Risk factors and treatment strategies for live birth in women with high BMI undergoing ART Authors: Yunhan Nie, Yuting Huang, Zhe Kuang, Mingru Yin, Xi Shen, Yali Liu, Hongyuan Gao, Hui Long, Li Wang Published: 2026 ABSTRACT BackgroundWomen with a high BMI represent a growing and challenging population in assisted reproductive technology, as known impairments to treatment outcomes. However, the specific risk factors contributing to recurrent live birth failure and the potential treatment strategies to overcome initial setbacks in this group remain unclear.MethodsIn this large-scale retrospective cohort study (2010–2023), 10,581 women were stratified into overweight (BMI 24–27.99 kg/m ², n = 8,350) and obese (BMI ≥ 28 kg/m ², n = 2,231) groups. Two-tiered analysis used generalized linear mixed models (GLMMs): (1) multilevel multinomial logistic regression to identify risk factors for recurrent failure; and (2) multilevel logistic regression to evaluate treatment strategies following initial failed frozen embryo transfers (FETs).ResultsIn the risk factor analysis, women with recurrent failure were older (33–35 years) and had poorer ovarian reserve (AFC ~ 7–9) than those with live birth (~ 31 years, AFC ~ 12–15). Multivariable analysis confirmed advanced maternal age as a key risk factor (aOR = 1.98, 95% CI: 1.85–2.12). Higher hMG dose per mature follicle increased single-failure risk versus < 350 IU: for 350–500 IU, aOR = 1.43, 95% CI: 1.16–1.76 (overweight) and 1.66, 95% CI: 1.14–2.43 (obese); for > 500 IU, aOR = 1.77, 95% CI: 1.45–2.16 (overweight) and 1.58, 95% CI: 1.1–2.25 (obese). Switching to PPOS—versus mild stimulation—was associated with a lower risk of continued failure (aOR = 0.54, 95% CI: 0.40–0.74). Propensity score matching confirmed PPOS positively related with embryological yields; live birth rates were higher but not statistically significant (45.7% vs. 43.7%). Blastocyst transfer (aOR = 0.54, 95% CI: 0.42–0.69) was associated with a lower risk of continued failure.ConclusionsThis study is, to our knowledge, one of the largest studies to systematically distinguish non-modifiable baseline risks (age and AFC) from modifiable ART-related factors (stimulation protocol, hMG dosage, and embryo transfer policy) for recurrent failure in a large population with high-BMI. This distinction provides preliminary insights to implement risk-stratified clinical approaches, especially for lower-risk patients. These approaches include adopting PPOS, optimizing gonadotropin dosage, and cautiously individualizing embryo transfer strategy, which may serve as potential therapeutic alternatives after initial setbacks.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s12916-026-05019-z. Background Obesity and its negative impacts on assisted reproduction treatment (ART)—including reduced ovarian response to gonadotropins [1], low embryonic quality [2], impaired endometrial receptivity [3], and decreased pregnancy and live birth rates [4]—are increasingly recognized. Evidence on body mass index (BMI) and live birth remains inconsistent [5, 6]. In the general obstetric population, Kalk et al. [5] found a U-shaped risk: underweight increases risks of preterm birth and fetal mortality, while obesity raises neonatal admissions for macrosomia and metabolic monitoring. In early ART outcomes, Chen et al. [6] reported no significant link between BMI and fresh embryo transfer success, though higher BMI was associated with increased gestational diabetes and hypertension. These general observation obscures substantial variability in patient outcomes: while some women with high BMI achieve live birth during their initial cycle, others succeed only after one or more unsuccessful attempts, and a significant subset experience recurrent failure [7, 8]. As a result, the factors associated with these distinct live birth trajectories—specifically, success, single-cycle failure, or recurrent failure—remain inadequately characterized in high-BMI populations [9, 10]. Moreover, potential treatment strategies that could improve live birth prospects for women with high-BMI following prior ART failures are currently lacking [11–14], underscoring the necessity for large-scale, longitudinal studies to inform clinical decision-making. In women with high BMI, factors influencing live birth outcomes—success, single failure, or recurrent failure may involve baseline patient characteristics (e.g., age, duration of infertility, antral follicle count (AFC)) and ART cycle features. These include controlled ovarian hyperstimulation (COH) parameters (e.g., COH protocol, gonadotropin dose, follicle-stimulating hormone (FSH), luteinizing hormone (LH), estradiol (E2), number of oocytes retrieved), in vitro fertilization/intracytoplasmic sperm injection (IVF/ICSI) procedures and frozen embryo transfer aspects (e.g., number and stage of transferred embryos, endometrial preparation) [1, 2]. Women with elevated BMI are often older and more likely to present with polycystic ovary syndrome [PCOS; recently proposed to be renamed polyendocrine metabolic ovarian syndrome (PMOS)] [15], conditions that may be associated with increased ovarian reserve, as reflected by AFC, and higher prevalence of secondary infertility [1, 8, 16]. Furthermore, obesity has been linked to diminished ovarian response during controlled ovarian stimulation, characterized by lower serum E2 and progesterone levels, necessitating higher doses of gonadotropins [1, 17, 18]. Reduced FSH levels on the day of trigger suggest impaired hypothalamic–pituitary responsiveness [18]. Nevertheless, there is still a lack of systematic and comprehensive evaluation of how population baseline characteristics and ART cycle profiles interact in high-BMI women experiencing recurrent live birth failure. For patients with prior ART cycle failures before live birth, it is unclear whether changes to COH protocols, IVF/ICSI procedures, or frozen embryo transfer (FET) strategies significantly influence subsequent success. Evidence shows that progestin-primed ovarian stimulation (PPOS) may reduce the negative effects of high BMI on embryo quality, clinical pregnancy rates [1], and cumulative live birth rates [19]. The GnRH-agonist long protocol may yield better pregnancy outcomes in fresh transfers among young women with obesity compared to HMG-based short protocols [20]. While ICSI may lower fertilization failure risk in women with obesity [21], it is not routinely recommended without male factor infertility due to no significant improvement in cumulative live birth rates [22]. Mild stimulation may lead to higher live birth rates and lower miscarriage rates than artificial cycles in women with overweight or obesity and PCOS (PMOS) [23]. Letrozole-based endometrial preparation appears non-inferior to programmed cycles in anovulatory women, offering a viable alternative [24]. However, existing evidence has critical gaps: (1) findings on optimal COH protocols and ART strategies are inconsistent and fragmented, with few head-to-head comparisons tailored to recurrent failure patients [1, 19–24]; (2) within-patient transition studies are virtually absent. Studies comparing sequential treatment transitions—from initial failure to live birth—within the same group are still needed. This study conducted a retrospective cohort analysis of clinical data from 10,581 women with high BMI (overweight and obese) and three live birth outcomes: success, single failure, or recurrent failure. The study tested its clinical hypotheses in two phases. First, we hypothesized that baseline patient characteristics and modifiable ART factors independently predict recurrent live birth failure—and identified key population-level and ART-related risk factors associated with it. Second, we hypothesized that customized clinical adjustments following an initial failure may mitigate the risk of continuous failure. Therefore, we explored transitional treatment strategies—including ovulation stimulation protocol selection, hMG dosage optimization, and management of modifiable factors—to reduce cycle number and potentially optimize treatment efficiency in patients who succeeded after prior failures. Methods Patients This large-scale, retrospective, population-based study analyzed clinical data from the Department of Assisted Reproduction at Shanghai Ninth People’s Hospital, Shanghai Jiao Tong University School of Medicine, covering 14 years (2010–2023). The Ethics Committee approved the study (Approval No. SH9H-2024-T134-1). We included 13,022 women with high BMI undergoing ART for infertility. Exclusion criteria: Asherman’s syndrome, major congenital uterine anomalies, no available embryos, fresh embryo transfer (to avoid confounding factors associated with the non - freezing status of embryos and the interference in the endometrium from COH), and donor sperm use (to avoid non-spousal genetic confounders). After exclusions, 10,581 women remained for analysis. Using Asian-specific BMI categories [25], we divided the cohort into two groups: BMI 24–27.99 kg/m² (n = 8,350) and BMI ≥ 28 kg/m² (n = 2,231). Study design In the primary analysis—which was conducted as a risk factor analysis—we examined population and ART cycle characteristics associated with live birth after FET in women with high BMI. They were divided into two BMI groups and categorized by live birth outcome: (1) Success—live birth achieved; (2) Single-failure—one failed FET cycle; (3) Recurrent-failure—at least two consecutive failed FET cycles. A live birth was defined as the delivery of at least one live-born infant. A failed FET cycle was defined as any FET cycle that did not result in a live birth, including implantation failure or pregnancy loss at any stage. In this study, “recurrent live birth failure” (non-standard terminology) is defined as failing to achieve a live birth after two or more FET cycles. It reflects cumulative inability to sustain pregnancy to term—distinct from recurrent implantation failure (RIF), which refers only to failure of initial embryo attachment [26, 27]. Using ≥ 2 cycles as the threshold helps distinguish persistent pathology from sporadic failure, consistent with distinct clinical features observed in our recurrent-failure cohort [28]. For ART cycle analysis, each woman’s cycles were included as follows: in the success and single-failure groups, only the FET cycle (successful or failed, respectively) and its matched IVF/ICSI cycle were included; in the recurrent-failure group, all failed FET cycles and their matched IVF/ICSI cycles were included. In the secondary analysis—a transitional strategy analysis—we focused only on women who achieved a live birth after at least one failed FET. For the ART cycle analysis, all unsuccessful cycles were aggregated and compared with the subsequent successful cycle. In this subgroup, we evaluated treatment transitions related to live birth. Controlled ovarian hyperstimulation (COH) and oocyte retrieval This study encompassed a diverse spectrum of controlled ovarian stimulation strategies, including both the long and short GnRH agonist protocols [29], the mild stimulation protocol [30], the GnRH antagonist protocol [31], and the PPOS-related protocols [1, 19]. Final oocyte maturation was triggered when at least one dominant follicle reached ≥ 18 mm, or two or more follicles measured ≥ 17–18 mm, depending on ovarian response. Triggering used one of three methods: [a] subcutaneous triptorelin (0.1–0.2 mg; Decapeptyl, Ferring Pharmaceuticals, China), [b] intramuscular hCG (1,000–5,000 IU; Lizhu Pharmaceutical Trading Co., China), or [c] a dual trigger with both agents at the same doses and routes. Oocytes were retrieved transvaginally under ultrasound guidance 32–40 hours post-trigger [1, 32–34], with aspiration of all follicles > 10 mm in diameter. Besides, the ‘hMG dose per follicle ≥ 14 mm’ was defined as the total dose of hMG administered during the COH cycle, divided by the total