Beyond metabolic comorbidity: metabolic health stratification reveals BMI–ART associations in tubal factor infertility

The baseline characteristics of the two cohorts are presented in Table S1. All baseline characteristics except metabolic indicators (BMI) showed no statistically significant differences between the two groups (age, FSH, AMH). BMI was significantly correlated with FSH ( r  = − 0.131 to − 0.123, both P  < 0.001), and both FSH and BMI were independently associated with oocyte yield in multivariable regression models (both P  < 0.001), satisfying the prerequisites for examining indirect effects (Table S1). Table  1 presents path coefficients and effect decomposition from SEM. The indirect pathway from BMI to oocyte yield via FSH was statistically significant in both cohorts (all P  < 0.01), accounting for 6.3% and 4.7% of the total association, respectively. Direct effects and total effects were positive and significant in all groups (all P   0.05). Interaction tests (BMI × cohort) further confirmed that cohort did not significantly modify the association between BMI and oocyte yield (all P  > 0.05). Table 1 Indirect pathway analysis of continuous BMI on oocyte yield via FSH Path/Effect Metabolically Normal ( N  = 1706) Metabolically Abnormal ( N  = 1781) Between-Cohort P Interaction P Est (95% CI) [ P ] Est (95% CI) [ P ] log_FSH ~ log_BMI (a) -0.123 (-0.201, -0.045) [0.002] -0.132 (-0.205, -0.059) [< 0.001] log_Oocytes ~ log_FSH (b) -0.167 (-0.230, -0.105) [< 0.001] -0.110 (-0.172, -0.048) [< 0.001] Indirect effect (a*b) 0.021 (0.006, 0.036) [0.007] 0.014 (0.003, 0.026) [0.013] 0.525 Direct effect (c’) 0.315 (0.194, 0.435) [< 0.001] 0.285 (0.169, 0.401) [< 0.001] 0.728 0.725 Total effect 0.335 (0.213, 0.457) [< 0.001] 0.299 (0.183, 0.415) [< 0.001] 0.677 0.552 Model fit indices (by cohort): Metabolically normal: χ²/df = 32.8/6, CFI = 0.986, TLI = 0.955, RMSEA = 0.051, SRMR = 0.014 Metabolically abnormal: χ²/df = 16.6/6, CFI = 0.995, TLI = 0.983, RMSEA = 0.032, SRMR = 0.010 Est = standardized coefficient. a, b, and c’ represent path coefficients: a = effect of BMI on FSH; b = effect of FSH on oocyte yield adjusting for BMI; c’ = direct effect of BMI on oocyte yield adjusting for FSH. Indirect effect (a×b) represents the indirect effect via FSH. Proportion of total association accounted for by the indirect pathway: 6.3% in metabolically normal cohort (0.021/0.335), 4.7% in metabolically abnormal cohort (0.014/0.299). Cohort interaction (BMI × cohort) was tested only for direct and total effects. The sample sizes for mediation analysis (metabolically healthy: n  = 1,706; metabolically unhealthy: n  = 1,781) are smaller than the total cohort ( n  = 1,849 and n  = 1,921, respectively) because mediation analysis requires complete data for all variables (BMI, FSH, oocyte yield, and all covariates). Cases with any missing values on these variables were excluded listwise Indirect pathway analysis of continuous BMI on oocyte yield via FSH Model fit indices (by cohort): Metabolically normal: χ²/df = 32.8/6, CFI = 0.986, TLI = 0.955, RMSEA = 0.051, SRMR = 0.014 Metabolically abnormal: χ²/df = 16.6/6, CFI = 0.995, TLI = 0.983, RMSEA = 0.032, SRMR = 0.010 Est = standardized coefficient. a, b, and c’ represent path coefficients: a = effect of BMI on FSH; b = effect of FSH on oocyte yield adjusting for BMI; c’ = direct effect of BMI on oocyte yield adjusting for FSH. Indirect effect (a×b) represents the indirect effect via FSH. Proportion of total association accounted for by the indirect pathway: 6.3% in metabolically normal cohort (0.021/0.335), 4.7% in metabolically abnormal cohort (0.014/0.299). Cohort interaction (BMI × cohort) was tested only for direct and total effects. The sample sizes for mediation analysis (metabolically healthy: n  = 1,706; metabolically unhealthy: n  = 1,781) are smaller than the total cohort ( n  = 1,849 and n  = 1,921, respectively) because mediation analysis