Ovarian reserve integrated damage (ORI) model for predicting ovarian function recovery after laparoscopic ovarian cystectomy for endometriosis: a randomized controlled trial

This prospective, randomized controlled trial was conducted at Chengdu First People’s Hospital from September 2023 to February 2025, with follow-up completed by August 2025. The study design, conduct, and reporting adhered to CONSORT 2010 guidelines for randomized trials and TRIPOD guidelines for prediction model development. Complete CONSORT 2010 and TRIPOD checklists are provided in Supplementary Appendix (Section J). Women aged 18–45 years with endometriosis-associated ovarian cysts scheduled for cystectomy were eligible for inclusion. Key exclusion criteria included previous ovarian surgery, history of chemotherapy or pelvic radiation, current hormonal therapy, and suspected malignant ovarian pathology. This study was conducted with approval from the Institutional Review Board of Chengdu first people’s hospital (IRB ID: 2022.KT.039), and all participants provided written informed consent. This trial was retrospectively registered in the Chinese Clinical Trial Registry (ChiCTR) under the registration number ChiCTR2500102270 and conducted in accordance with the Declaration of Helsinki. The sample size was calculated for the RCT component based on the primary endpoint of AMH decline > 30% at 6 months. Assuming a 15% event rate in the control group and 5% in the intervention group, with 80% power and α = 0.05, 90 participants were required. We targeted 120 participants to account for 25% attrition. The final analysis included 101 participants, which exceeded the minimum required sample size. Participants were randomly assigned 1:1:1 to receive one of three hemostatic methods using computer-generated block randomization: (1) oxidised cellulose polymer (Surgicel Original, Ethicon), (2) bipolar coagulation (power setting 30 W), or (3) suturing (2 –0 V-Loc™, Covidien™). All surgical procedures were performed by experienced gynecologic surgeons using standardized techniques. After cyst removal, hemostasis was achieved according to the assigned method. The primary surgeon was blinded to the randomization sequence until after cyst removal. Participants were randomized using a computer-generated sequence with block randomization (block size = 6), stratified by age (< 35 vs. ≥ 35 years) and maximum cyst diameter (< 5 cm vs. ≥ 5 cm). Allocation was concealed using a centralized web-based randomization system. Blinding was implemented for outcome assessors: AMH laboratory personnel and sonographers were blinded to treatment allocation. Participants and surgeons could not be blinded to the surgical intervention. Following anesthesia induction, patients were placed in the supine position with Trendelenburg tilt (head down and hips elevated). A Hangt Port (Beijing Hangtian Kadi) was inserted through the umbilicus. For cases where the ovarian cyst was adherent to surrounding tissues, it was first mobilized. The ovarian cortex was then incised longitudinally using monopolar scissors along the antimesenteric border. The cleavage plane between the ovarian cortex and cyst wall was identified, and the cyst wall was carefully dissected away from the ovarian parenchyma for complete cyst excision.No hemostatic measures were applied during the cyst dissection phase. After completing cyst enucleation, different hemostatic techniques were applied according to group assignment: Bipolar coagulation group: hemostasis was achieved using bipolar forceps at 30 W power, applied minimally to preserve ovarian tissue. If inadequate hemostasis occurred, laparoscopic suturing was performed as rescue therapy. Oxidized cellulose polymer group: a single piece of oxidized regenerated cellulose (2.5 × 5.1 cm, Ethicon) was placed on the bleeding ovarian defect and compressed for 5–10 min. Suturing was implemented if hemostasis failed within 10 min. Suture group: continuous suturing of residual ovarian tissue was performed using 2 –0 V-Loc™ (Covidien™) absorbable barbed suture. All procedures were performed exclusively for ovarian endometriomas, with no cases involving broad ligament endometriosis excision included in this study. Detailed surgical methods and equipment specifications are provided in Supplementary Appendix (Section N). Primary outcomes included changes in ovarian reserve markers at 6 months postoperatively: - Anti-Müllerian hormone (AMH) levels (ng/ml) - Antral follicle count (AFC) - Peak systolic velocity (PSV) (cm/s). Secondary outcomes included operative time, estimated blood loss, complication rates, and postoperative recovery parameters. Ovarian reserve markers were assessed preoperatively, at 1 month, and at 6 months postoperatively. AMH levels were measured using electrochemiluminescence immunoassay (Cobas e601, Roche Diagnostics) with a detection limit of 0.01 ng/ml. AFC was assessed by transvaginal ultrasound (Voluson E10, GE Healthcare) during the early