number of follicles reaching a mean diameter of ≥ 14 mm, as measured by transvaginal ultrasound on the day of trigger. Laboratory protocols Fertilization of retrieved oocytes was performed using either conventional IVF or ICSI. On day 3 post-fertilization, all embryos were morphologically evaluated according to the Cummins criteria [35], which assesses blastomere number, symmetry, and degree of fragmentation. A differential cryopreservation strategy was subsequently applied based on the number of high-quality cleavage-stage embryos. Top-quality embryos, defined as grade I–II 8-cell blastomere embryos, were vitrified at the cleavage stage. For patients yielding fewer than six high-quality embryos, all were vitrified at the cleavage stage. When more than six high-quality embryos were obtained, the surplus embryos were further cultured to the blastocyst stage. Additionally, embryos that failed to meet the quality criteria on day 3 (grade III–IV) were also extended in culture to assess their potential for blastocyst development. Blastocysts were graded on day 5 or day 6 using the Gardner and Schoolcraft classification system [36], which evaluates blastocoel expansion, inner cell mass (ICM), and trophectoderm (TE) morphology. Only blastocysts exhibiting good morphological quality were selected for cryopreservation. FET and endometrial preparation Personalized endometrial preparation strategies were designed for each FET cycle based on the patient’s clinical characteristics [37, 38]. Protocol selection was guided by menstrual regularity: patients with irregular cycles underwent a mild stimulation protocol using letrozole (2.5 or 5 mg daily for 5 days starting on menstrual cycle day 3 [MC3]), while those with regular cycles were eligible for either a natural cycle (NC) or late follicular phase stimulation with hMG. In NC cycles, follicular monitoring began on MC12 or upon detection of a LH surge; if follicular development was inadequate (dominant follicle < 10 mm on MC12), the protocol was converted to hMG stimulation. A hormone replacement therapy (HRT) regimen was reserved for patients with a history of thin endometrium (≤ 6 mm) or prior implantation failure with other protocols. For all non-HRT protocols, ovulation was triggered with 5000 IU of urinary hCG upon meeting the following criteria: (a) dominant follicle diameter ≥ 17 mm, (b) endometrial thickness > 8 mm, (c) serum E2 levels > 150 pg/mL, and (d) serum progesterone < 1 ng/mL. Luteal phase support with progesterone was initiated 2–3 days after hCG administration. All embryo transfers, limited to a maximum of two embryos, were performed under transabdominal ultrasound guidance, with transfer timing based on embryo developmental stage: 4–5 days post-hCG for cleavage-stage embryos and 6–7 days post-hCG for blastocysts. Statistical analysis Data analysis was conducted using IBM SPSS Statistics (Version 26.0.0; IBM Corp., Armonk, NY, USA), R software (version 4.5.2, R Core Team, Vienna, Austria), and GraphPad Prism (Version 10.1.2; Dotmatics, Boston, MA, USA). Continuous variables are reported as mean ± standard deviation (SD), and intergroup differences were assessed using the Kruskal-Wallis test or Mann-Whitney test. Categorical variables are presented as frequencies and percentages (n, %), with group comparisons performed using either the chi-square test or Fisher’s exact test, as appropriate. Post hoc analyses incorporated Bonferroni correction to adjust for multiple comparisons, and statistical significance was defined as P < 0.05. Repeated-measures ANOVA was employed to examine longitudinal changes in FSH, LH, E2, and progesterone levels across treatment phases among women with overweight or obesity, stratified by live birth outcome. Fold changes were normalized relative to the Success group, which was assigned a reference value of 1. We used Generalized Linear Mixed Models (GLMMs) with a patient-level random intercept to account for multiple ART cycles per patient. For risk factor analysis, multilevel multinomial logistic regression—using live birth as the reference outcome—assessed how baseline patient characteristics and cycle-level confounders influenced the three live birth outcomes (Table 3). For transitional strategy analysis, multilevel binary logistic regression identified predictors of success in subsequent cycles among patients who failed to achieve live birth in prior cycles (Table 5). P-values in the multivariable regression models (Tables 3 and 5) were adjusted using the False Discovery Rate (FDR) correction. To address clinical selection bias and confounding by indication, we used Propensity Score Matching (PSM), matching patients transitioning to the PPOS protocol with those transitioning to mild stimulation based on maternal age, BMI, infertility type and cause, and ovarian reserve markers. Missing data were handled as follows: variables with < 5% missingness were excluded; hormonal parameters with higher missingness were imputed using Multiple Imputation by Chained Equations (MICE), incorporating baseline characteristics, treatment factors, and prior hormone levels. Results Identify risk factors for recurrent live birth failure in women with high BMI, focusing on baseline and ART characteristics Study design flowchart Figure 1 shows the patient selection criteria. Of 13,022 women, 10,581 with high BMI underwent 16,746 IVF/ICSI cycles; 15,305 FET cycles were included in the analysis. Two analyses were conducted: (1) identifying risk factors for recurrent live birth failure in high-BMI women, stratified by outcome—success, single failure, or recurrent failure; and (2) evaluating transitional treatment strategies by comparing successful and failed ART cycles in high-BMI women who achieved live birth after prior failures. Fig. 1Patient grouping and analysis flowchart. The diagram outlines the inclusion and exclusion criteria for the final study population, which was used for two analyses: (1) a between-patient comparison of cohorts stratified by BMI and live birth outcome, and (2) an intra-individual comparison of failed versus successful cycles within the same patient. A live birth was defined as the delivery of at least one live-born infant. A failed FET cycle was defined as any FET cycle that did not result in a live birth, including implantation failure or pregnancy loss at any stage Patient grouping and analysis flowchart. The diagram outlines the inclusion and exclusion criteria for the final study population, which was used for two analyses: (1) a between-patient comparison of cohorts stratified by BMI and live birth outcome, and (2) an intra-individual comparison of failed versus successful cycles within the same patient. A live birth was defined as the delivery of at least one live-born infant. A failed FET cycle was defined as any FET cycle that did not result in a live birth, including implantation failure or pregnancy loss at any stage Baseline characteristics Table 1 shows baseline characteristics of women with overweight or obesity, stratified by BMI (24–27.99 vs. ≥28 kg/m²) and live birth outcome (success, single failure, recurrent failure). Women achieving live birth were younger (BMI 24–27.99: 31.6 ± 4.09 vs. 35.12 ± 5.77 vs. 35.18 ± 5.74; BMI ≥ 28: 30.88 ± 3.94 vs. 33.44 ± 5.82 vs. 33.63 ± 5.41, P < 0.001) and had shorter infertility durations (P < 0.001). In the BMI 24–27.99 group, live birth was linked to lower rates of regular menstrual cycles (52.25% vs. 58.31% and 58.69%, P < 0.001), likely due to higher PCOS (PMOS) prevalence (17.98% vs. 10.72% and 10.59%, P < 0.001). Similar trends occurred in the BMI ≥ 28 group (42.77% vs. 45.52% and 48.67%; PCOS (PMOS): 33.54% vs. 23.82% and 26.33%, P < 0.001). Secondary infertility was more common among those without live birth (BMI 24–27.99: 58.24% and 57.75% vs. 45.86%; BMI ≥ 28: 44.81% and 49.43% vs. 40.03%, P < 0.001). Tubal and mixed factors were the main causes of infertility across subgroups (tubal: 28.6%–41.25%; mixed: 26.72%–37.12%). The live birth group had a higher proportion of first-cycle IVF/ICSI and FET treatments (P < 0.001). Table 1Baseline characteristics of overweight and obese women, categorized by live birth outcomesBaseline characteristicsBMI 24-27.99BMI ≥ 28P valueSuccess(live birth)Single-failure(FET = 1)Recurrent-failure (FET ≥ 2)Success(live birth)Single-failure(FET = 1)Recurrent-failure (FET ≥ 2) Number of women (n) 4911140820311279424528Female age (mean, SD)31.6 ± 4.09a35.12 ± 5.77b35.18 ± 5.74b30.88 ± 3.94c33.44 ± 5.82d33.63 ± 5.41d < 0.001 Female BMI (mean, SD)25.46 ± 1.09a25.46 ± 1.1a25.5 ± 1.1a30.57 ± 3.16b30.61 ± 2.53b30.55 ± 2.95b < 0.001 Number of women with regular menstrual cycle (24–38 d) (n, %)2566 (52.25%) a821 (58.31%) b1192 (58.69%) b547 (42.77%) c193 (45.52%) ac257 (48.67%) ac < 0.001 Infertility type (n, %) < 0.001 Primary infertility2659 (54.14%) a588 (41.76%) b858 (42.25%) b767 (59.97%) c234 (55.19%) ac267 (50.57%) aSecondary infertility2252 (45.86%) a820 (58.24%) b1173 (57.75%) b512 (40.03%) c190 (44.81%) ac261 (49.43%) aCause of infertility (n, %) < 0.001 Tubal factor2026 (41.25%) a498 (35.37%) bc765 (35.37%) ac388 (30.34%) b124 (29.25%) b151 (28.6%) bMale cause553 (11.26%) a117 (8.31%) b178 (8.76%) b119 (15.17%) ab38 (8.96%) ab45 (8.52%) abOvulatory factor448 (9.12%) a179 (12.71%) bc206 (10.14%) ac194 (15.2%) b65 (15.33%) b69 (13.07%) bcEndometrium factor109 (2.22%) ab40 (2.84%) ab59 (2.9%) b16 (1.25%) a6 (1.42%) ab10 (1.89%) abMixed causes1312 (26.72%) a406 (28.84%) ab609 (29.99%) ab414 (32.37%) bc129 (30.42%) abc196 (37.12%) cUnexplained factors463 (9.43%) a168 (11.93%) ab214 (10.54%) ab148 (11.57%) ab62 (14.62%) b57 (10.8%) abPCOS (PMOS) (n, %)883 (17.98%) a151 (10.72%) b215 (10.59%) b429 (33.54%) c101 (23.82%) d139 (26.33%) d < 0.001 Duration of infertility (year) 3.5 ± 3.09a3.93 ± 3.63ab4.08 ± 3.87b3.85 ± 2.96b4.22 ± 3.82b4.17 ± 3.63b < 0.001 Number of people with different IVF/ICSI cycles (n, %) < 0.001 13304 (67.28%) a647 (45.95%) b675 (33.23%) c875 (68.41%) a210 (49.53%) b212 (40.15%) b2965 (19.65%) a364 (25.85%) b511 (25.16%) b252 (19.7%) a117 (27.59%) b117 (22.16%) ab≥ 3642 (13.07%) a397 (28.2%) b845 (41.61%) c152 (11.88%) a97 (22.88%) b199 (37.69%) cNumber of people with different FET cycles (n, %) < 0.001 12638 (53.72%) a1408 (100%) b0 (0%) c753 (58.87%) d424 (100%) b0 (0%) c21374 (27.98%) a0 (0%) b1097 (54.01%) c324 (25.33%) a0 (0%) b298 (56.44%) c≥ 3899 (18.31%) a0 (0%) b934 (45.99%) c202 (15.79%) a0 (0%) b230 (43.56%) cA live birth was defined as the delivery of at least one live-born infant. A failed FET cycle was defined as any FET cycle that did not result in a live birth, including implantation failure or pregnancy loss at any stage. Kruskal-Wallis test, chi-square test, and Bonferroni correction were used. Different letters a, b, c, and d represent significant differences between groups. Statistical significance was defined as P < 0.05. BMI: Body Mass Index, PCOS: Polycystic Ovary Syndrome, PMOS: Polyendocrine Metabolic Ovarian Syndrome, IVF: In Vitro Fertilization, ICSI: Intracytoplasmic Sperm Injection, FET: Frozen Embryo Transfer Table 1Baseline characteristics of overweight and obese women, categorized by live birth outcomesBaseline characteristicsBMI 24-27.99BMI ≥ 28P valueSuccess(live birth)Single-failure(FET = 1)Recurrent-failure (FET ≥ 2)Success(live birth)Single-failure(FET = 1)Recurrent-failure (FET ≥ 2) Number of women (n) 4911140820311279424528Female age (mean, SD)31.6 ± 4.09a35.12 ± 5.77b35.18 ± 5.74b30.88 ± 3.94c33.44 ± 5.82d33.63 ± 5.41d < 0.001 Female BMI (mean, SD)25.46 ± 1.09a25.46 ± 1.1a25.5 ± 1.1a30.57 ± 3.16b30.61 ± 2.53b30.55 ± 2.95b < 0.001 Number of women with regular menstrual cycle (24–38 d) (n, %)2566 (52.25%) a821 (58.31%) b1192 (58.69%) b547 (42.77%) c193 (45.52%) ac257 (48.67%) ac < 0.001 Infertility type (n, %) < 0.001 Primary infertility2659 (54.14%) a588 (41.76%) b858 (42.25%) b767 (59.97%) c234 (55.19%) ac267 (50.57%) aSecondary