requires complete data for all variables (BMI, FSH, oocyte yield, and all covariates). Cases with any missing values on these variables were excluded listwise Pathway analyses using the counterfactual framework and the product-of-coefficients method yielded results consistent with the SEM analysis (Table S2). When BMI was analyzed as a categorical variable, oocyte yield showed a progressively increasing trend from underweight to overweight/obese (Table S3), consistent with the dose–response relationship observed in the continuous analysis. No significant differences were detected between cohorts, confirming the consistency of this indirect pathway across metabolic status. The study included 2,981 IVF cycles and 307 ICSI cycles. After confirming no significant interactions between BMI and cohort for any outcome (all interaction P  > 0.05), data from the two cohorts were pooled for subsequent analyses. In ICSI cycles, underweight exhibited significantly higher rates of Day 3 good-quality embryo rate (+ 11.4%, adjusted P  < 0.01) and good-quality blastocyst formation rate (+ 5.5%, adjusted P  = 1.00). No significant differences were observed in IVF cycles (Table  2 ). Table 2 Effect of BMI as a categorical variable on embryo development outcomes Indicator (Method) BMI Cycles / Rate ( n / N ) 1 OR (95% CI) [ P /Pcorr] 2 RDadj (%) 3 MII oocyte rate (ICSI) Normal 198 / 73.8% (2161/2928) 1.000 (ref) 0.0% UW 28 / 74.2% (299/403) 1.071 (0.833–1.378) [0.59 / 1.00] + 1.3% OW/OB 81 / 70.4% (796/1130) 0.816 (0.693–0.960) [0.01 / 0.10] -4.1% 2PN rate (IVF) Normal 1983 / 63.1% (16671/26428) 1.000 (ref) 0.0% UW 297 / 64.4% (2636/4093) 1.092 (1.018–1.172) [0.01 / 0.10] + 2.0% OW/OB 701 / 60.5% (5538/9149) 0.866 (0.823–0.912) [< 0.01 / <0.01] -3.4% 2PN rate (ICSI) Normal 198 / 79.1% (1710/2161) 1.000 (ref) 0.0% UW 28 / 79.3% (237/299) 1.004 (0.734–1.374) [0.98 / 1.00] + 0.1% OW/OB 81 / 74.6% (594/796) 0.805 (0.656–0.989) [0.04 / 0.27] -3.9% D3QE rate (IVF) Normal 1976 / 27.9% (4644/16671) 1.000 (ref) 0.0% UW 297 / 28.0% (739/2636) 1.018 (0.927–1.118) [0.71 / 1.00] + 0.4% OW/OB 697 / 27.3% (1510/5538) 0.972 (0.905–1.044) [0.44 / 1.00] -0.6% D3QE rate (ICSI) Normal 197 / 22.8% (390/1710) 1.000 (ref) 0.0% UW 28 / 35.0% (83/237) 1.759 (1.295–2.390) [<0.01 / <0.01] + 11.4% OW/OB 81 / 22.6% (134/594) 0.978 (0.770–1.242) [0.85 / 1.00] -0.4% QB rate (IVF) Normal 603 / 65.7% (3038/4624) 1.000 (ref) 0.0% UW 85 / 64.8% (486/750) 1.010 (0.854–1.195) [0.91 / 1.00] + 0.2% OW/OB 237 / 63.3% (1214/1919) 0.857 (0.761–0.966) [0.01 / 0.08] -3.5% QB rate (ICSI) Normal 77 / 61.4% (404/658) 1.000 (ref) 0.0% UW 13 / 64.9% (74/114) 1.292 (0.820–2.033) [0.27 / 1.00] + 5.5% OW/OB 34 / 58.5% (121/207) 0.864 (0.605–1.234) [0.42 / 1.00] -3.4% 1 Cycles and event rates (n/N with %) derived from pooled cohort 2 ORs adjusted for covariates listed in methods; raw / Bonferroni-corrected P (threshold α = 0.05/7 ≈ 0.0071) shown in brackets 3 RDadj: adjusted rate difference (%) standardized to metabolically normal, normal-weight women Abbreviations : D3QE rate Day 3 good-quality embryo rate, QB rate good-quality blastocyst formation rate, Ref. reference, UW  underweight, OW/OB  overweight/obese Effect of BMI as a categorical variable on embryo development outcomes 1 Cycles and event rates (n/N with %) derived from pooled cohort 2 ORs adjusted for covariates listed in methods; raw / Bonferroni-corrected P (threshold α = 0.05/7 ≈ 0.0071) shown in brackets 3 RDadj: adjusted rate difference (%) standardized to metabolically normal, normal-weight women Abbreviations : D3QE rate Day 3 good-quality embryo rate, QB rate good-quality blastocyst formation rate, Ref. reference, UW  underweight, OW/OB  overweight/obese Overweight/obesity was associated with lower 2PN rate in both IVF (-3.4%, adjusted P  < 0.01) and ICSI cycles (-3.9%, adjusted P  = 0.27), reduced good blastocyst formation rate in IVF (-3.5%, adjusted P  = 0.08) and ICSI (-3.4%, adjusted P  = 