follicular phase (days 2–5) by experienced sonographers blinded to treatment allocation. PSV measurements were obtained using color Doppler ultrasound with standardized settings. We developed two logistic regression–based prediction models to estimate the risk of poor ovarian recovery (> 30% decline in AMH at 6 months) after laparoscopic ovarian cystectomy: ORI-PreOp (preoperative predictors only) and ORI-PeriOp (preoperative plus intraoperative predictors). Candidate predictors were prespecified from clinical evidence, including age, baseline AMH, AFC, PSV, cyst diameter, surgical method, blood loss, and operative time. Models were derived using Firth’s penalized logistic regression to address the low event rate (4.0%). Internal validation used 1000 bootstrap samples and 5-fold cross-validation, with performance evaluated by discrimination (AUC), calibration, and decision curve analysis. Missing data (< 5%) were addressed by multiple imputation. A total of 101 patients with complete follow-up provided adequate power for model development. Analyses were conducted in R (logistf package) and Python (scikit-learn). For clinical interpretation, patients were stratified into low ( 30%) risk categories, with a prespecified decision threshold of 15%. Complete model specification including all coefficients, intercepts, and performance metrics is provided in Supplementary Appendix (Sections A-F). Sample size estimation suggested that approximately 30 patients per group would be sufficient to detect a 15% difference in AMH decline rates between groups with 80% power at a two-sided α = 0.05. In practice, 38, 31, and 37 patients were included in the three study groups, slightly exceeding the initial estimation. Continuous variables were tested for normality using the Shapiro–Wilk test. Normally distributed data are presented as mean ± SD and compared using one-way ANOVA with post hoc Tukey test. Non-normally distributed data are presented as median (IQR) and compared using the Kruskal–Wallis test with post hoc Mann–Whitney U tests and Bonferroni correction. Normality of continuous variables was assessed using the Shapiro-Wilk test. Normally distributed data are presented as mean ± standard deviation and compared using one-way analysis of variance (ANOVA) with Tukey post-hoc test. Non-normally distributed data are presented as median (interquartile range) and compared using Kruskal-Wallis test with Mann-Whitney U tests (Bonferroni corrected). Categorical variables are presented as count (percentage) and compared using Chi-square test or Fisher’s exact test, as appropriate. The primary outcome was defined as an AMH decline of more than 30% at 6 months, which occurred in 4 of 101 patients (4.0% event rate). Given the low event rate, model performance was evaluated using both receiver operating characteristic area under the curve (AUC-ROC) and precision-recall area under the curve (PR-AUC). PR-AUC was selected as the primary performance metric because it directly reflects the model’s ability to identify positive cases and is particularly suitable for low-event-rate clinical prediction models. A PR-AUC value exceeding 0.30 was considered to indicate excellent clinical utility in low-event-rate scenarios. Missing data (< 5%) were addressed by multiple imputation with chained equations. Statistical analyses were performed using SPSS version 26.0 (IBM Corp., Armonk, NY, USA) for descriptive statistics and group comparisons. Predictive modeling and advanced analyses were conducted using R software (version 4.3.2, R Foundation for Statistical Computing, Vienna, Austria) with the ‘glmnet’ package for penalized logistic regression. A two-sided P-value < 0.05 was considered statistically significant. All continuous variables are reported as mean ± standard deviation or median (interquartile range). Categorical variables are presented as frequency (percentage). 95% confidence intervals were calculated for primary outcome measures. The statistical significance level was set at α = 0.05 (two-sided). The primary analysis was a per-protocol analysis including only participants with pathologically confirmed endometriosis who completed the study protocol. No data were missing for the primary outcome due to 100% follow-up completion.