infertility2252 (45.86%) a820 (58.24%) b1173 (57.75%) b512 (40.03%) c190 (44.81%) ac261 (49.43%) aCause of infertility (n, %) < 0.001 Tubal factor2026 (41.25%) a498 (35.37%) bc765 (35.37%) ac388 (30.34%) b124 (29.25%) b151 (28.6%) bMale cause553 (11.26%) a117 (8.31%) b178 (8.76%) b119 (15.17%) ab38 (8.96%) ab45 (8.52%) abOvulatory factor448 (9.12%) a179 (12.71%) bc206 (10.14%) ac194 (15.2%) b65 (15.33%) b69 (13.07%) bcEndometrium factor109 (2.22%) ab40 (2.84%) ab59 (2.9%) b16 (1.25%) a6 (1.42%) ab10 (1.89%) abMixed causes1312 (26.72%) a406 (28.84%) ab609 (29.99%) ab414 (32.37%) bc129 (30.42%) abc196 (37.12%) cUnexplained factors463 (9.43%) a168 (11.93%) ab214 (10.54%) ab148 (11.57%) ab62 (14.62%) b57 (10.8%) abPCOS (PMOS) (n, %)883 (17.98%) a151 (10.72%) b215 (10.59%) b429 (33.54%) c101 (23.82%) d139 (26.33%) d < 0.001 Duration of infertility (year) 3.5 ± 3.09a3.93 ± 3.63ab4.08 ± 3.87b3.85 ± 2.96b4.22 ± 3.82b4.17 ± 3.63b < 0.001 Number of people with different IVF/ICSI cycles (n, %) < 0.001 13304 (67.28%) a647 (45.95%) b675 (33.23%) c875 (68.41%) a210 (49.53%) b212 (40.15%) b2965 (19.65%) a364 (25.85%) b511 (25.16%) b252 (19.7%) a117 (27.59%) b117 (22.16%) ab≥ 3642 (13.07%) a397 (28.2%) b845 (41.61%) c152 (11.88%) a97 (22.88%) b199 (37.69%) cNumber of people with different FET cycles (n, %) < 0.001 12638 (53.72%) a1408 (100%) b0 (0%) c753 (58.87%) d424 (100%) b0 (0%) c21374 (27.98%) a0 (0%) b1097 (54.01%) c324 (25.33%) a0 (0%) b298 (56.44%) c≥ 3899 (18.31%) a0 (0%) b934 (45.99%) c202 (15.79%) a0 (0%) b230 (43.56%) cA live birth was defined as the delivery of at least one live-born infant. A failed FET cycle was defined as any FET cycle that did not result in a live birth, including implantation failure or pregnancy loss at any stage. Kruskal-Wallis test, chi-square test, and Bonferroni correction were used. Different letters a, b, c, and d represent significant differences between groups. Statistical significance was defined as P < 0.05. BMI: Body Mass Index, PCOS: Polycystic Ovary Syndrome, PMOS: Polyendocrine Metabolic Ovarian Syndrome, IVF: In Vitro Fertilization, ICSI: Intracytoplasmic Sperm Injection, FET: Frozen Embryo Transfer Baseline characteristics of overweight and obese women, categorized by live birth outcomes A live birth was defined as the delivery of at least one live-born infant. A failed FET cycle was defined as any FET cycle that did not result in a live birth, including implantation failure or pregnancy loss at any stage. Kruskal-Wallis test, chi-square test, and Bonferroni correction were used. Different letters a, b, c, and d represent significant differences between groups. Statistical significance was defined as P < 0.05. BMI: Body Mass Index, PCOS: Polycystic Ovary Syndrome, PMOS: Polyendocrine Metabolic Ovarian Syndrome, IVF: In Vitro Fertilization, ICSI: Intracytoplasmic Sperm Injection, FET: Frozen Embryo Transfer ART characteristics In Table 2, the success (live birth) group had significantly higher AFC, greater hMG dosage, and longer hMG stimulation duration, but lower hMG consumption per mature follicle (BMI 24–27.99: 318.47 ± 313.17 vs. 527.4 ± 503.97 vs. 502.38 ± 453.64; BMI ≥ 28: 362.86 ± 384.02 vs. 563.71 ± 521.65 vs. 612.98 ± 626.92, P < 0.001), and yielded more oocytes than both single- and recurrent-failure groups. Across both BMI categories, PPOS-based protocols were more common in the success group (BMI 24–27.99: 72.13% vs. 63.26% vs. 63.41%; BMI ≥ 28: 73.2% vs. 64.1% vs. 63.6%, P < 0.001), while mild stimulation was predominant in failed groups. Estradiol levels per mature follicle on trigger day were similar across live birth outcomes within each BMI group, though levels in the obese group were 20–30 pg/mL lower than in the overweight group. Table 2ART cycle characteristics of overweight and obese women, categorized by live birth outcomesIVF/ICSI characteristicsBMI 24-27.99BMI ≥ 28P valueSuccess(live birth)Single-failure(FET = 1)Recurrent-failure (FET ≥ 2)Success(live birth)Single-failure(FET = 1)Recurrent-failure (FET ≥ 2) Number of women (n) 4911140820311279424528 Number of COH cycles (n) 49412790560313367521324AFC (mean, SD)12.56 ± 8.2a6.6 ± 6.93b6.8 ± 6.67b14.57 ± 9.36c9.33 ± 8.92d8.62 ± 8.32d < 0.001 Type of COH protocol (n, %) < 0.001 Mild stimulation protocol528 (10.69%) a418 (14.98%) bcd892 (15.92%) d160 (11.98%) ac137 (18.22%) bd237 (17.9%) bdGnRH agonist long protocol75 (1.52%) a24 (0.86%) a92 (1.64%) a16 (1.20%) a6 (0.80%) a22 (1.66%) aGnRH agonist short protocol419 (8.48%) a96 (3.44%) b269 (4.8%) bc88 (6.59%) ac31 (4.12%) bc57 (4.31%) bcGnRH antagonist protocol174 (3.52%) a148 (5.3%) b253 (4.52%) ab45 (3.37%) ab33 (4.39%) ab67 (5.06%) abPPOS-related protocols3564 (72.13%) a1765 (63.26%) b3553 (63.41%) b978 (73.20%) a482 (64.10%) b842 (63.60%) bUndefined181 (3.66%) a339 (12.15%) b544 (9.71%) c49 (3.67%) a63 (8.38%) bc99 (7.48%) chMG dose (IU) (mean, SD)2000.36 ± 763.52a1556.34 ± 1038.31b1614.55 ± 969.1b2325.79 ± 1035.74c1972.29 ± 1217.23a2047.83 ± 1245.81a < 0.001 hMG treatment duration (mean, SD)9.36 ± 2.77a7.52 ± 4.21b7.81 ± 3.97c10.15 ± 3.5d8.76 ± 4.53e9.21 ± 4.57ae < 0.001 Follicles ≥ 14 mm at retrieval (mean, SD)9.99 ± 7.12a4.75 ± 5.19b5.13 ± 5.35c10.43 ± 7.29a5.95 ± 6.21d5.96 ± 6.15d < 0.001 hMG dose per follicle ≥ 14 mm (IU) (mean, SD)318.47 ± 313.17a527.4 ± 503.97b502.38 ± 453.64b362.86 ± 384.02c563.71 ± 521.65bd612.98 ± 626.92d < 0.001 E2 level per follicle ≥ 14 mm on trigger day (pg/mL) (mean, SD)354.95 ± 244.78a338.50 ± 209.10a354.80 ± 239.81a323.60 ± 253.72b321.29 ± 239.93b318.82 ± 181.69b < 0.001 Oocytes retrieved (mean, SD)11.69 ± 8.02a4.85 ± 5.88b5.37 ± 5.84c11.61 ± 8.21a5.48 ± 6.13c6.05 ± 6.58c < 0.001 Number of IVF/ICSI cycles (n, %) < 0.001 IVF2876 (58.21%) a1574 (56.42%) a3299 (58.88%) a752 (56.29%) a413 (54.92%) a739 (55.82%) aICSI1394 (28.21%) a1062 (38.06%) b2037 (36.36%) b346 (25.9%) a278 (36.97%) b499 (37.69%) bIVF+ICSI671 (13.58%) a154 (5.52%) bc267 (4.77%) c238 (17.81%) d61 (8.11%) b86 (6.5%) bcFertilized oocytes (mean, SD)8.88 ± 6.12a3.42 ± 4.36b3.97 ± 4.41c8.92 ± 6.32a3.94 ± 4.64c4.46 ± 4.88c < 0.001 Total embryos (mean, SD)8.09 ± 5.64a2.98 ± 4.05b3.47 ± 4.06c8.14 ± 5.87a3.49 ± 4.3c3.97 ± 4.5c < 0.001 High-quality embryos (mean, SD)4.44 ± 3.6a1.36 ± 2.1b1.86 ± 2.37c4.53 ± 3.66a1.56 ± 2.26b2.11 ± 2.59d < 0.001 FET characteristics BMI 24-27.99 BMI ≥ 28 P value Success (live birth) Single-failure (FET = 1) Recurrent-failure (FET ≥ 2) Success (live birth) Single-failure (FET = 1) Recurrent-failure (FET ≥ 2) Number of FET cycles (n) 49111408582212794241461Endometrium preparation protocol (n, %) < 0.001 Natural cycle747 (15.21%) a230 (16.34%) a836 (14.36%) a128 (10.01%) b33 (7.78%) b158 (10.81%) bHormone replacement cycle1691 (34.43%) a574 (40.77%) bc2625 (45.09%) cd515 (40.27%) b171 (40.33%) abcd691 (47.30%) dLetrozole stimulation cycle1553 (31.62%) a358 (25.43%) bc1361 (23.38%) c463 (36.2%) d147 (34.67%)ad415 (28.41%) abhMG late stimulation cycle902 (18.37%) a233 (16.55%) ab943 (16.2%) ab172 (13.45%) bc68 (16.04%) abc180 (12.32%) cOthers18 (0.37%) abc13 (0.92%) cd57 (0.98%) d1 (0.08%) b5 (1.18%) acd17 (1.16%) dEndometrial thickness (mm) (mean, SD)10.83 ± 2.44a10.41 ± 2.72b10.16 ± 2.53c10.91 ± 2.38ab10.57 ± 2.76abc10.52 ± 3.36b < 0.001 Number of transferred embryos (mean, SD)1.79 ± 0.41a1.59 ± 0.51b1.72 ± 0.46c1.76 ± 0.43a1.52 ± 0.5b1.69 ± 0.47c < 0.001 Stage of transferred embryos (n, %)0.466Cleavage embryo transfer4004 (81.53%)1151 (81.75%)4829 (82.94%)1048 (81.94%)339 (79.95%)1180 (80.77%)Blastocyst embryo transfer885 (18.02%)248 (17.69%)961 (16.6%)227 (17.75%)82 (19.43%)273 (18.83%)Combined22 (0.45%)9 (0.64%)32 (0.55%)4 (0.31%)3 (0.71%)8 (0.55%)Biochemical pregnancy rate (n, %)4911 (100%) a542 (38.49%) b1358 (23.33%) c1279 (100%) a188 (44.34%) b420 (28.75%) d < 0.001 Implantation rate (mean, SD)73.39 ± 25a24.64 ± 39.25b9.75 ± 25.31c74.64 ± 25.2a28.93 ± 41.05b13.43 ± 29.93d < 0.001 Clinical pregnancy rate (n, %)4911 (100%) a438 (31.11%) b848 (14.57%) c100% a155 (36.56%) b275 (18.82%) d < 0.001 Miscarriage rate (n, %)0 (0%) a438 (100%) b848 (100%) b0 (0%) a155 (100%) b275 (100%) b < 0.001 Live birth rate (n, %)4911 (100%) a0 (0%) b0 (0%) b100% a0 (0%) b0 (0%) b < 0.001 A live birth was defined as the delivery of at least one live-born infant. A failed FET cycle was defined as any FET cycle that did not result in a live birth, including implantation failure or pregnancy loss at any stage. Kruskal-Wallis test, chi-square test, and Bonferroni correction were used. Different letters a, b, c, d, and e represent significant differences between groups. Statistical significance was defined as P < 0.05. BMI: Body Mass Index, IVF: In Vitro Fertilization, ICSI: Intracytoplasmic Sperm Injection, FET: Frozen Embryo Transfer, COH: Controlled Ovarian Hyperstimulation, AFC: Antral Follicle Count, PPOS: Progestin-Primed Ovarian Stimulation, hMG: Human Menopausal Gonadotropin, E2: Estradiol Table 2ART cycle characteristics of overweight and obese women, categorized by live birth outcomesIVF/ICSI characteristicsBMI 24-27.99BMI ≥ 28P valueSuccess(live birth)Single-failure(FET = 1)Recurrent-failure (FET ≥ 2)Success(live birth)Single-failure(FET = 1)Recurrent-failure (FET ≥ 2) Number of women (n) 4911140820311279424528 Number of COH cycles (n) 49412790560313367521324AFC (mean, SD)12.56 ± 8.2a6.6 ± 6.93b6.8 ± 6.67b14.57 ± 9.36c9.33 ± 8.92d8.62 ± 8.32d < 0.001 Type of COH protocol (n, %) < 0.001 Mild stimulation protocol528 (10.69%) a418 (14.98%) bcd892 (15.92%) d160 (11.98%) ac137 (18.22%) bd237 (17.9%) bdGnRH agonist long protocol75 (1.52%) a24 (0.86%) a92 (1.64%) a16 (1.20%) a6 (0.80%) a22 (1.66%) aGnRH agonist short protocol419 (8.48%) a96 (3.44%) b269 (4.8%) bc88 (6.59%) ac31 (4.12%) bc57 (4.31%) bcGnRH antagonist protocol174 (3.52%) a148 (5.3%) b253 (4.52%) ab45 (3.37%) ab33 (4.39%) ab67 (5.06%) abPPOS-related protocols3564 (72.13%) a1765 (63.26%) b3553 (63.41%) b978 (73.20%) a482 (64.10%) b842 (63.60%) bUndefined181 (3.66%) a339 (12.15%) b544 (9.71%) c49 (3.67%) a63 (8.38%) bc99 (7.48%) chMG dose (IU) (mean, SD)2000.36 ± 763.52a1556.34 ± 1038.31b1614.55 ± 969.1b2325.79 ± 1035.74c1972.29 ± 1217.23a2047.83 ± 1245.81a < 0.001 hMG treatment duration (mean, SD)9.36 ± 2.77a7.52 ± 4.21b7.81 ± 3.97c10.15 ± 3.5d8.76 ± 4.53e9.21 ± 4.57ae < 0.001 Follicles ≥ 14 mm at retrieval (mean, SD)9.99 ± 7.12a4.75 ± 5.19b5.13 ± 5.35c10.43 ± 7.29a5.95 ± 6.21d5.96 ± 6.15d < 0.001 hMG dose per follicle ≥ 14 mm (IU) (mean, SD)318.47 ± 313.17a527.4 ± 503.97b502.38 ± 453.64b362.86 ± 384.02c563.71 ± 521.65bd612.98 ± 626.92d < 0.001 E2 level per follicle ≥ 14 mm on trigger day (pg/mL) (mean, SD)354.95 ± 244.78a338.50 ± 209.10a354.80 ± 239.81a323.60 ± 253.72b321.29 ± 239.93b318.82 ± 181.69b < 0.001 Oocytes retrieved (mean, SD)11.69 ± 8.02a4.85 ± 5.88b5.37 ± 5.84c11.61 ± 8.21a5.48 ± 6.13c6.05 ± 6.58c < 0.001 Number of IVF/ICSI cycles (n, %) < 0.001 IVF2876 (58.21%) a1574 (56.42%) a3299 (58.88%) a752 (56.29%) a413 (54.92%) a739 (55.82%) aICSI1394 (28.21%) a1062 (38.06%) b2037 (36.36%) b346 (25.9%) a278 (36.97%) b499 (37.69%) bIVF+ICSI671 (13.58%) a154 (5.52%) bc267 (4.77%) c238 (17.81%) d61 (8.11%) b86 (6.5%) bcFertilized oocytes (mean, SD)8.88 ± 6.12a3.42 ± 4.36b3.97 ± 4.41c8.92 ± 6.32a3.94 ± 4.64c4.46 ± 4.88c < 0.001 Total embryos (mean, SD)8.09 ± 5.64a2.98 ± 4.05b3.47 ± 4.06c8.14 ± 5.87a3.49 ± 4.3c3.97 ± 4.5c < 0.001 High-quality embryos (mean, SD)4.44 ± 3.6a1.36 ± 2.1b1.86 ± 2.37c4.53 ± 3.66a1.56 ± 2.26b2.11 ± 2.59d < 0.001 FET characteristics BMI 24-27.99 BMI ≥ 28 P value Success (live birth) Single-failure (FET = 1) Recurrent-failure (FET ≥ 2) Success (live birth) Single-failure (FET = 1) Recurrent-failure (FET ≥ 2) Number of FET cycles (n) 49111408582212794241461Endometrium preparation protocol (n, %) < 0.001 Natural cycle747 (15.21%) a230 (16.34%) a836 (14.36%) a128 (10.01%) b33 (7.78%) b158 (10.81%) bHormone replacement cycle1691 (34.43%) a574 (40.77%) bc2625 (45.09%) cd515 (40.27%) b171 (40.33%) abcd691 (47.30%) dLetrozole stimulation