1.00), and lower MII oocyte rate in ICSI (-4.1%, adjusted P  = 0.10) (Table  2 ). Continuous analysis further supported the validity of cohort pooling (all interaction P  > 0.05). Increasing BMI was significantly associated with lower 2PN rate in IVF cycles (OR = 0.498, 95% CI: 0.421–0.588, adjusted P  < 0.01) and ICSI cycles (OR = 0.397, 95% CI: 0.205–0.771, adjusted P  = 0.04), as well as lower MII oocyte rate in ICSI cycles (OR = 0.436, 95% CI: 0.260–0.731, adjusted P  = 0.01) (Table S4). Most analyses were adequately powered (> 80%); underpowered comparisons (Day 3 good-quality embryo rate and good blastocyst formation rate) may require larger sample sizes for definitive confirmation (Table S4). The analysis included 2,530 cycles for clinical outcomes. No significant interactions were detected between BMI and cohort for any outcome (all interaction P  > 0.05), supporting pooling of cohorts for subsequent analyses. No associations remained significant after correction for multiple comparisons. However, a clinically meaningful effect size was noted for the lean group, which exhibited a lower miscarriage rate (-3.5%, adjusted P  = 1.00). No other clinically meaningful effects were observed (all absolute rate differences ≤ 2.5%) (Table  3 ). Continuous analysis further supported cohort pooling (all interaction P  > 0.05). Log-BMI showed no statistically significant associations with any outcomes (all p  > 0.05). Post-hoc power analysis revealed limited statistical power for detecting modest effects (range: 23.5%–74.2%). Combining power and effect size analyses, the observed effect sizes for clinical pregnancy and live birth were small (absolute rate differences ≤ 2.5%), indicating limited clinical relevance. In contrast, the effect size for miscarriage (absolute rate difference: -3.5%) was clinically meaningful but underpowered, warranting further investigation in larger cohorts (Table S5). Table 3 Effect of BMI as a categorical variable on clinical outcomes Outcome BMI Group Rate % ( n / N )¹ OR (95% CI) [ P / Pcorr]² RDadj (%)³ Clinical pregnancy rate Normal 67.2% (1096/1631) 1.000 (ref.) 0.0 UW 65.7% (140/213) 1.27 (0.87–1.85) [0.223 / 1.000] -0.9 OW/OB 68.4% (469/686) 0.81 (0.59–1.12) [0.206 / 1.000] + 2.5 Miscarriage rate Normal 11.4% (119/1047) 1.000 (ref.) 0.0 UW 5.3% (7/131) 0.62 (0.26–1.53) [0.302 / 1.000] -3.5 OW/OB 13.6% (60/441) 0.73 (0.41–1.28) [0.266 / 1.000] -0.4 Live birth rate Normal 58.7% (928/1582) 1.000 (ref.) 0.0 UW 60.8% (124/204) 1.24 (0.86–1.80) [0.249 / 1.000] + 1.4 OW/OB 57.9% (381/658) 0.92 (0.67–1.26) [0.596 / 1.000] + 2.3 Bonferroni correction (3 outcomes × 2 BMI comparisons = 6 tests, threshold α = 0.05/6 ≈ 0.0083). other Notes are the same as in Table  2 Effect of BMI as a categorical variable on clinical outcomes Bonferroni correction (3 outcomes × 2 BMI comparisons = 6 tests, threshold α = 0.05/6 ≈ 0.0083). other Notes are the same as in Table  2

Approval for this retrospective study was obtained from the Reproductive Medicine Ethics Committee of Jiangxi Maternal and Child Health Hospital (Approval No. SZYY-202602), which also waived the informed consent requirement. The study included women undergoing their first IVF/ICSI cycle after controlled ovarian stimulation at the same center between January 2009 and March 2026. All first IVF/ICSI cycles with oocyte retrieval yield > 0, without donor gametes, preimplantation genetic testing, azoospermia, chromosomal abnormalities, or natural cycles were initially included. Cycles were excluded if any of the following were present: (1) established metabolic or endocrine disorders that could independently affect ART outcomes or ovarian function, including diagnosed type 1 or type 2 diabetes requiring pharmacological treatment, severe dyslipidemia requiring ongoing statin or fibrate therapy, uncontrolled thyroid disorders, hyperprolactinemia, Cushing’s syndrome, congenital adrenal hyperplasia, androgen-secreting tumors, or