A total of 185 patients were assessed for eligibility between September 2023 and August 2025. Of these, 12 patients were excluded during screening (6.5%) due to not meeting inclusion criteria ( n  = 5), logistical reasons ( n  = 4), lost to follow-up ( n  = 2), and other medical conditions ( n  = 1). A total of 173 patients were randomized to receive oxidised cellulose polymer ( n  = 60), bipolar coagulation ( n  = 56), or suture ( n  = 57) for ovarian cystectomy hemostasis. Post-enrollment, 63 patients were excluded due to non-endometriosis diagnosis (36.4%). The remaining 110 patients with confirmed endometriosis were included in the analysis, with 101 patients completing the 6-month follow-up period (Fig.  1 ). Baseline characteristics were well-balanced across treatment groups (Table  1 ). The mean age was 32.4 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 5.6 years, and all patients had confirmed endometriosis. No significant differences were observed in baseline ovarian reserve markers except for AFC ( P =0.025) and maximum cyst diameter ( P  = 0.036) between groups. Table 1 Baseline characteristics and operative outcomes Characteristic Oxidised ( n  = 37) Bipolar ( n  = 33) Suture ( n  = 31) P value Demographics  Age, years 32.0 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 5.5 32.7 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 6.9 33.6 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 4.1 0.369b  BMI, kg/m² 22.5 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 1.4 22.5 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 2.1 22.8 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 1.7 0.736a Cyst characteristics  Left location 14 (38%) 12 (36%) 10 (32%) 0.887c  Diameter, mm 59.6 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 5.4 62.7 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 5.1 61.8 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 4.7 0.036a Baseline ovarian reserve  AMH, ng/ml 3.75 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 1.03 3.90 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 1.23 3.58 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 1.37 0.574a  AFC, count 8.9 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 1.1 8.2 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 1.0 8.4 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 1.4 0.025b  PSV, m/s 0.15 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 0.02 0.13 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 0.03 0.15 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 0.04 0.176a Operative outcomes  Operation time, min 31.0 (25–36) 33.0 (27–37) 36.0 (33–42) 0.005b  Blood loss, ml 34.7 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 7.1 35.0 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 15.1 37.9 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 11.7 0.473a  Hemoglobin drop, g/L 14.0 (8–21) 20.0 (14–25) 27.0 (18–37) 0.000b  Hospital stay, days 3.0 (3–4) 3.0 (3–4) 4.0 (4–5) 0.001b Data are presented as mean ± standard deviation, median (interquartile ranges) or frequencies. Data are presented as medians (interquartile ranges). a: ANOVA (for normally distributed continuous variables), b: Kruskal-Wallis test (for non-normally distributed continuous variables), c: Chi-square test (for categorical variables). Detailed baseline characteristics table is provided in Supplementary Appendix Baseline characteristics and operative outcomes Data are presented as mean ± standard deviation, median (interquartile ranges) or frequencies. Data are presented as medians (interquartile ranges). a: ANOVA (for normally distributed continuous variables), b: Kruskal-Wallis test (for non-normally distributed continuous variables), c: Chi-square test (for categorical variables). Detailed baseline characteristics table is provided in Supplementary Appendix Detailed participant flow with specific reasons for exclusion at each step is provided in Supplementary Appendix (Section J). Complete baseline characteristics by treatment group are provided in Supplementary Appendix (Section K). Significant differences in AMH decline rates were observed between treatment groups at both 1 and 6 months (Fig.  2 ). At 6 months, mean AMH decline was 5.973.1% for oxidised cellulose polymer, 19.067.6% for bipolar coagulation, and 7.225.2% for suture (P<0.001) (Table 2 ). The AMH decline rate was 5.97±3.1% (95% CI: 4.9–7.1%) in the oxidised cellulose polymer group, 19.06±7.6% (95% CI: 16.4–21.8%) in the bipolar coagulation group, and 7.22±5.2% (95% CI: 5.3–9.1%) in the suture group. The difference between groups was statistically significant (F=52.3, P<0.001). Table 2 Ovarian reserve markers and decline rates Characteristic Oxidised ( n  = 37) Bipolar ( n  = 33) Suture ( n  = 31) P value AMH levels (ng/ml)  Baseline 3.75 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 1.03 3.90 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 1.23 3.58 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 1.37 0.574a  6 months 3.50 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 0.94 3.09 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 0.88 3.30 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 1.24 0.260a  Decline (%) 5.97 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 3.1 19.06 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 7.6 7.22 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 5.2 0.000b AFC levels  Baseline 8.9 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 1.1 8.2 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 1.0 8.4 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 1.4 0.025b  6 months 7.6 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 