cycle1553 (31.62%) a358 (25.43%) bc1361 (23.38%) c463 (36.2%) d147 (34.67%)ad415 (28.41%) abhMG late stimulation cycle902 (18.37%) a233 (16.55%) ab943 (16.2%) ab172 (13.45%) bc68 (16.04%) abc180 (12.32%) cOthers18 (0.37%) abc13 (0.92%) cd57 (0.98%) d1 (0.08%) b5 (1.18%) acd17 (1.16%) dEndometrial thickness (mm) (mean, SD)10.83 ± 2.44a10.41 ± 2.72b10.16 ± 2.53c10.91 ± 2.38ab10.57 ± 2.76abc10.52 ± 3.36b < 0.001 Number of transferred embryos (mean, SD)1.79 ± 0.41a1.59 ± 0.51b1.72 ± 0.46c1.76 ± 0.43a1.52 ± 0.5b1.69 ± 0.47c < 0.001 Stage of transferred embryos (n, %)0.466Cleavage embryo transfer4004 (81.53%)1151 (81.75%)4829 (82.94%)1048 (81.94%)339 (79.95%)1180 (80.77%)Blastocyst embryo transfer885 (18.02%)248 (17.69%)961 (16.6%)227 (17.75%)82 (19.43%)273 (18.83%)Combined22 (0.45%)9 (0.64%)32 (0.55%)4 (0.31%)3 (0.71%)8 (0.55%)Biochemical pregnancy rate (n, %)4911 (100%) a542 (38.49%) b1358 (23.33%) c1279 (100%) a188 (44.34%) b420 (28.75%) d < 0.001 Implantation rate (mean, SD)73.39 ± 25a24.64 ± 39.25b9.75 ± 25.31c74.64 ± 25.2a28.93 ± 41.05b13.43 ± 29.93d < 0.001 Clinical pregnancy rate (n, %)4911 (100%) a438 (31.11%) b848 (14.57%) c100% a155 (36.56%) b275 (18.82%) d < 0.001 Miscarriage rate (n, %)0 (0%) a438 (100%) b848 (100%) b0 (0%) a155 (100%) b275 (100%) b < 0.001 Live birth rate (n, %)4911 (100%) a0 (0%) b0 (0%) b100% a0 (0%) b0 (0%) b < 0.001 A live birth was defined as the delivery of at least one live-born infant. A failed FET cycle was defined as any FET cycle that did not result in a live birth, including implantation failure or pregnancy loss at any stage. Kruskal-Wallis test, chi-square test, and Bonferroni correction were used. Different letters a, b, c, d, and e represent significant differences between groups. Statistical significance was defined as P < 0.05. BMI: Body Mass Index, IVF: In Vitro Fertilization, ICSI: Intracytoplasmic Sperm Injection, FET: Frozen Embryo Transfer, COH: Controlled Ovarian Hyperstimulation, AFC: Antral Follicle Count, PPOS: Progestin-Primed Ovarian Stimulation, hMG: Human Menopausal Gonadotropin, E2: Estradiol ART cycle characteristics of overweight and obese women, categorized by live birth outcomes Success (live birth) Single-failure (FET = 1) Success (live birth) Single-failure (FET = 1) A live birth was defined as the delivery of at least one live-born infant. A failed FET cycle was defined as any FET cycle that did not result in a live birth, including implantation failure or pregnancy loss at any stage. Kruskal-Wallis test, chi-square test, and Bonferroni correction were used. Different letters a, b, c, d, and e represent significant differences between groups. Statistical significance was defined as P < 0.05. BMI: Body Mass Index, IVF: In Vitro Fertilization, ICSI: Intracytoplasmic Sperm Injection, FET: Frozen Embryo Transfer, COH: Controlled Ovarian Hyperstimulation, AFC: Antral Follicle Count, PPOS: Progestin-Primed Ovarian Stimulation, hMG: Human Menopausal Gonadotropin, E2: Estradiol The Success (live birth) group more frequently used combined IVF and ICSI across both BMI categories, while ICSI alone was more common in the single-failure and recurrent-failure groups. Success groups were linked to better embryological outcomes, including more fertilized oocytes, greater total embryo yield and more high-quality embryos. Notably, the recurrent-failure group did not show significantly worse outcomes than the single-failure group in either BMI category. Endometrial preparation protocols differed significantly among groups. Hormone replacement therapy was least used in the Success group and increased in frequency from single- to recurrent-failure groups (BMI 24–27.99: 34.43% vs. 40.77% vs. 45.09%; BMI ≥ 28: 40.27% vs. 40.33% vs. 47.30%, P < 0.001). The Success group had slightly greater endometrial thickness and a significantly higher mean number of transferred embryos than the Failure group in both BMI categories (P < 0.001). No significant differences were observed in embryo developmental stage at transfer. Implantation and live birth rates differed significantly across groups, as expected. Hormonal changes during COH, including FSH, LH, E2, and P, are shown in Fig. 2A–H. In both BMI groups, women in the failure groups had similar serum FSH and LH levels during COH and on the trigger day compared to the success group, but significantly higher levels the day after triggering (Fig. 2A–B and E–F). Despite elevated gonadotropins, E2 and P secretion was lower in the failure groups from the start of COH through the trigger day and the following day (Fig. 2C–D and G–H). Normalized line plots in Fig. 2I–J compare key stimulation parameters—age, AFC, hMG dose per mature follicle, pituitary FSH, ovarian E2, and oocyte yield—between single- and recurrent-failure groups. Compared with the success group, failure groups were older, required higher hMG doses per follicle, had similar FSH levels, reduced E2 production, and yielded fewer oocytes. Fig. 2Hormonal dynamics during COH and comparison of key patient and ART cycle parameters by live birth outcome and BMI. (A–H) Longitudinal changes in serum FSH (A, E), LH (B, F), E2 (C, G), and progesterone (P) (D, H) measured at baseline, treatment days 5–7, trigger day, and post-trigger. Panels A–D show data for women with BMI 24–27.99; panels E–H for BMI ≥ 28. (I–J) Normalized comparison of key patient and ART cycle parameters relative to the success group (set to 1), including female age, AFC, hMG dose per mature follicle, hormone levels at different time points, and oocytes retrieved, for BMI 24–27.99 (I) and BMI ≥ 28 (J). Data are mean ± SD. Repeated-measures ANOVA was used for statistical analysis in panels A–H. Black asterisk: single failure vs. live birth reference group; blue asterisk: recurrent failure vs. live birth reference group. In panels I and J, *P < 0.05, **P < 0.01, ***P < 0.001 vs. reference group. Abbreviations: AFC, Antral Follicle Count; BMI, Body Mass Index; COH, Controlled Ovarian Hyperstimulation; E2, Estradiol; FET, Frozen Embryo Transfer; FSH, Follicle-Stimulating Hormone; hMG, Human Menopausal Gonadotropin; LH, Luteinizing Hormone; P, Progesterone; SD, Standard Deviation Hormonal dynamics during COH and comparison of key patient and ART cycle parameters by live birth outcome and BMI. (A–H) Longitudinal changes in serum FSH (A, E), LH (B, F), E2 (C, G), and progesterone (P) (D, H) measured at baseline, treatment days 5–7, trigger day, and post-trigger. Panels A–D show data for women with BMI 24–27.99; panels E–H for BMI ≥ 28. (I–J) Normalized comparison of key patient and ART cycle parameters relative to the success group (set to 1), including female age, AFC, hMG dose per mature follicle, hormone levels at different time points, and oocytes retrieved, for BMI 24–27.99 (I) and BMI ≥ 28 (J). Data are mean ± SD. Repeated-measures ANOVA was used for statistical analysis in panels A–H. Black asterisk: single failure vs. live birth reference group; blue asterisk: recurrent failure vs. live birth reference group. In panels I and J, *P < 0.05, **P < 0.01, ***P < 0.001 vs. reference group. Abbreviations: AFC, Antral Follicle Count; BMI, Body Mass Index; COH, Controlled Ovarian Hyperstimulation; E2, Estradiol; FET, Frozen Embryo Transfer; FSH, Follicle-Stimulating Hormone; hMG, Human Menopausal Gonadotropin; LH, Luteinizing Hormone; P, Progesterone; SD, Standard Deviation Risk factors associated with recurrent live birth failure A multivariable multinomial logistic regression analysis assessed the impact of patient and ART cycle characteristics on live birth outcomes, stratified by BMI (Table 3). Advanced maternal age consistently predicted live birth failure across both BMI groups. In women with BMI 24–27.99 kg/m², older age was significantly associated with increased odds of both single-failure (aOR = 1.61, 95% CI: 1.47–1.76) and recurrent-failure (aOR = 1.98, 95% CI: 1.85–2.12; both P < 0.001, PFDR < 0.001). A similarly potent effect was observed in the BMI ≥ 28 kg/m² group, where age elevated the risk for both failure types (both PFDR < 0.001). Secondary infertility was associated with a 29% increase in the odds of single-failure within the BMI 24–27.99 group (PFDR = 0.008), though this association was not statistically significant in the BMI ≥ 28 cohort. Notably, PCOS (PMOS) appeared to be a protective factor against recurrent-failure in the BMI 24–27.99 group (aOR = 0.75, PFDR = 0.018). Regarding ovarian reserve, higher basal LH and basal P levels showed an increasing trend in the odds of recurrent-failure in the BMI 24–27.99 group (PFDR = 0.089 and 0.036, respectively). Table 3A multilevel multinomial logistic regression analysis to assess the impact of potential confounding factors on the live birth outcomesIVF/ICSI characteristicsBMI 24-27.99 (Success cycles as ref.)BMI ≥ 28 (Success cycles of as ref.)Adjusted OR (95% CI)P, PFDRAdjusted OR (95% CI)P, PFDR Female age (n) Single-failure1.61 (1.47, 1.76)< 0.001, < 0.0011.71 (1.45, 2.01)< 0.001, < 0.001Recurrent-failure1.98 (1.85, 2.12)< 0.001, < 0.0011.84 (1.62, 2.08)< 0.001, < 0.001 Duration of infertility (y) Single-failure1.07 (0.99, 1.15)0.071, 0.1341.05 (0.91, 1.2)0.511, 0.758Recurrent-failure1.06 (1, 1.12)0.042, 0.0890.98 (0.88, 1.09)0.707, 0.856 Infertility type (Primary as ref.) Single-failure1.29 (1.1, 1.52)0.002, 0.0080.94 (0.7, 1.27)0.702, 0.856Recurrent-failure1.1 (0.98, 1.23)0.119, 0.21.16 (0.93, 1.44)0.192, 0.588 PCOS (PMOS) (not as ref.) Single-failure0.91 (0.68, 1.22)0.522, 0.6220.73 (0.45, 1.17)0.192, 0.588Recurrent-failure0.75 (0.61, 0.91)0.005, 0.0180.91 (0.64, 1.29)0.593, 0.817 AFC (n) Single-failure0.86 (0.78, 0.96)0.005, 0.021.02 (0.86, 1.21)0.83, 0.872Recurrent-failure0.95 (0.89, 1.02)0.158, 0.2340.89 (0.79, 1.02)0.089, 0.424 Basal FSH (mIU/mL) Single-failure0.98 (0.9, 1.07)0.629, 0.6841.09 (0.95, 1.25)0.232, 0.624Recurrent-failure1.02 (0.96, 1.08)0.625, 0.6841.03 (0.91, 1.15)0.674, 0.856 Basal LH (mIU/mL) Single-failure1.2 (1.09, 1.31)< 0.001, < 0.0011.07 (0.92, 1.25)0.36, 0.676Recurrent-failure1.08 (1, 1.17)0.041, 0.0891.06 (0.94, 1.21)0.345, 0.668 Basal E2 (pg/mL) Single-failure1.02 (0.93, 1.11)0.683, 0.7171.06 (0.93, 1.21)0.384, 0.7Recurrent-failure1 (0.94, 1.07)0.936, 0.9510.92 (0.8, 1.04)0.18, 0.588 Basal P (ng/mL) Single-failure1.06 (0.97, 1.14)0.197, 0.2710.91 (0.75, 1.1)0.31, 0.644Recurrent-failure1.08 (1.02, 1.14)0.012, 0.0360.97 (0.88, 1.08)0.585, 0.817 Type of COH protocol (Mild stimulation protocol as ref.) GnRH agonist long protocolSingle-failure0.73 (0.33, 1.63)0.446, 0.5531.97 (0.6, 6.41)0.262, 0.624Recurrent-failure1.51 (0.98, 2.32)0.063, 0.1211.16 (0.44, 3.05)0.764, 0.859GnRH agonist short protocolSingle-failure0.74 (0.48, 1.15)0.178, 0.2510.56 (0.26, 1.21)0.142, 0.518Recurrent-failure0.94 (0.71, 1.24)0.649, 0.6940.75 (0.43, 1.31)0.312, 0.644GnRH antagonist protocolSingle-failure1.59 (1.05, 2.4)0.029, 0.0761.24 (0.6, 2.54)0.561, 0.809Recurrent-failure1.52 (1.12, 2.07)0.008, 0.0281.91 (1.1, 3.31)0.022, 0.158PPOS-related protocolsSingle-failure0.89 (0.7, 1.13)0.349, 0.4710.68 (0.45, 1.02)0.064, 0.332Recurrent-failure0.93 (0.78, 1.11)0.424, 0.5360.93 (0.67, 1.28)0.654, 0.856 hMG treatment duration (d) Single-failure0.74 (0.64, 0.85)< 0.001, < 0.0010.89 (0.68, 1.17)0.4, 0.702Recurrent-failure0.85 (0.76, 0.94)0.002, 0.0090.94 (0.77, 1.14)0.513, 0.758 hMG dose (IU) Single-failure1.16 (1.01, 1.34)0.035, 0.0831.17 (0.89, 1.54)0.258, 0.624Recurrent-failure1.18 (1.06, 1.31)0.003, 0.0111.29 (1.05, 1.57)0.014, 0.123 hMG dose per follicle≥14 mm (IU) (< 350IU as ref.) 350–500 IUSingle-failure1.43 (1.16, 1.76)0.001, 0.0041.66 (1.14, 2.43)0.009, 0.109Recurrent-failure1.16 (1, 1.35)0.054, 0.1110.98 (0.72, 1.32)0.873, 0.902> 500IUSingle-failure1.77 (1.45, 2.16)< 0.001, < 0.0011.58 (1.1, 2.25)0.012, 0.123Recurrent-failure1.18 (1.02, 1.37)0.031, 0.0771.1 (0.84, 1.44)0.47, 0.739 Insemination method (IVF as ref.) ICSISingle-failure1.56 (1.31, 1.85)< 0.001, < 0.0011.42 (1.03, 1.96)0.034, 