severe hepatic/renal/pituitary/adrenal dysfunction. Use of metformin, oral contraceptives, or other metabolic-modifying agents within 3 months was also excluded. Mild, untreated metabolic abnormalities (e.g., borderline dyslipidemia, pre-diabetes, or stage 1 hypertension without medication) were not exclusion criteria, as these were used for subsequent metabolic subgrouping; (2) reproductive system abnormalities (severe male factor requiring donor sperm, uterine structural abnormalities, history of gestational trophoblastic disease, stage III-IV endometriosis, history of ovarian surgery, endometrioma, or pelvic radiotherapy/chemotherapy); or (3) history of recurrent miscarriage or recurrent pregnancy loss. From the remaining cycles, we selected women with tubal factor infertility confirmed by hysterosalpingography or laparoscopy, who also had normal ovarian reserve (AMH ≥ 1.1 ng/mL, FSH  7). Women with PCOS, endometriosis (any stage), or any indication of diminished reserve (age > 37 years, AMH < 1.1, FSH ≥ 10, AFC ≤ 7) were excluded at this stage. The final tubal factor cohort was divided into two subgroups based on metabolic status: metabolically healthy (met none of the criteria below) and metabolically unhealthy (met at least one). The metabolic abnormality criteria used only among the retained participants (i.e., after excluding those with treated or severe metabolic diseases) were defined as the presence of any of the following mild, untreated metabolic abnormalities, measured at baseline: (1) Elevated blood pressure: systolic BP ≥ 130 mmHg or diastolic BP ≥ 85 mmHg, without antihypertensive medication; (2) Dysglycemia: fasting glucose ≥ 5.6 mmol/L (100 mg/dL) or HbA1c ≥ 5.7%, without glucose-lowering medication or insulin; (3) Low HDL-C: 3.89 mmol/L (150 mg/dL), without lipid-lowering therapy. Participants meeting none of these criteria were classified as metabolically healthy; those meeting at least one were classified as metabolically unhealthy. This definition is consistent with established metabolic health phenotyping in the literature [ 12 ] and is deliberately restricted to untreated, subclinical metabolic derangements to avoid overlap with exclusion criteria. The stepwise selection process is summarized in Fig. 1 . After applying the above criteria, the final study cohort comprised 3,770 women: 1,849 metabolically healthy and 1,921 metabolically unhealthy. Fig. 1 Flow diagram of patient selection Flow diagram of patient selection Ovarian stimulation was individualized based on age, BMI, and ovarian reserve. Protocols included GnRH agonist (long, ultra-long, or short), GnRH antagonist, and other regimens (e.g., mild stimulation or progestin-primed ovarian stimulation, PPOS). Trigger criteria were at least 2–3 follicles ≥ 18 mm, with trigger methods including hCG, GnRH agonist, or both. Oocyte retrieval was performed 34–36 h post-trigger. IVF or ICSI was performed according to standard indications. Embryos were cultured to day 3 or day 5/6, and the number of embryos transferred was determined based on embryo quality, age, and previous cycle history. Outcomes were classified as embryological outcomes and clinical outcomes. Oocyte yield (total oocytes retrieved); MII oocyte rate (MII/retrieved oocytes, only in ICSI cycles); 2PN rate (2PN/MII oocytes for ICSI, or 2PN/retrieved oocytes for IVF); day 3 good-quality embryo rate (good-quality embryos/2PN zygotes), where good-quality embryos were defined as 7–9 cells, grade 1–2, and symmetric; and good-quality blastocyst formation rate (good-quality blastocysts/blastocysts cultured), analyzed only in cycles with complete blastocyst culture (no day 3 transfer or cryopreservation). Good-quality blastocysts were defined as those eligible for transfer or cryopreservation (≥ 3BC or ≥3CB). Clinical pregnancy (gestational sac at 4–5 weeks post-transfer), miscarriage (pregnancy loss before 28 weeks), and live birth (delivery of at least one live