1.1 5.7 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 1.1 6.9 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 1.5 0.000b  Decline (%) 14.0 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 8.7 30.7 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 11.0 17.5 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 11.6 0.000b PSV levels (cm/s)  Baseline 0.15 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 0.02 0.13 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 0.03 0.15 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 0.04 0.176a  6 months 0.14 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 0.02 0.11 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 0.03 0.14 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 0.04 0.000a  Decline (%) 5.3 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 1.2 20.0 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 2.5 6.8 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 1.7 0.000b  Preservation (%) 94.7 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 1.2 80.0 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 2.5 93.2 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 1.7 0.000b Data are presented as mean ± standard deviation. a: ANOVA (for normally distributed continuous variables), b: Kruskal-Wallis test (for non-normally distributed continuous variables) Ovarian reserve markers and decline rates Data are presented as mean ± standard deviation. a: ANOVA (for normally distributed continuous variables), b: Kruskal-Wallis test (for non-normally distributed continuous variables) Fig. 2 Temporal Dynamics of Serum AMH Levels. Boxplots showing AMH decline patterns across treatment groups. ( A ) AMH decline rate at 1 month postoperatively. ( B ) AMH decline rate at 6 months postoperatively. The median AMH decline at 6 months was 6.70% for oxidised cellulose polymer, 16.65% for bipolar coagulation, and 5.40% for suture. Mean values showed similar patterns: oxidised cellulose polymer (5.97%), bipolar coagulation (19.06%), and suture (7.22%). The bipolar coagulation group showed significantly higher AMH decline compared to other groups (P < 0.001) Temporal Dynamics of Serum AMH Levels. Boxplots showing AMH decline patterns across treatment groups. ( A ) AMH decline rate at 1 month postoperatively. ( B ) AMH decline rate at 6 months postoperatively. The median AMH decline at 6 months was 6.70% for oxidised cellulose polymer, 16.65% for bipolar coagulation, and 5.40% for suture. Mean values showed similar patterns: oxidised cellulose polymer (5.97%), bipolar coagulation (19.06%), and suture (7.22%). The bipolar coagulation group showed significantly higher AMH decline compared to other groups (P < 0.001) Similar patterns were observed for AFC decline at 6 months (Fig.  3 ). Oxidised cellulose polymer showed superior AFC preservation (mean decline 14.0 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 8.7%) compared to bipolar coagulation (30.7 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 11.0%, P <0.001) and suture (17.5 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 11.6%, P =0.023) (Table  2 ). Fig. 3 AFC Decline Analysis at 6 Months: Boxplot comparing AFC decline rates at 6 months postoperatively across treatment groups. Oxidised cellulose polymer showed the best AFC preservation (mean decline 14.0%) compared to bipolar coagulation (30.7%, P  < 0.001) and suture (17.5%, P  = 0.023). The bipolar coagulation group demonstrated significantly higher AFC decline, indicating poorer ovarian reserve preservation compared to non-thermal hemostatic methods AFC Decline Analysis at 6 Months: Boxplot comparing AFC decline rates at 6 months postoperatively across treatment groups. Oxidised cellulose polymer showed the best AFC preservation (mean decline 14.0%) compared to bipolar coagulation (30.7%, P  < 0.001) and suture (17.5%, P  = 0.023). The bipolar coagulation group demonstrated significantly higher AFC decline, indicating poorer ovarian reserve preservation compared to non-thermal hemostatic methods PSV decline rates followed the same pattern (Fig.  4 ), with oxidised cellulose polymer demonstrating the best preservation (mean decline 5.3 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 1.2%) compared to bipolar coagulation (20.0 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 2.5%, P <0.001) and suture (6.8 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 1.7%, P <0.001) (Table  2 ). Comparative analysis across all ovarian reserve markers demonstrated consistent patterns of preservation with oxidised cellulose polymer (Fig.  5 ). Fig. 4 PSV Decline Analysis at 6 Months: Boxplot showing PSV decline rates at 6 months postoperatively. Oxidised cellulose polymer demonstrated superior ovarian blood flow preservation (mean decline 5.3%) compared to bipolar coagulation (20.0%, P  < 0.001) and suture (6.8%, P  < 0.001). The results indicate that oxidised cellulose