0.19Recurrent-failure1.31 (1.15, 1.48)< 0.001, < 0.0011.13 (0.88, 1.44)0.34, 0.668IVF+ICSISingle-failure1.02 (0.8, 1.32)0.858, 0.8870.93 (0.62, 1.4)0.729, 0.856Recurrent-failure1.05 (0.89, 1.24)0.553, 0.6470.99 (0.74, 1.33)0.946, 0.946 Endometrium preparation protocol (Natural cycle as ref.) Hormone replacement cycleSingle-failure1.17 (0.94, 1.45)0.167, 0.241.31 (0.79, 2.17)0.301, 0.644Recurrent-failure1.49 (1.27, 1.74)< 0.001, < 0.0011.33 (0.94, 1.87)0.107, 0.444Letrozole stimulation cycleSingle-failure1.1 (0.87, 1.4)0.419, 0.5361.52 (0.92, 2.54)0.105, 0.444Recurrent-failure1.14 (0.96, 1.34)0.134, 0.2140.98 (0.69, 1.4)0.928, 0.943hMG late stimulation cycleSingle-failure1 (0.78, 1.28)0.969, 0.9691.9 (1.09, 3.29)0.023, 0.158Recurrent-failure1.06 (0.89, 1.26)0.499, 0.6071.06 (0.71, 1.57)0.79, 0.859 Endometrial thickness (mm) Single-failure0.91 (0.85, 0.98)0.012, 0.0360.92 (0.79, 1.06)0.259, 0.624Recurrent-failure0.84 (0.8, 0.88)< 0.001, < 0.0010.96 (0.88, 1.05)0.408, 0.702 Two embryos transfer (one as ref.) Single-failure0.64 (0.59, 0.69)< 0.001, < 0.0010.57 (0.49, 0.65)< 0.001, < 0.001Recurrent-failure0.84 (0.79, 0.89)< 0.001, < 0.0010.86 (0.78, 0.96)0.007, 0.109 Blastocyst embryo transfer (cleavage as ref.) Single-failure0.77 (0.63, 0.94)0.012, 0.0360.75 (0.51, 1.08)0.123, 0.475Recurrent-failure0.96 (0.83, 1.11)0.582, 0.6641.1 (0.84, 1.44)0.477, 0.739A live birth was defined as the delivery of at least one live-born infant. A failed FET cycle was defined as any FET cycle that did not result in a live birth, including implantation failure or pregnancy loss at any stage. Analysis was performed using multilevel multinomial logistic regression with ‘Success cycles’ as the reference outcome. Statistical significance was defined as P < 0.05. Multicollinearity was assessed and ruled out (variance inflation factor < 10 for all predictors). Infertility causes were also included in the regression, but results are not shown. Abbreviations: BMI, Body Mass Index; IVF, In Vitro Fertilization; ICSI, Intracytoplasmic Sperm Injection; AFC, Antral Follicle Count; FSH, Follicle-Stimulating Hormone; LH, Luteinizing Hormone; E2, Estradiol; P, Progesterone; COH, Controlled Ovarian Hyperstimulation; PPOS, Progestin-Primed Ovarian Stimulation Table 3A multilevel multinomial logistic regression analysis to assess the impact of potential confounding factors on the live birth outcomesIVF/ICSI characteristicsBMI 24-27.99 (Success cycles as ref.)BMI ≥ 28 (Success cycles of as ref.)Adjusted OR (95% CI)P, PFDRAdjusted OR (95% CI)P, PFDR Female age (n) Single-failure1.61 (1.47, 1.76)< 0.001, < 0.0011.71 (1.45, 2.01)< 0.001, < 0.001Recurrent-failure1.98 (1.85, 2.12)< 0.001, < 0.0011.84 (1.62, 2.08)< 0.001, < 0.001 Duration of infertility (y) Single-failure1.07 (0.99, 1.15)0.071, 0.1341.05 (0.91, 1.2)0.511, 0.758Recurrent-failure1.06 (1, 1.12)0.042, 0.0890.98 (0.88, 1.09)0.707, 0.856 Infertility type (Primary as ref.) Single-failure1.29 (1.1, 1.52)0.002, 0.0080.94 (0.7, 1.27)0.702, 0.856Recurrent-failure1.1 (0.98, 1.23)0.119, 0.21.16 (0.93, 1.44)0.192, 0.588 PCOS (PMOS) (not as ref.) Single-failure0.91 (0.68, 1.22)0.522, 0.6220.73 (0.45, 1.17)0.192, 0.588Recurrent-failure0.75 (0.61, 0.91)0.005, 0.0180.91 (0.64, 1.29)0.593, 0.817 AFC (n) Single-failure0.86 (0.78, 0.96)0.005, 0.021.02 (0.86, 1.21)0.83, 0.872Recurrent-failure0.95 (0.89, 1.02)0.158, 0.2340.89 (0.79, 1.02)0.089, 0.424 Basal FSH (mIU/mL) Single-failure0.98 (0.9, 1.07)0.629, 0.6841.09 (0.95, 1.25)0.232, 0.624Recurrent-failure1.02 (0.96, 1.08)0.625, 0.6841.03 (0.91, 1.15)0.674, 0.856 Basal LH (mIU/mL) Single-failure1.2 (1.09, 1.31)< 0.001, < 0.0011.07 (0.92, 1.25)0.36, 0.676Recurrent-failure1.08 (1, 1.17)0.041, 0.0891.06 (0.94, 1.21)0.345, 0.668 Basal E2 (pg/mL) Single-failure1.02 (0.93, 1.11)0.683, 0.7171.06 (0.93, 1.21)0.384, 0.7Recurrent-failure1 (0.94, 1.07)0.936, 0.9510.92 (0.8, 1.04)0.18, 0.588 Basal P (ng/mL) Single-failure1.06 (0.97, 1.14)0.197, 0.2710.91 (0.75, 1.1)0.31, 0.644Recurrent-failure1.08 (1.02, 1.14)0.012, 0.0360.97 (0.88, 1.08)0.585, 0.817 Type of COH protocol (Mild stimulation protocol as ref.) GnRH agonist long protocolSingle-failure0.73 (0.33, 1.63)0.446, 0.5531.97 (0.6, 6.41)0.262, 0.624Recurrent-failure1.51 (0.98, 2.32)0.063, 0.1211.16 (0.44, 3.05)0.764, 0.859GnRH agonist short protocolSingle-failure0.74 (0.48, 1.15)0.178, 0.2510.56 (0.26, 1.21)0.142, 0.518Recurrent-failure0.94 (0.71, 1.24)0.649, 0.6940.75 (0.43, 1.31)0.312, 0.644GnRH antagonist protocolSingle-failure1.59 (1.05, 2.4)0.029, 0.0761.24 (0.6, 2.54)0.561, 0.809Recurrent-failure1.52 (1.12, 2.07)0.008, 0.0281.91 (1.1, 3.31)0.022, 0.158PPOS-related protocolsSingle-failure0.89 (0.7, 1.13)0.349, 0.4710.68 (0.45, 1.02)0.064, 0.332Recurrent-failure0.93 (0.78, 1.11)0.424, 0.5360.93 (0.67, 1.28)0.654, 0.856 hMG treatment duration (d) Single-failure0.74 (0.64, 0.85)< 0.001, < 0.0010.89 (0.68, 1.17)0.4, 0.702Recurrent-failure0.85 (0.76, 0.94)0.002, 0.0090.94 (0.77, 1.14)0.513, 0.758 hMG dose (IU) Single-failure1.16 (1.01, 1.34)0.035, 0.0831.17 (0.89, 1.54)0.258, 0.624Recurrent-failure1.18 (1.06, 1.31)0.003, 0.0111.29 (1.05, 1.57)0.014, 0.123 hMG dose per follicle≥14 mm (IU) (< 350IU as ref.) 350–500 IUSingle-failure1.43 (1.16, 1.76)0.001, 0.0041.66 (1.14, 2.43)0.009, 0.109Recurrent-failure1.16 (1, 1.35)0.054, 0.1110.98 (0.72, 1.32)0.873, 0.902> 500IUSingle-failure1.77 (1.45, 2.16)< 0.001, < 0.0011.58 (1.1, 2.25)0.012, 0.123Recurrent-failure1.18 (1.02, 1.37)0.031, 0.0771.1 (0.84, 1.44)0.47, 0.739 Insemination method (IVF as ref.) ICSISingle-failure1.56 (1.31, 1.85)< 0.001, < 0.0011.42 (1.03, 1.96)0.034, 0.19Recurrent-failure1.31 (1.15, 1.48)< 0.001, < 0.0011.13 (0.88, 1.44)0.34, 0.668IVF+ICSISingle-failure1.02 (0.8, 1.32)0.858, 0.8870.93 (0.62, 1.4)0.729, 0.856Recurrent-failure1.05 (0.89, 1.24)0.553, 0.6470.99 (0.74, 1.33)0.946, 0.946 Endometrium preparation protocol (Natural cycle as ref.) Hormone replacement cycleSingle-failure1.17 (0.94, 1.45)0.167, 0.241.31 (0.79, 2.17)0.301, 0.644Recurrent-failure1.49 (1.27, 1.74)< 0.001, < 0.0011.33 (0.94, 1.87)0.107, 0.444Letrozole stimulation cycleSingle-failure1.1 (0.87, 1.4)0.419, 0.5361.52 (0.92, 2.54)0.105, 0.444Recurrent-failure1.14 (0.96, 1.34)0.134, 0.2140.98 (0.69, 1.4)0.928, 0.943hMG late stimulation cycleSingle-failure1 (0.78, 1.28)0.969, 0.9691.9 (1.09, 3.29)0.023, 0.158Recurrent-failure1.06 (0.89, 1.26)0.499, 0.6071.06 (0.71, 1.57)0.79, 0.859 Endometrial thickness (mm) Single-failure0.91 (0.85, 0.98)0.012, 0.0360.92 (0.79, 1.06)0.259, 0.624Recurrent-failure0.84 (0.8, 0.88)< 0.001, < 0.0010.96 (0.88, 1.05)0.408, 0.702 Two embryos transfer (one as ref.) Single-failure0.64 (0.59, 0.69)< 0.001, < 0.0010.57 (0.49, 0.65)< 0.001, < 0.001Recurrent-failure0.84 (0.79, 0.89)< 0.001, < 0.0010.86 (0.78, 0.96)0.007, 0.109 Blastocyst embryo transfer (cleavage as ref.) Single-failure0.77 (0.63, 0.94)0.012, 0.0360.75 (0.51, 1.08)0.123, 0.475Recurrent-failure0.96 (0.83, 1.11)0.582, 0.6641.1 (0.84, 1.44)0.477, 0.739A live birth was defined as the delivery of at least one live-born infant. A failed FET cycle was defined as any FET cycle that did not result in a live birth, including implantation failure or pregnancy loss at any stage. Analysis was performed using multilevel multinomial logistic regression with ‘Success cycles’ as the reference outcome. Statistical significance was defined as P < 0.05. Multicollinearity was assessed and ruled out (variance inflation factor < 10 for all predictors). Infertility causes were also included in the regression, but results are not shown. Abbreviations: BMI, Body Mass Index; IVF, In Vitro Fertilization; ICSI, Intracytoplasmic Sperm Injection; AFC, Antral Follicle Count; FSH, Follicle-Stimulating Hormone; LH, Luteinizing Hormone; E2, Estradiol; P, Progesterone; COH, Controlled Ovarian Hyperstimulation; PPOS, Progestin-Primed Ovarian Stimulation A multilevel multinomial logistic regression analysis to assess the impact of potential confounding factors on the live birth outcomes A live birth was defined as the delivery of at least one live-born infant. A failed FET cycle was defined as any FET cycle that did not result in a live birth, including implantation failure or pregnancy loss at any stage. Analysis was performed using multilevel multinomial logistic regression with ‘Success cycles’ as the reference outcome. Statistical significance was defined as P < 0.05. Multicollinearity was assessed and ruled out (variance inflation factor < 10 for all predictors). Infertility causes were also included in the regression, but results are not shown. Abbreviations: BMI, Body Mass Index; IVF, In Vitro Fertilization; ICSI, Intracytoplasmic Sperm Injection; AFC, Antral Follicle Count; FSH, Follicle-Stimulating Hormone; LH, Luteinizing Hormone; E2, Estradiol; P, Progesterone; COH, Controlled Ovarian Hyperstimulation; PPOS, Progestin-Primed Ovarian Stimulation Risk factors related to COH characteristics were further analyzed. In the BMI 24–27.99 group, the GnRH antagonist protocol was associated with higher odds of recurrent-failure compared to the mild stimulation protocol (aOR = 1.52, PFDR = 0.028). While the aOR for the antagonist protocol in the BMI ≥ 28 group was 1.91, this association did not reach the threshold for FDR-corrected significance (PFDR = 0.158), suggesting a potential clinical trend. Regarding medication, a higher hMG dose per follicle (≥ 14 mm) emerged as a critical risk factor. Compared to the reference dose (< 350 IU), an hMG dose > 500 IU significantly predicted single-failure in the BMI 24–27.99 group (aOR = 1.77, PFDR < 0.001), while a dose of 350–500 IU showed a similar risk (aOR = 1.43, PFDR = 0.004). In the BMI ≥ 28 group, a dose of 350–500 IU and > 500 IU also showed a potential association with single-failure (aOR = 1.66, PFDR = 0.109; aOR = 1.58, PFDR = 0.123). In terms of laboratory and transfer parameters, ICSI was consistently associated with higher failure risks in the BMI 24–27.99 group (single: aOR = 1.56; recurrent: aOR = 1.31; both PFDR < 0.001). Within the BMI ≥ 28 group, ICSI showed a similar directional trend for single-failure (PFDR = 0.190). Hormone replacement cycles for endometrial preparation were associated with increased recurrent-failure in the BMI 24–27.99 group (aOR = 1.49, PFDR < 0.001). Double embryo transfer (DET) emerged as a robust protective factor, significantly reducing the odds of single-failure by 36% to 43% across both BMI groups, and recurrent-failure by 16% in the BMI 24–27.99 group (all PFDR < 0.001). A similar 14% reduction in recurrent-failure was observed in the BMI ≥ 28 group, though this did not remain statistically significant after FDR correction (PFDR = 0.109). Finally, blastocyst transfer appeared to lower the odds of single-failure in the BMI 24–27.99 group (aOR = 0.77, PFDR = 0.036). Identify transitional strategies linked to live birth after failed frozen embryo transfers in women with high BMI Characteristics of success vs. failure in ART cycles Table 4 compares failed and successful cycles in high-BMI women. Successful cycles were linked to higher baseline AFC (13.28 vs. 9.9; 15.21 vs. 12.59; P < 0.001), greater use of PPOS protocols (68.44% vs. 54.88%; 72.66% vs. 61.15%; P < 0.001), longer hMG stimulation, higher total hMG doses, but lower hMG dose per mature follicle (289.37 vs. 438.08 IU; 325.83 vs. 535.12 IU; P < 0.001). These cycles also had more follicles > 14 mm on trigger day and nearly twice as many oocytes retrieved (12.93 vs. 6.5; 13.05 vs. 6.15; P < 0.001). Serum hormone levels (FSH, LH, E2, P) during COH and key cycle parameters (Additional file 1: Fig. S1A–J) resembled those in Fig. 2. Table 4ART cycle characteristics within live birth success or failure in high - BMI women with prior failed attemptsIVF/ICSI characteristicsBMI 24-27.99BMI ≥ 28FailureSuccessP valueFailureSuccessP value Number of women (n) 2273575Female age (mean, SD)31.56 ± 4.1530.82 ± 4.01 Number of COH cycles (n) 19262262*471578COH attempt number (n, %) < 0.001 < 0.001 1825 (42.83%) a1346 (59.5%) b204 (43.31%) a342 (59.17%) b2449 (23.3%) a537 (23.7%) a105 (22.29%) a139 (24.05%) a≥ 3652 (33.9%) a379 (16.8%) b162 (34.39%) a97 (16.78%) bAFC (mean, SD)9.9 ± 7.7413.28 ± 8.47 < 0.001 12.59 ± 9.2915.21 ± 9.48 < 0.001 Type of COH protocol (n, %) < 0.001 < 0.001 Mild stimulation protocol374 (19.4%) a236 (10.4%) b93 (19.75%) a41 (7.09%) bGnRH agonist long protocol61 (3.2%) a71 (3.1%) a7 (1.49%) a14 (2.42%) aGnRH agonist short protocol145 (7.5%) a190 (8.4%) a27 (5.73%) a43 (7.44%) aGnRH antagonist protocol141 (7.32%) a124 (5.48%) b18 (3.82%) a32 (5.54%) aPPOS-related protocols1057 (54.88%) a1548 (68.44%) b288 (61.15%) a420 (72.66%) bUndefined148 (7.7%) a93 (4.1%) b38 (8.07%) a28 (4.84%) bhMG dose (IU) (mean, SD)1833.32 ± 820.272057.76 ± 739.92 < 0.001 2207.08 ± 1118.442464.27 ± 985.42 < 0.001 hMG treatment duration (mean, SD)8.87 ± 3.099.51 ± 2.55 < 0.001 10.05 ± 4.0610.47 ± 3.37 < 0.001 Follicles > 14 mm at retrieval (mean, SD)6.42 ± 5.1310.91 ± 7.27 < 0.001 6.8 ± 5.8912.04 ± 7.7 < 0.001 hMG dose per follicle ≥ 14 mm (IU) (mean, SD)438.08 ± 382.46289.37 ± 275.95 < 0.001 535.12 ± 505.06325.83 ± 349.02 < 0.001 E2 level per follicle ≥ 14 mm on trigger day (pg/mL) (mean, SD)356.18 ± 264.80361.88 ± 278.870.66321.11 ± 244.06315.66 ± 223.640.27Oocytes retrieved (mean, SD)6.5 ± 5.4312.93 ± 8.09 < 0.001 6.15 ± 5.3313.05 ± 8.12 < 0.001 Number of IVF/ICSI cycles (n, %) < 0.001 < 0.001 IVF1128 (58.6%) a1301 (57.5%) b245 (52.02%) a317 (54.84%) aICSI689 (35.8%) a664 (29.4%) a187 (39.7%) a169 (29.24%) bIVF+ICSI109 (5.7%) a297 (13.1%) b39 (8.28%) a92 (15.92%) bFertilized oocytes (mean, SD)4.59 ± 3.759.88 ± 6.13 < 0.001 4.35 ± 3.8410.14 ± 6.34 < 0.001 Total embryos (mean, SD)3.93 ± 3.298.93 ± 5.6 < 0.001 3.86 ± 3.459.22 ± 5.9 < 0.001 High-quality embryos (mean, SD)1.75 ± 1.774.96 ± 3.6 < 0.001 1.46 ± 1.425.1 ± 3.64 < 0.001 FET characteristics BMI 24-27.99 BMI ≥ 28 Failure Success P value Failure Success P value Number of FET cycles (n) 35962273872575Endometrium preparation protocol (n, %) 0.03 0.01 Natural cycle483 (13.43%) a290 (12.76%) a92 (10.55%) a46 (8%) aHormone replacement cycle1335 (37.12%) a939 (41.31%) b344 (39.45%) a267 (46.43%) bLetrozole stimulation cycle1144 (31.81%) a681 (29.96%) a298 (34.17%) a200 (34.78%) ahMG late stimulation cycle611 (16.99%) a353 (15.53%) a133 (15.25%) a62 (10.78%) bOthers23 (0.64%) a10 (0.44%) a5 (0.57%) a0 (0%) aEndometrial thickness (mm) (mean, SD)10.73 ± 2.5310.74 ± 2.40.3810.89 ± 2.6610.89 ± 2.340.44Number of transferred embryos (mean, SD)1.73 ± 0.471.84 ± 0.38 < 0.001 1.65 ± 0.51.81 ± 0.4 < 0.001 Stage of transferred embryos (n, %) < 0.001 0.13Cleavage-stage2917 (81.59%) a1799 (79.18%) b686 (79.21%)466 (81.18%)Blastocyst-stage648 (18.13%) a455 (20.03%) a180 (20.79%)106 (18.47%)Combined10 (0.28%) a18 (0.79%) b0 (0%)2 (0.35%)Biochemical pregnancy rate (n, %)916 (25.5%)2273 (100%) < 0.001 232 (26.6%)575 (100%) < 0.001 Implantation rate (mean, SD)10.34 ± 26.9270.9 ± 25.51 < 0.001 11.89 ± 29.4970.09 ± 25.79 < 0.001 Clinical pregnancy rate (n, %)521 (14.5%)2273 (100%) < 0.001 135 (15.5%)575 (100%) < 0.001 Miscarriage rate (n, %)521 (100%)0 (0%) < 0.001 135 (100%)0 (0%) < 0.001 A live birth was defined as the delivery of at least one live-born infant. A failed FET cycle was defined as any FET cycle that did not result in a live birth, including implantation failure or pregnancy loss at any stage. Mann-Whitney test, chi-square test, and Bonferroni correction were used. Different letters a and b represent significant differences between groups. Statistical significance was defined as P < 0.05. *Several COH cycles missing from the electric database. BMI: Body Mass Index, IVF: In Vitro Fertilization, ICSI: Intracytoplasmic Sperm Injection, FET: Frozen Embryo Transfer, COH: Controlled Ovarian Hyperstimulation, AFC: Antral Follicle Count, PPOS: Progestin-Primed Ovarian Stimulation, hMG: Human Menopausal Gonadotropin, E2: Estradiol Table 4ART cycle characteristics within live birth success or failure in high - BMI women with prior failed attemptsIVF/ICSI characteristicsBMI 24-27.99BMI ≥ 28FailureSuccessP valueFailureSuccessP value Number of women (n) 2273575Female age (mean, SD)31.56 ± 4.1530.82 ± 4.01 Number of COH cycles (n) 19262262*471578COH attempt number (n, %) < 0.001 < 0.001 1825 (42.83%) a1346 (59.5%) b204 (43.31%) a342 (59.17%) b2449 (23.3%) a537 (23.7%) a105 (22.29%) a139 (24.05%) a≥ 3652 (33.9%) a379 (16.8%) b162 (34.39%) a97 (16.78%) bAFC (mean, SD)9.9 ± 7.7413.28 ± 8.47 < 0.001 12.59 ± 9.2915.21 ± 9.48 < 0.001 Type of COH protocol (n, %) < 0.001 < 0.001 Mild stimulation protocol374 (19.4%) a236 (10.4%) b93 (19.75%) a41 (7.09%) bGnRH agonist long protocol61 (3.2%) a71 (3.1%) a7 (1.49%) a14 (2.42%) aGnRH agonist short protocol145 (7.5%) a190 (8.4%) a27 (5.73%) a43 (7.44%) aGnRH antagonist protocol141 (7.32%) a124 (5.48%) b18 (3.82%) a32 (5.54%) aPPOS-related protocols1057 (54.88%) a1548 (68.44%) b288 (61.15%) a420 (72.66%) bUndefined148 (7.7%) a93 (4.1%) b38 (8.07%) a28 (4.84%) bhMG dose (IU) (mean, SD)1833.32 ± 820.272057.76 ± 739.92 < 0.001 2207.08 ± 1118.442464.27 ± 985.42 < 0.001 hMG treatment duration (mean, SD)8.87 ± 3.099.51 ± 2.55 < 0.001 10.05 ± 4.0610.47 ± 3.37 < 0.001 Follicles > 14 mm at retrieval (mean, SD)6.42 ± 5.1310.91 ± 7.27 < 0.001 6.8 ± 5.8912.04 ± 7.7 < 0.001 hMG dose per follicle ≥ 14 mm (IU) (mean, SD)438.08 ± 382.46289.37 ± 275.95 < 0.001 535.12 ± 505.06325.83 ± 349.02 < 0.001 E2 level per follicle ≥ 14 mm on trigger day (pg/mL) (mean, SD)356.18 ± 264.80361.88 ± 278.870.66321.11 ± 244.06315.66 ± 223.640.27Oocytes retrieved (mean, SD)6.5 ± 5.4312.93 ± 8.09 < 0.001 6.15 ± 5.3313.05 ± 8.12 < 0.001 Number of IVF/ICSI cycles (n, %) < 0.001 < 0.001 IVF1128 (58.6%) a1301 (57.5%) b245 (52.02%) a317 (54.84%) aICSI689 (35.8%) a664 (29.4%) a187 (39.7%) a169 (29.24%) bIVF+ICSI109 (5.7%) a297 (13.1%) b39 (8.28%) a92 (15.92%) bFertilized oocytes (mean, SD)4.59 ± 3.759.88 ± 6.13 < 0.001 4.35 ± 3.8410.14 ± 6.34 < 0.001 Total embryos (mean, SD)3.93 ± 3.298.93 ± 5.6 < 0.001 3.86 ± 3.459.22 ± 5.9 < 0.001 High-quality embryos (mean, SD)1.75 ± 1.774.96 ± 3.6 < 0.001 1.46 ± 1.425.1 ± 3.64 < 0.001 FET characteristics BMI 24-27.99 BMI ≥ 28 Failure Success P value Failure Success P value Number of FET cycles (n) 35962273872575Endometrium preparation protocol (n, %) 0.03 0.01 Natural cycle483 (13.43%) a290 (12.76%) a92 (10.55%) a46 (8%) aHormone replacement cycle1335 (37.12%) a939 (41.31%) b344 (39.45%) a267 (46.43%) bLetrozole stimulation cycle1144 (31.81%) a681 (29.96%) a298 (34.17%) a200 (34.78%) ahMG late stimulation cycle611 (16.99%) a353 (15.53%) a133 (15.25%) a62 (10.78%) bOthers23 (0.64%) a10 (0.44%) a5 (0.57%) a0 (0%) aEndometrial thickness (mm) (mean, SD)10.73 ± 2.5310.74 ± 2.40.3810.89 ± 2.6610.89 ± 2.340.44Number of transferred embryos (mean, SD)1.73 ± 0.471.84 ± 0.38 < 0.001 1.65 ± 0.51.81 ± 0.4 < 0.001 Stage of transferred embryos (n, %) < 0.001 0.13Cleavage-stage2917 (81.59%) a1799 (79.18%) b686 (79.21%)466 (81.18%)Blastocyst-stage648 (18.13%) a455 (20.03%) a180 (20.79%)106 (18.47%)Combined10 (0.28%) a18 (0.79%) b0 (0%)2 (0.35%)Biochemical pregnancy rate (n, %)916 (25.5%)2273 (100%) < 0.001 232 (26.6%)575 (100%) < 0.001 Implantation rate (mean, SD)10.34 ± 26.9270.9 ± 25.51 < 0.001 11.89 ± 29.4970.09 ± 25.79 < 0.001 Clinical pregnancy rate (n, %)521 (14.5%)2273 (100%) < 0.001 135 (15.5%)575 (100%) < 0.001 Miscarriage rate (n, %)521 (100%)0 (0%) < 0.001 135 (100%)0 (0%) < 0.001 A live birth was defined as the delivery of at least one live-born infant. A failed FET cycle was defined as any FET cycle that did not result in a live birth, including implantation failure or pregnancy loss at any stage. Mann-Whitney test, chi-square test, and Bonferroni correction were used. Different letters a and b represent significant differences between groups. Statistical significance was defined as P < 0.05. *Several COH cycles missing from the electric database. BMI: Body Mass Index, IVF: In Vitro Fertilization, ICSI: Intracytoplasmic Sperm Injection, FET: Frozen Embryo Transfer, COH: Controlled Ovarian Hyperstimulation, AFC: Antral Follicle Count, PPOS: Progestin-Primed Ovarian Stimulation, hMG: Human Menopausal Gonadotropin, E2: Estradiol ART cycle characteristics within live birth success or failure in high - BMI women with prior failed attempts hMG dose per follicle ≥ 14 mm (IU) (mean, SD) A live birth was defined as the delivery of at least one live-born infant. A failed FET cycle was defined as any FET cycle that did not result in a live birth, including implantation failure or pregnancy loss at any stage. Mann-Whitney test, chi-square test, and Bonferroni correction were used. Different letters a and b represent significant differences between groups. Statistical significance was defined as P < 0.05. *Several COH cycles missing from the electric database. BMI: Body Mass Index, IVF: In Vitro Fertilization, ICSI: Intracytoplasmic Sperm Injection, FET: Frozen Embryo Transfer, COH: Controlled Ovarian Hyperstimulation, AFC: Antral Follicle Count, PPOS: Progestin-Primed Ovarian Stimulation, hMG: Human Menopausal Gonadotropin, E2: Estradiol IVF+ICSI was more common in successful cycles, while ICSI alone dominated in failed ones. Embryological outcomes were better in successful cycles: over twice as many fertilized oocytes, more total embryos, and nearly three times as many high-quality embryos. In FET, hormone replacement therapy was used more often in successful cycles in both BMI groups (41.31% vs. 37.12%, P = 0.03; 46.43% vs. 39.45%, P = 0.01). Successful cycles had a higher mean number of embryos transferred. In the BMI 24–27.99 group, fewer cleavage-stage transfers occurred in successful cycles (79.18% vs. 81.59%), though this difference was not significant in the BMI ≥ 28 group. Endometrial thickness did not differ significantly. As expected, biochemical, implantation, and clinical pregnancy rates were higher in successful cycles. Determinants of successful transition strategies from unsuccessful to live birth cycles A multivariable logistic regression identified factors associated with live birth in patients with prior failed ART cycles (Table 5). Among women with BMI 24–27.99, higher AFC was associated with higher live birth odds (aOR = 0.70, PFDR < 0.001). PPOS-based COH protocols were linked to a lower risk of subsequent treatment failure than mild stimulation in the BMI 24–27.99 group (aOR = 0.54, 95% CI: 0.40–0.74, PFDR < 0.001). Regarding medication dynamics, while a higher total hMG dose was associated with live birth odds (aOR = 0.56, PFDR < 0.001), prolonged treatment duration (aOR = 1.43, PFDR = 0.001) and excessive hMG doses per mature follicle (350–500 IU: aOR = 2.28, PFDR < 0.001; >500 IU: aOR = 1.31, PFDR = 0.027) significantly decreased the live birth odds. DET consistently mitigated live birth failure odds, yielding a 26% reduction in the BMI 24–27.99 group (aOR = 0.74, PFDR < 0.001) and a 38% reduction in the BMI ≥ 28 group (aOR = 0.62; PFDR < 0.001). Furthermore, blastocyst transfer was significantly associated with better live birth odds in the BMI 24–27.99 group (aOR = 0.54, PFDR < 0.001). Table 5A multilevel logistic regression analysis to assess the impact of potential confounding factors on the live birth outcomes among patients who transitioned following failed FET cyclesIVF/ICSI characteristicsBMI 24-27.99(Success cycles of one person as ref.)BMI ≥ 28(Success cycles of one person as ref.)Adjusted OR(95% CI)P, PFDRAdjusted OR(95% CI)P, PFDR AFC (n) 0.7 (0.62, 0.8)< 0.001, < 0.0011 (0.87, 1.14)0.965, 0.965 Basal FSH (mIU/mL) 1.1 (1, 1.22)0.058, 0.1230.91 (0.8, 1.04)0.162, 0.505 Basal LH (mIU/mL) 0.99 (0.88, 1.11)0.881, 0.8811.05 (0.92, 1.2)0.492, 0.721 Basal E2 (pg/mL) 1.01 (0.91, 1.11)0.874, 0.8811.06 (0.86, 1.31)0.589, 0.774 Basal P (ng/mL) 1.05 (0.96, 1.15)0.263, 0.3621.02 (0.9, 1.15)0.8, 0.88 Type of COH protocol Mild stimulation protocolReferenceReferenceGnRH agonist long