infant after 28 weeks). Mediating variable: basal FSH, measured on day 2–3 of the menstrual cycle. Continuous variables were non-normally distributed (Shapiro–Wilk test, all P  < 0.05) and thus presented as medians (IQR) with comparisons by Wilcoxon rank-sum test. Pearson correlation assessed BMI–FSH relationships. Multivariable linear regression models assessed associations of log-transformed BMI and log-transformed baseline FSH with log-transformed oocyte yield, adjusting for age, log-transformed AMH, log-transformed baseline LH, infertility type (primary/secondary), log-transformed infertility duration, treatment protocol (agonist/antagonist/special), and log-transformed hCG day E 2 . Before conducting mediation analysis, restricted cubic spline regression was employed to examine the nonlinear relationship between BMI and oocyte yield. The results supported the linearity assumption, justifying the use of linear regression models in the mediation framework. To explore whether baseline FSH is involved in the association between BMI and oocyte yield, cohort-specific pathway analyses employed three complementary approaches to ensure robustness: (1) product-of-coefficients method with bias-corrected bootstrap resampling (5,000 iterations); (2) structural equation modeling (SEM) to assess overall model fit; and (3) counterfactual-based approaches to further examine the indirect pathway. Wald tests compared effect estimates between cohorts, and likelihood ratio tests assessed cohort interactions. To complement the continuous analysis and explore nonlinear relationships, BMI was categorized per Chinese criteria: underweight (< 18.5 kg/m²), normal weight (18.5–23.9 kg/m², reference), and overweight/obese (≥ 24.0 kg/m²). For embryological outcomes (MII oocyte rate, 2PN rate, day 3 good-quality embryo rate, and good-quality blastocyst formation rate), binomial logistic regression models with a cbind(successes, failures) structure were used, where the numerator and denominator represented event counts and total oocytes/embryos at risk, respectively. Robust standard errors were applied to account for potential within-cycle clustering. Multivariable logistic regression models evaluated associations between BMI categories and two sets of outcomes: (1) embryonic outcomes; and (2) clinical outcomes. BMI was categorized according to the Chinese criteria described above. All models adjusted for age, infertility type, log-transformed infertility duration, treatment protocol, log-transformed AMH, log-transformed total Gn dose, and log-transformed E 2 and P levels on hCG day. Clinical outcomes models additionally adjusted for embryological and transfer factors, including insemination method (IVF/other), 2PN embryos count, D3 good-quality embryos count, embryo type (blastocyst/cleavage), transfer type (single/multiple), and number of good-quality embryos transferred. To determine whether the effect of BMI on outcomes differed by metabolic status, we tested for interactions between BMI category (or log-BMI for continuous analysis) and cohort using likelihood ratio tests comparing models with and without the interaction term. All interaction P values were > 0.05 for both categorical and continuous analyses, indicating no evidence of interaction. Thus, data from both cohorts were pooled for the main analyses, with metabolic status included as a covariate. For categorical BMI analysis, associations were estimated using the above models, and adjusted absolute rate differences were calculated via marginal standardization. For supplementary analysis, log-transformed BMI was included as a continuous predictor to assess dose–response and validate categorical findings. Post-hoc power analyses were conducted under the conservative assumption of a model R² of 0.1 to distinguish between underpowered findings and truly negligible effects. All statistical analyses were performed using R software (version 4.3.2). A two-tailed P  < 0.05 was considered statistically significant unless otherwise specified for multiple comparisons.

