 1 ]. Abnormalities in gonadotropins secretion, ovarian folliculogenesis, steroidogenesis, and insulin secretion have been observed in individuals with PCOS [ 2 ]. Recent studies have indicated that genetic predispositions, epigenetic alterations, environmental factors, oxidative stress, chronic low-grade inflammation, mitochondrial dysfunction, and metabolic disorders are involved in the aetiology of PCOS and thus affect normal ovarian function [ 3 – 6 ]. Although PCOS manifests heterogeneously, IR emerges as a central pathophysiological hub, affecting 35–80% of patients and driving both metabolic dysregulation and reproductive dysfunction [ 7 ]. From a mechanistic perspective, IR impairs endometrial receptivity in PCOS patients by reducing glucose transporter 4 expression, leading to insufficient glucose supply in endometrial cells and finally causing abortion [ 8 ]. IR not only disrupts physiological glucose homeostasis but also serves as a pivotal driver in the pathogenesis and progression of metabolic disorders, including dyslipidemia, non-alcoholic fatty liver disease, atherosclerosis, and type 2 diabetes mellitus (T2DM) [ 9 ]. IR also leads to a series of cellular reactions that affect the clinical characteristics of PCOS. For instance, IR can lower the levels of hepatic sex hormone-binding globulin (SHBG), increase the pituitary responsiveness to gonadotropin-releasing hormone (GnRH), stimulate theca cells to produce androgens and elevate both total and free testosterone levels, thereby contributing to hyperandrogenism [ 9 ]. This is a key factor in ovulatory dysfunction and infertility [ 10 ]. Beyond the systemic metabolic effect of IR, accumulating evidence suggests that IR may directly impair clinical reproductive outcomes among patients with or without PCOS who are undergoing in vitro fertilization (IVF) [ 11 , 12 ]. It is widely acknowledged that a significant proportion of women with PCOS, ranging from 50 to 70%, are at an elevated risk of miscarriage, with insulin resistance (IR) identified as key contributing factors [ 13 – 15 ]. A meta-analysis study with 6137 PCOS patients revealed that IR was associated with a decreased risk of clinical pregnancy rate (OR: 0.77, 95% CI: 0.59 to 0.99, p  = 0.042) and an increased risk of spontaneous abortion (OR: 1.11, 95% CI: 1.02 to 1.22, p  = 0.017) in patients undergoing assisted reproductive technology (ART) treatment [ 16 ]. Another meta-analysis involving 11,182 patients with PCOS further corroborated the detrimental impact of IR on spontaneous abortion risk (MD: 0.32, 95% CI: 0.15 to 0.49) [ 17 ]. Despite these associations, the mechanistic interplay between IR and early reproductive processes—particularly oocyte fertilization competence and embryonic development—remains poorly elucidated. Recent studies highlight IR-associated reductions in high-quality embryo rates (36.8% vs. 39.7%,  p  = 0.005) and metabolic syndrome-driven oocyte quality deterioration in PCOS [ 18 – 20 ]. Pathogenic models implicate altered ovarian microenvironment, PCOS granulosa cell metabolic stress, mitochondrial dysfunction and autophagy imbalance as disruptors of oocyte-somatic cell crosstalk, potentially impairing oocyte meiotic competence [ 21 – 23 ]. While PCOS exhibits a significantly lower proportion of mature metaphase II (MII) (56.42% vs. 69.35% vs. 66.27%, p  < 0.001) oocyte compared to infertility with endometriosis or tubal factors, accompanied by markedly diminished meiotic spindle present (50.5% vs. 66.0% vs. 62.3%, p  < 0.001) [ 24 ], conflicting morphometric analyses report comparable ultrastructural anomalies between PCOS and non-PCOS cohorts (35.3% vs. 25.0%, p  = 0.720) [ 25 , 26 ] and similar fertilization rates (73.6% vs. 74.8%) [ 27 ]. These paradoxes suggest that IR-driven embryonic compromise may arise not from overt oocyte morphological defects but through metabolic perturbations affecting post-fertilization developmental competence. To address these gaps, we hypothesize that IR severity in PCOS patients negatively correlates with fertilization efficiency and early embryonic developmental potential, independent of obesity or ovarian stimulation protocols. By leveraging a large-scale retrospective cohort with standardized IVF/ICSI procedures, this study aims to analyze the relationship between fasting plasma glucose (FPG), fasting insulin (FINS) levels and key embryological outcomes, including fertilization, 2-pronuclei (2PN) formation, cleavage kinetics, and blastocyst quality. Multivariable regression models, adjusted for BMI, antral follicle count (AFC), and gonadotropin dosage, will elucidate the mechanistic relationships between IR severity and fertilization efficiency in this PCOS population. This study aims to clarify how metabolic stress impacts the earliest stages of human reproduction in PCOS.

Supplementary Material 1. Supplementary Material 2. Supplementary Material 1. Supplementary Material 2.