protocol0.61 (0.33, 1.11)0.107, 0.1880.76 (0.28, 2.09)0.598, 0.774GnRH agonist short protocol0.58 (0.37, 0.93)0.023, 0.0530.56 (0.28, 1.14)0.112, 0.481GnRH antagonist protocol1.09 (0.68, 1.76)0.721, 0.7970.48 (0.24, 0.98)0.043, 0.316PPOS-related protocols0.54 (0.4, 0.74)< 0.001, < 0.0010.79 (0.5, 1.24)0.298, 0.505 hMG treatment duration (d) 1.43 (1.19, 1.73)< 0.001, 0.0011.14 (0.9, 1.45)0.284, 0.505 hMG dose (IU) 0.56 (0.45, 0.68)< 0.001, < 0.0010.87 (0.69, 1.11)0.262, 0.505 hMG dose per follicle≥14 mm < 350ReferenceReference350–5002.28 (1.68, 3.08)< 0.001, < 0.0011.33 (0.86, 2.06)0.202, 0.505> 5001.31 (1.07, 1.62)0.01, 0.0271.05 (0.81, 1.36)0.711, 0.869 Insemination method IVFReferenceReferenceICSI1.23 (0.98, 1.56)0.077, 0.1471.04 (0.79, 1.36)0.8, 0.88IVF+ICSI1.19 (0.86, 1.63)0.293, 0.3621.24 (0.88, 1.75)0.219, 0.505 Endometrium preparation protocol Natural cycleReferenceReferenceHormone replacement cycle0.8 (0.6, 1.06)0.122, 0.1960.61 (0.39, 0.96)0.032, 0.316Letrozole stimulation cycle0.85 (0.63, 1.14)0.28, 0.3620.66 (0.42, 1.03)0.069, 0.379hMG late stimulation cycle0.93 (0.68, 1.27)0.65, 0.7581.02 (0.61, 1.7)0.945, 0.965 Endometrial thickness (mm) 0.94 (0.86, 1.04)0.231, 0.3470.93 (0.83, 1.06)0.277, 0.505 Number of transferred embryos 1ReferenceReference20.74 (0.68, 0.82)< 0.001, < 0.0010.62 (0.54, 0.72)< 0.001, < 0.001 Stage of transferred embryos Cleavage embryo transferReferenceReferenceBlastocyst embryo transfer0.54 (0.42, 0.69)< 0.001, < 0.0010.77 (0.55, 1.08)0.131, 0.481A live birth was defined as the delivery of at least one live-born infant. A failed FET cycle was defined as any FET cycle that did not result in a live birth, including implantation failure or pregnancy loss at any stage. Analysis was performed using multilevel logistic regression with ‘Success cycles’ as the reference outcome. Statistical significance was defined as P < 0.05. Multicollinearity was assessed and ruled out (variance inflation factor < 10 for all predictors). Abbreviations: BMI, Body Mass Index; IVF, In Vitro Fertilization; ICSI, Intracytoplasmic Sperm Injection; AFC, Antral Follicle Count; FSH, Follicle-Stimulating Hormone; LH, Luteinizing Hormone; E2, Estradiol; P, Progesterone; COH, Controlled Ovarian Hyperstimulation; PPOS, Progestin-Primed Ovarian Stimulation Table 5A multilevel logistic regression analysis to assess the impact of potential confounding factors on the live birth outcomes among patients who transitioned following failed FET cyclesIVF/ICSI characteristicsBMI 24-27.99(Success cycles of one person as ref.)BMI ≥ 28(Success cycles of one person as ref.)Adjusted OR(95% CI)P, PFDRAdjusted OR(95% CI)P, PFDR AFC (n) 0.7 (0.62, 0.8)< 0.001, < 0.0011 (0.87, 1.14)0.965, 0.965 Basal FSH (mIU/mL) 1.1 (1, 1.22)0.058, 0.1230.91 (0.8, 1.04)0.162, 0.505 Basal LH (mIU/mL) 0.99 (0.88, 1.11)0.881, 0.8811.05 (0.92, 1.2)0.492, 0.721 Basal E2 (pg/mL) 1.01 (0.91, 1.11)0.874, 0.8811.06 (0.86, 1.31)0.589, 0.774 Basal P (ng/mL) 1.05 (0.96, 1.15)0.263, 0.3621.02 (0.9, 1.15)0.8, 0.88 Type of COH protocol Mild stimulation protocolReferenceReferenceGnRH agonist long protocol0.61 (0.33, 1.11)0.107, 0.1880.76 (0.28, 2.09)0.598, 0.774GnRH agonist short protocol0.58 (0.37, 0.93)0.023, 0.0530.56 (0.28, 1.14)0.112, 0.481GnRH antagonist protocol1.09 (0.68, 1.76)0.721, 0.7970.48 (0.24, 0.98)0.043, 0.316PPOS-related protocols0.54 (0.4, 0.74)< 0.001, < 0.0010.79 (0.5, 1.24)0.298, 0.505 hMG treatment duration (d) 1.43 (1.19, 1.73)< 0.001, 0.0011.14 (0.9, 1.45)0.284, 0.505 hMG dose (IU) 0.56 (0.45, 0.68)< 0.001, < 0.0010.87 (0.69, 1.11)0.262, 0.505 hMG dose per follicle≥14 mm < 350ReferenceReference350–5002.28 (1.68, 3.08)< 0.001, < 0.0011.33 (0.86, 2.06)0.202, 0.505> 5001.31 (1.07, 1.62)0.01, 0.0271.05 (0.81, 1.36)0.711, 0.869 Insemination method IVFReferenceReferenceICSI1.23 (0.98, 1.56)0.077, 0.1471.04 (0.79, 1.36)0.8, 0.88IVF+ICSI1.19 (0.86, 1.63)0.293, 0.3621.24 (0.88, 1.75)0.219, 0.505 Endometrium preparation protocol Natural cycleReferenceReferenceHormone replacement cycle0.8 (0.6, 1.06)0.122, 0.1960.61 (0.39, 0.96)0.032, 0.316Letrozole stimulation cycle0.85 (0.63, 1.14)0.28, 0.3620.66 (0.42, 1.03)0.069, 0.379hMG late stimulation cycle0.93 (0.68, 1.27)0.65, 0.7581.02 (0.61, 1.7)0.945, 0.965 Endometrial thickness (mm) 0.94 (0.86, 1.04)0.231, 0.3470.93 (0.83, 1.06)0.277, 0.505 Number of transferred embryos 1ReferenceReference20.74 (0.68, 0.82)< 0.001, < 0.0010.62 (0.54, 0.72)< 0.001, < 0.001 Stage of transferred embryos Cleavage embryo transferReferenceReferenceBlastocyst embryo transfer0.54 (0.42, 0.69)< 0.001, < 0.0010.77 (0.55, 1.08)0.131, 0.481A live birth was defined as the delivery of at least one live-born infant. A failed FET cycle was defined as any FET cycle that did not result in a live birth, including implantation failure or pregnancy loss at any stage. Analysis was performed using multilevel logistic regression with ‘Success cycles’ as the reference outcome. Statistical significance was defined as P < 0.05. Multicollinearity was assessed and ruled out (variance inflation factor < 10 for all predictors). Abbreviations: BMI, Body Mass Index; IVF, In Vitro Fertilization; ICSI, Intracytoplasmic Sperm Injection; AFC, Antral Follicle Count; FSH, Follicle-Stimulating Hormone; LH, Luteinizing Hormone; E2, Estradiol; P, Progesterone; COH, Controlled Ovarian Hyperstimulation; PPOS, Progestin-Primed Ovarian Stimulation A multilevel logistic regression analysis to assess the impact of potential confounding factors on the live birth outcomes among patients who transitioned following failed FET cycles A live birth was defined as the delivery of at least one live-born infant. A failed FET cycle was defined as any FET cycle that did not result in a live birth, including implantation failure or pregnancy loss at any stage. Analysis was performed using multilevel logistic regression with ‘Success cycles’ as the reference outcome. Statistical significance was defined as P < 0.05. Multicollinearity was assessed and ruled out (variance inflation factor < 10 for all predictors). Abbreviations: BMI, Body Mass Index; IVF, In Vitro Fertilization; ICSI, Intracytoplasmic Sperm Injection; AFC, Antral Follicle Count; FSH, Follicle-Stimulating Hormone; LH, Luteinizing Hormone; E2, Estradiol; P, Progesterone; COH, Controlled Ovarian Hyperstimulation; PPOS, Progestin-Primed Ovarian Stimulation To validate PPOS efficacy versus mild stimulation and reduce indication bias, we performed propensity score matching (PSM; Additional file 2: Table S1). After PSM, baseline characteristics were well balanced between the PPOS (n = 387) and mild stimulation (n = 222) groups (all standardized mean differences < 0.1). In the matched cohort, PPOS yielded significantly better embryological outcomes: more oocytes retrieved (6.87 vs. 5.53, P = 0.001), fertilized oocytes (5.21 vs. 4.15, P < 0.001), total embryos (4.63 vs. 3.78, P = 0.002), and high-quality embryos (2.66 vs. 2.24, P = 0.006). Clinical pregnancy (55.0% vs. 50.0%) and live birth rates (45.7% vs. 43.7%) were numerically higher with PPOS but not statistically significant. Discussion Main findings Our analysis of over 10,000 women with high BMI aimed to identify key factors behind recurrent live birth failure in ART and explore potential treatment strategies for success. We found that non-modifiable factors—advanced maternal age and low AFC—are baseline risks. More importantly, modifiable treatment factors strongly linked to failure include using the GnRH antagonist protocol, excessive gonadotropin dosage relative to follicular response (“follicle-level overstimulation”), suboptimal endometrial preparation (especially HRT in women with overweight), and single-embryo transfer. In turn, a reduced risk of continued failure was associated with switching to the PPOS protocol, reducing gonadotropin dose per mature follicle, judicious, case-specific double-embryo transfer, and using blastocyst embryos. These results provide cohort-based associations that may inform approaches to improving live birth rates in this population. Non-modifiable factors link to live birth failure: advanced maternal age and low AFC Our study identified advanced maternal age as a universal risk factor for live birth failure, with distinct clinical and ovarian reserve profiles varying by BMI. In the overweight group, women with live birth failure were on average 35.1–35.2 years old—3.5–3.6 years older than those who succeeded. In the obese group, affected women averaged 33.4–33.6 years, 2.5–2.7 years older than successful cases. These results support established evidence that advanced age reduces oocyte competence and IVF success [39, 40]. Multivariate analysis (Table 3) confirmed age as an independent predictor of single and recurrent failure across all BMI groups. Beyond age, ovarian reserve markers—particularly low AFC—was strong predictors of failure. Women with overweight who experienced treatment failure had a mean AFC of 6.6–6.8, nearly half that of successful cases (12.6). Women with obesity with live birth failure also had lower AFCs (9.3 and 8.6) than those who succeeded (14.6), though the gap was smaller. Notably, baseline AFC was higher in women with obesity than with overweight, likely due to higher PCOS (PMOS) prevalence. Multivariate analysis showed low AFC as an independent risk factor for single failure in women with overweight. A similar trend was observed in women with obesity but did not reach significance, possibly due to limited sample size. Mechanistically, low AFC reflects impaired folliculogenesis and reduced oocyte competence, leading to poor embryo development, implantation, and adverse reproductive outcomes [19, 41, 42]. Modifiable factors linked to live birth failure: GnRH antagonist protocol, excessive gonadotropin dosing relative to follicular response, suboptimal endometrial preparation, and single-embryo transfer The COH strategy is a key risk factor for recurrent live birth failure in women with high BMI. Table 2 shows clear links between COH protocols and live birth outcomes. Mild stimulation and GnRH antagonist protocols were more common in both single and recurrent failure groups among women with overweight or obesity, indicating lower success rates. In contrast, the GnRH agonist short protocol was less used in failed cycles. Notably, PPOS-related protocols showed the strongest positive association; they were underrepresented in failures and dominant in successes across BMI groups. Multivariate analysis (Table 3) identified the GnRH antagonist protocol as an independent risk factor for live birth failure in women with overweight, though this effect was not statistically significant in the obese group. This finding aligns with clinical evidence suggesting that women with overweight or obesity, especially those with PCOS (PMOS), have lower implantation and live birth rates with the GnRH antagonist protocol than normal-weight women [43]. Therefore, this protocol should be used cautiously in high-BMI patients. A key finding is that excessive hMG dose per mature follicle—not total hMG dose—is the core issue. Failed cycles involved significantly lower AFC (≈ 50% less) and fewer mature follicles, yet received only slightly lower total hMG doses. Consequently, hMG per mature follicle was 1.6–1.7× higher in failed cycles (502–527 IU vs. 318 IU in women with overweight; 564–613 IU vs. 363 IU in women with obesity). Estradiol per mature follicle remained similar across groups, confirming normal follicular function—indicating high doses did not recruit more follicles but overstimulated fewer ones, likely compromising oocyte quality. Multivariate analysis confirmed hMG per mature follicle as a strong independent risk factor for recurrent live birth failure in women with overweight or obesity. Longer hMG duration was protective in women with overweight, suggesting adequate stimulation time supports maturation. These findings suggest that “follicle-level overstimulation”—as reflected by