In this study, we introduced metabolic health stratification into reproductive medicine and systematically evaluated BMI–ART associations in a homogeneous population without ovarian reserve impairment or reproductive endocrinopathies. Key findings were as follows: no significant interactions were detected between BMI and metabolic status for any outcome, suggesting that the impact of BMI on ART outcomes may not be substantially modified by metabolic health in this population. First, BMI was positively associated with oocyte yield, and the indirect pathway through FSH accounted for 4.7%–6.3% of this association. Second, overweight/obesity was associated with lower MII oocyte rate, fertilization rate, and blastocyst quality. Third, after rigorous adjustment, BMI was not independently associated with clinical pregnancy or live birth, whereas a clinically meaningful reduction in miscarriage rate was observed in underweight women. Previous studies on BMI and oocyte yield have yielded inconsistent results [ 13 , 14 ]. Some found no differences across BMI categories in unselected populations using PPOS or antagonist protocols [ 15 , 16 ], while others reported a positive association in both PCOS and non-PCOS patients treated with antagonist/agonist protocols [ 17 ]. These discrepancies likely reflect variations in treatment protocols and patient populations, both of which are closely tied to endocrine status and ovarian reserve. This highlights a key strength of our study: by excluding patients with conditions affecting ovarian reserve or confounding endocrinopathies, we isolated the effects of metabolic status and BMI. Moreover, previous studies were largely observational and did not explore mechanisms. In contrast, our pathway analysis suggested a modest but consistent indirect pathway from BMI to oocyte yield through basal FSH—a pre-treatment pathway independent of stimulation protocols—highlighting the potential role of baseline endocrine status. Unlike conventional regression, our SEM approach demonstrated excellent model fit, and results were consistent across three complementary analytical approaches, supporting the robustness of our findings. Studies on BMI and embryonic outcomes have yielded conflicting results, likely due to heterogeneity in study populations. Some reported comparable blastocyst formation and fertilization rates across BMI categories [ 18 , 19 ], while others observed lower 2PN and good-quality embryo rates in higher BMI groups [ 20 , 21 ]; still others found higher usable embryo and blastocyst rates [ 3 ]. In PCOS populations, higher BMI was associated with lower fertilization and good-quality embryo rates [ 22 ]. By rigorously excluding conditions that may confound this association—including polycystic ovary syndrome, diminished ovarian reserve, and other reproductive endocrinopathies—our study enables a more isolated assessment of the independent effect of BMI on embryonic outcomes. The association between BMI and fresh embryo transfer outcomes remains controversial. A large study ( n = 30,746, excluding PCOS) reported no differences across BMI categories [ 1 ]. Another study of 5,981 women with poor prognosis found comparable pregnancy rates in obese women but increased miscarriage and reduced live birth rates [ 23 ]. A PCOS study ( n = 1,704) observed no outcome differences [ 24 ]. Notably, none of these studies incorporated power or effect size analyses [ 25 ]. In contrast, our study ( n = 2,530) integrated logistic regression with power and effect size analyses, enabling a more comprehensive interpretation of both statistical and clinical significance. For clinical pregnancy and live birth, the effect sizes were small, suggesting limited clinical relevance even if statistical significance were achieved. For miscarriage, although not statistically significant after