polymer better maintains ovarian perfusion, which is crucial for follicular development and ovarian function PSV Decline Analysis at 6 Months: Boxplot showing PSV decline rates at 6 months postoperatively. Oxidised cellulose polymer demonstrated superior ovarian blood flow preservation (mean decline 5.3%) compared to bipolar coagulation (20.0%, P  < 0.001) and suture (6.8%, P  < 0.001). The results indicate that oxidised cellulose polymer better maintains ovarian perfusion, which is crucial for follicular development and ovarian function Fig. 5 Comparative Analysis of Ovarian Reserve Decline. Boxplots comparing the percentage decline across ovarian reserve markers for each treatment group. ( A ) AMH decline at 1 month postoperatively. ( B ) AMH decline at 6 months postoperatively. ( C ) AFC decline at 6 months postoperatively. ( D ) PSV decline at 6 months postoperatively. The comparative analysis demonstrates consistent patterns of ovarian reserve preservation, with oxidised cellulose polymer showing the least decline across all parameters (AMH: 6.0%, AFC: 14.0%, PSV: 5.3%), followed by suture (AMH: 7.2%, AFC: 17.5%, PSV: 6.8%), and bipolar coagulation showing the highest decline (AMH: 19.1%, AFC: 30.7%, PSV: 20.0%) Comparative Analysis of Ovarian Reserve Decline. Boxplots comparing the percentage decline across ovarian reserve markers for each treatment group. ( A ) AMH decline at 1 month postoperatively. ( B ) AMH decline at 6 months postoperatively. ( C ) AFC decline at 6 months postoperatively. ( D ) PSV decline at 6 months postoperatively. The comparative analysis demonstrates consistent patterns of ovarian reserve preservation, with oxidised cellulose polymer showing the least decline across all parameters (AMH: 6.0%, AFC: 14.0%, PSV: 5.3%), followed by suture (AMH: 7.2%, AFC: 17.5%, PSV: 6.8%), and bipolar coagulation showing the highest decline (AMH: 19.1%, AFC: 30.7%, PSV: 20.0%) Two clinical prediction models (ORI-PreOp and ORI-PeriOp) were developed and validated for predicting poor ovarian recovery (> 30% AMH decline). At 6 months, the ORI-PreOp model achieved an AUC of 0.794 (95% CI: 0.650–0.900), while the ORI-PeriOp model achieved excellent performance with an AUC of 0.936 (95% CI: 0.850–0.980). At 1 month, performance was lower but the ORI-PeriOp model remained superior (AUC = 0.809, 95% CI: 0.650–0.900 vs. 0.546, 95% CI: 0.384–0.618). The ORI-PeriOp model demonstrated exceptional performance in identifying high-risk patients, achieving a PR-AUC of 0.31. Given the low event rate (4 events in 101 patients, 4.0% prevalence), this represents a 675% improvement over random classifier baseline (PR-AUC = 0.04) and exceeds the clinical utility threshold (PR-AUC > 0.30) for low-event-rate prediction models. The incorporation of intraoperative variables significantly enhanced model performance, with the ORI-PeriOp model showing superior discrimination compared to the ORI-PreOp model (PR-AUC 0.31 vs. 0.18) (Table  3 ). Table 3 Univariate analysis of predictors for AMH Decline > 30% at 6 months Variable Test Statistic P value Age 0.079 0.429a AMH 0.192 0.054a AFC -0.248 0.012a PSV -0.238 0.017a CystSize 0.165 0.099a BloodLoss -0.064 0.522a OperationTime -0.131 0.193a SurgeryType 0.008 0.929b Univariate analysis showing associations between preoperative and intraoperative variables and the primary outcome of > 30% AMH decline at 6 months. Only AFC ( p  = 0.012) and PSV ( p  = 0.017) showed statistically significant associations. Continuous variables were analyzed using Pearson correlation, categorical variables using chi-square test Univariate analysis of predictors for AMH Decline > 30% at 6 months Univariate analysis showing associations between preoperative and intraoperative variables and the primary outcome of > 30% AMH decline at 6 months. Only AFC ( p  = 0.012) and PSV ( p  = 0.017) showed statistically significant associations. Continuous variables were analyzed using Pearson correlation, categorical variables using chi-square test The ORI-PeriOp model was developed using Firth’s penalized logistic regression with the following architecture: \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned}\mathbf{ORI}\boldsymbol-\mathbf{PeriOp}&={\mathrm\beta}_0+{\mathrm\beta}_1\times\mathrm{Age}+{\mathrm\beta}_2\times\mathrm{AMH}\\&+{\mathrm\beta}_3\times\mathrm{AFC}+{\mathrm\beta}_4\times\mathrm{PSV}\\&+{\mathrm\beta}_5\times\mathrm{CystSize}\\&+{\mathrm\beta}_6\times\mathrm{SurgeryType\_Coagulation}\\&+{\mathrm\beta}_7\times\mathrm{SurgeryType\_Suture}\\&+{\mathrm\beta}_8\times\mathrm{BloodLoss}\\&+{\mathrm\beta}_9\times\mathrm{OperationTime} \end{aligned}$$\end{document} Internal validation using 5-fold stratified cross-validation and 1000 bootstrap samples confirmed model stability with good calibration (calibration slope = 1.08) (Figs.  6 and 7 ). Decision curve analysis demonstrated higher net benefit for ORI-PeriOp across most clinically relevant thresholds (Figs.  8 , 9 and 10 ). Complete model