dose per mature follicle—may be an important factor associated with live birth failure, and align with prior IVF studies showing supraphysiological gonadotropin dosing fails to improve recruitment and may reduce oocyte/embryo quality (e.g., Shen et al. found higher doses linked to fewer high-quality embryos [44]; Munch et al. reported lower live birth rates with higher doses in fresh cycles [45, 46]). In our cohort with a high BMI, COH induced a hormonal pattern characterized as “effective-central, ineffective-peripheral”, which was associated with groups experiencing live birth failure. Centrally, FSH and LH levels were similar during COH but markedly elevated after trigger versus the control live birth group (Fig. 2). Peripherally, poor ovarian response was evident from high hMG doses and persistently low E2. This pattern reflects a failed follicular response to strong pituitary stimulation, which impairs the follicular-to-luteal transition. This may reduce embryo implantation and early pregnancy success [47]. A secondary analysis (Additional file 1: Fig. S1) confirmed these changes were stimulation-induced, not baseline features, in high-BMI patients who failed to achieve live birth. Our findings suggest that optimizing FET outcomes requires BMI-stratified endometrial preparation and transfer strategies. In women with overweight, non-physiological protocols like HRT increase the risk of recurrent failure. Conversely, in women with obesity, hMG late stimulation raises the risk of single failure compared to natural cycles. This points to different underlying mechanisms: women with overweight may retain partial hypothalamic-pituitary-ovarian axis function [48–50], allowing quasi-physiological protocols to maintain oocyte–endometrium synchronization and ovary-uterus communication. Standard HRT, however, may disrupt these delicate processes. In contrast, women with obesity likely have more severe hormonal and metabolic dysregulation [51, 52], making milder interventions such as late stimulation insufficient. Furthermore, a thicker endometrium was significantly associated with a lower risk of both single and recurrent failure in women with overweight. Interestingly, this protective association was not observed in women with obesity. Regarding embryo transfer, transferring two embryos significantly reduced recurrent failure risk in both groups (P < 0.001). This is consistent with evidence that multiple embryo transfer increases per-cycle pregnancy chances, although it also carries a higher risk of multiple gestation [53]. For women with elevated BMI and lower implantation efficiency, transferring more embryos may help compensate for reduced success rates [54]. Overall, these results support tailoring FET protocols to obesity severity rather than adopting a one-size-fits-all approach. Future prospective studies are needed to confirm these differences and establish evidence-based, BMI-guided FET strategies. Transitional strategies associated with a reduced risk of continued failure: PPOS protocol, lower gonadotropin dose per mature follicle, double-embryo transfer, and blastocyst-stage embryo transfer Our within-patient analysis reveals a key shift in ovarian stimulation among successful cycles: increased use of the PPOS protocol—especially in patients with overweight or obesity, who adopted PPOS more often in successful versus prior failed cycles. Multivariate analysis showed that PPOS adoption is independently associated with a lower risk of subsequent failure in patients without persistent high-risk factors. This aligns with prior evidence showing PPOS yields more oocytes and higher-quality embryos than mild protocols in women with advanced age or diminished ovarian reserve [55, 56]. This transition is typically phenotype-driven—not random or universal. In clinical practice, high-BMI women often present with overlapping PCOS (PMOS) features or high ovarian reserve [1, 57]. Based on over a decade of institutional experience, we prefer PPOS for these patients: it reliably prevents premature LH surges and supports a freeze-all strategy [38, 58]. Although phenotype-driven clinical preference may introduce confounding by indication, our PSM analysis—balancing key covariates (maternal age, BMI, ovarian reserve)—showed that PPOS transition is associated with better embryonic outcomes, though residual confounding cannot be excluded. However, PPOS is not universally superior: studies show no significant difference in oocyte yield or cumulative live birth rates compared with conventional GnRH antagonist protocols—in both general [59, 60] and poor-prognosis patients [61]. Moreover, PPOS requires a freeze-all strategy, delaying pregnancy and increasing financial and psychological burdens [62, 63]. Thus, while highly beneficial for our cohort, PPOS use must be individualized. Efficient hMG use—defined as total dose per mature follicle (≥ 14 mm)—is a key predictor of success in PPOS cycles. Because progesterone suppresses granulosa cell proliferation and antral follicle growth, PPOS requires higher total hMG doses than conventional protocols [64]. Women with overweight or obesity who achieved live birth used ~ 12% more total hMG—likely the minimum needed to overcome progesterone’s cytostatic effect. However, hMG dose per mature follicle is more clinically informative: higher per-follicle doses independently predict cycle failure (Table 5). In live birth cycles, average exposure was 289 IU (overweight) and 326 IU (obese); in failed cycles, it rose by 50–65% (438 IU and 535 IU). These findings suggest that while sufficient total hMG is essential to counteract progesterone, exceeding ~ 350 IU per mature follicle may impair oocyte competence. The hMG-to-follicle ratio may serve as a potential marker of diminished ovarian reserve and could be explored as a response-guided dosing tool. When this ratio exceeds ~ 350 IU/follicle, diminishing returns may occur, and further dose escalation may not improve outcomes; prospective trials are needed to confirm its utility as an actionable intervention target. Endometrial preparation protocol (Natural, HRT, or Letrozole) was not independently associated with live birth in women with overweight or obesity—confirmed by multivariable logistic regression (Table 5). This explains seemingly conflicting findings: although HRT use was higher in successful cycles (Table 4), it correlated with higher failure risk in women with overweight (Table 3). These opposing patterns reflect confounding by indication and clinical personalization. In women with overweight, HRT may disrupt the partially preserved HPO axis, impairing oocyte-endometrium synchrony (Table 3). The observed HRT use in successful cycles probably reflects clinicians’ strategic switch to HRT after prior natural-cycle failure to stabilize the implantation window (Table 4). However, once key covariates (e.g., PPOS protocol, hMG dose) are adjusted for, endometrium protocol loses predictive power (Table 5). Thus, for high-BMI patients, oocyte quality and stimulation efficiency—not endometrial preparation—may drive outcomes. No consistent effect of preparation method was found; prospective studies are needed to clarify its role. Transferring two embryos lowered live birth failure risk by 26–38% in women with overweight (aOR 0.74, P < 0.001) or obesity (aOR 0.62, P < 0.001). However, double embryo transfer (DET) carries a high multiple gestation risk [65] and is not universally recommended. Elective single embryo transfer (SET) remains the standard for maternal and neonatal safety. DET represents a pragmatic choice—favoring per-cycle success in challenging cases despite obstetric risks. Blastocyst transfer is associated with better outcomes than cleavage-stage transfer in overweight women. For patients with prior failed cycles, embryo selection and developmental stage matter more than endometrial preparation adjustments. Other potential confounders not considered: prepregnancy blood pressure (BP) Emerging evidence identifies prepregnancy blood pressure (BP) as a key modifiable factor for ART success—especially in high-BMI patients. Large-scale studies show that even systolic BP within the normal range inversely correlates with live birth rates. Elevated BP impairs oocyte competence and clinical outcomes [66, 67], and Stage 1 hypertension (≥ 130/80 mmHg) significantly increases pregnancy loss [68]. In the obese cohort of our center, hypertension was linked to markedly lower live birth rates (19.8% vs. 31.0% in normotensive patients) [1]. Yet BP is not routinely measured in IVF practice of our center [69, 70], so only a small subset had BP data—excluding it from primary analysis to avoid selection bias. Given its impact on vascular function and endometrial receptivity, pre-ART BP optimization warrants prospective study in high-BMI patients. Limitations This study has several limitations. This single-center retrospective study lacked randomization, potentially introducing selection bias and limiting causal inference and generalizability. Prospective, multicenter studies are needed to validate the findings. Second, although the cohort is large, using BMI for stratification is a crude method that does not capture metabolic heterogeneity, possibly underpowering some subgroup analyses. Future prospective studies are encouraged to incorporate more detailed measures, such as Waist-to-Hip Ratio (WHR), to more effectively elucidate the nuanced impact of abdominal obesity on ART outcomes. Third, prepregnancy blood pressure, insulin resistance, dyslipidemia, chronic inflammation and paternal BMI were either not routinely recorded or had high rates of missingness—so we excluded them from our final models, leaving potential residual confounding. Fourth, our transitional analysis was restricted to the 2,848 women who ultimately achieved a live birth. By excluding those with persistent failures, this analysis introduced a potential conditioning-on-success bias. Fifth, excluding fresh transfers and donor sperm minimized confounding but limited the generalizability of our findings and the applicability of our transition strategies to patients utilizing these approaches. Nonetheless, the large sample size strengthens the results, supports hypothesis generation, and highlights the need for prospective, randomized trials to validate BMI-stratified treatment strategies. Conclusions In conclusion, this large-scale study identifies key risk factors associated with recurrent live birth failure in women with elevated BMI. Non-modifiable factors include advanced maternal age (≥ 35 for overweight, ≥ 33 for obese) and low AFC (≤ 7 for overweight, ≤ 9 for obese). Modifiable ART-related factors are the use of GnRH antagonist protocol, excessive hMG dose per mature follicle (> 500 IU), and single-embryo transfer. For lower-risk patients—such as younger women (~ 31 years) with adequate AFC (≥ 10)—transtition strategies associated with favorable therapeutic options include using the PPOS protocol, optimizing gonadotropin dose (< 350 IU per mature follicle), and judicious, case-specific double-embryo transfer and blastocyst culture. This study is, to our knowledge, the first to systematically differentiate modifiable from non-modifiable risk factors for recurrent live birth failure in women with high BMI undergoing ART. This distinction may enable more precise live birth outcome prediction and may facilitates the development of personalized, risk-stratified interventions, particularly in lower-risk patients. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1: Additional file 1: Fig. S1. Additional file 2: Table S1 Comparison of clinical outcomes and baseline variables in the propensity score-matched cohort (Mild Stimulation vs. PPOS). Supplementary Material 1: Additional file 1: Fig. S1. Additional file 2: Table S1 Comparison of clinical outcomes and baseline variables in the propensity score-matched cohort (Mild Stimulation vs. PPOS).