correction, the effect size (absolute rate difference: -3.5%) indicates a clinically meaningful protective effect associated with underweight status, warranting further investigation. Importantly, no significant interactions were detected between BMI category and metabolic status for any outcome, suggesting that the detrimental effects of elevated BMI on ART are primarily driven by adiposity per se rather than by its associated metabolic comorbidities. This study has several methodological strengths. Incorporating metabolic health stratification allowed us to disentangle BMI effects from associated metabolic factors; the homogeneous study population (tubal factor infertility) minimized confounding by common reproductive endocrinopathies; and rigorous covariate adjustment, multiple testing correction, and post-hoc power analyses reduced the risks of type I and type II errors. This is also the first study to explore FSH as a potential intermediate variable in the BMI–oocyte yield association, using three complementary analytical approaches. Several limitations should be acknowledged, including the retrospective design, which precludes causal inference for any of the observed associations—including the indirect pathway findings. Therefore, these results should be considered hypothesis-generating rather than conclusive evidence of a causal pathway. Additional limitations include the lack of assessment of dynamic BMI changes (e.g., weight loss interventions) and the single-center setting, which may limit generalizability to populations with different ethnic compositions or clinical practices. Additionally, because clinical outcomes were restricted to fresh embryo transfer cycles, cycles with freeze-all or no embryo transfer were excluded, potentially introducing selection bias and limiting generalizability to broader ART populations.

Body mass index (BMI) is a key metabolic indicator associated with assisted reproductive technology (ART) outcomes. Obese women generally have poorer outcomes [ 1 , 2 ], yet findings across studies remain inconsistent [ 3 ]. This heterogeneity suggests that the effect of BMI may be modified by other factors [ 4 , 5 ], including the patient’s metabolic status (e.g., dyslipidemia, hypertension) and the underlying infertility etiology (e.g., polycystic ovary syndrome [PCOS], diminished ovarian reserve [DOR]) [ 6 ]. In clinical practice, these two factors often intertwine, making it difficult to disentangle their independent effects. Normal weight does not necessarily imply metabolic health; indeed, evidence supports the existence of metabolically unhealthy normal-weight individuals and metabolically healthy obese phenotypes [ 7 , 8 ]. Metabolic health stratification has been widely adopted in cardiovascular and metabolic disease research [ 9 , 10 ]. However, its application in reproductive medicine remains limited [ 11 ]. Consequently, whether and how the association between BMI and ART outcomes differs between metabolically healthy and unhealthy individuals remains largely unknown. To address these gaps, we introduce metabolic health stratification into reproductive medicine to examine whether the impact of BMI on ART outcomes differs by metabolic status. Focusing on women with tubal factor infertility—a homogeneous population with preserved ovarian reserve and no confounding reproductive endocrinopathies—we systematically evaluated the associations between BMI and outcomes spanning oocyte yield, embryogenesis, and clinical pregnancy. Furthermore, we explored whether follicle-stimulating hormone is involved in the BMI–oocyte yield association, an indirect pathway not previously examined in metabolically stratified populations. This study represents the first integrated analysis of metabolic stratification and pathway exploration to determine whether metabolic status modifies the reproductive consequences of adiposity.

Supplementary Material 1. Supplementary Material 1.