specification including all coefficients, performance metrics, and clinical implementation guidelines is provided in Supplementary Appendix (Sections A-E). Fig. 6 ORI Model Performance and ROC Curve Analysis: Four-panel figure showing comprehensive model performance analysis based on clinical data. Panel A displays ROC curves for both models at 6 months, demonstrating ORI-PeriOp’s superior performance (AUC 0.936 vs. 0.794). Panel B shows 1-month ROC curves with ORI-PeriOp also performing better (AUC 0.809 vs. 0.546). Panel C compares AUC values across time points, showing ORI-PeriOp’s consistent superiority at both time points. Panel D summarizes key performance metrics and event rates. The analysis reveals excellent predictive performance with ORI-PeriOp showing outstanding discrimination, particularly at 6 months ORI Model Performance and ROC Curve Analysis: Four-panel figure showing comprehensive model performance analysis based on clinical data. Panel A displays ROC curves for both models at 6 months, demonstrating ORI-PeriOp’s superior performance (AUC 0.936 vs. 0.794). Panel B shows 1-month ROC curves with ORI-PeriOp also performing better (AUC 0.809 vs. 0.546). Panel C compares AUC values across time points, showing ORI-PeriOp’s consistent superiority at both time points. Panel D summarizes key performance metrics and event rates. The analysis reveals excellent predictive performance with ORI-PeriOp showing outstanding discrimination, particularly at 6 months Fig. 7 ORI Calibration Analysis: Six-panel calibration analysis showing model calibration based on clinical data. Panels A - D display calibration curves for both models, with ORI-PeriOp showing excellent calibration at 6 months (slope 0.95) and ORI-PreOp showing good calibration (slope 0.88). Panel E compares calibration slopes across models and time points, demonstrating that higher-performing models show better calibration. Panel F provides calibration assessment summary. The analysis reveals excellent calibration performance, particularly for the superior ORI-PeriOp model at 6 months ORI Calibration Analysis: Six-panel calibration analysis showing model calibration based on clinical data. Panels A - D display calibration curves for both models, with ORI-PeriOp showing excellent calibration at 6 months (slope 0.95) and ORI-PreOp showing good calibration (slope 0.88). Panel E compares calibration slopes across models and time points, demonstrating that higher-performing models show better calibration. Panel F provides calibration assessment summary. The analysis reveals excellent calibration performance, particularly for the superior ORI-PeriOp model at 6 months Fig. 8 ORI Variable Importance and Forest Plot Analysis: Comprehensive variable importance analysis based on clinical data from Firth penalized logistic regression. Panel A shows forest plot for ORI-PreOp (preoperative variables only) with AMH as the strongest predictor (OR = 2.75, p  = 0.023). Panel B displays ORI-PeriOp forest plot including intraoperative variables, with oxidised cellulose showing protective effect (OR = 0.40, p  = 0.130). Panel C compares variable importance between models, showing consistent AMH importance across models. Panel D summarizes key findings, highlighting AMH’s significant predictive value and the important contribution of intraoperative variables to model performance ORI Variable Importance and Forest Plot Analysis: Comprehensive variable importance analysis based on clinical data from Firth penalized logistic regression. Panel A shows forest plot for ORI-PreOp (preoperative variables only) with AMH as the strongest predictor (OR = 2.75, p  = 0.023). Panel B displays ORI-PeriOp forest plot including intraoperative variables, with oxidised cellulose showing protective effect (OR = 0.40, p  = 0.130). Panel C compares variable importance between models, showing consistent AMH importance across models. Panel D summarizes key findings, highlighting AMH’s significant predictive value and the important contribution of intraoperative variables to model performance Fig. 9 ORI Internal Validation Results: Six-panel internal validation analysis based on clinical data. Panels A - B show bootstrap distributions with excellent stability, particularly for ORI-PeriOp (SD = 0.038 at 6 months). Panel C summarizes validation metrics with minimal optimism correction (< 0.02), indicating excellent internal validity. Panel D displays 5-fold cross-validation results demonstrating outstanding consistency. Panel E compares performance standard deviations across validation methods, showing superior stability for ORI-PeriOp. Panel F provides validation assessment confirming robust model performance suitable for clinical application ORI Internal Validation Results: Six-panel internal validation analysis based on clinical data. Panels A - B show bootstrap distributions with excellent stability, particularly for ORI-PeriOp (SD = 0.038 at 6 months). Panel C summarizes validation metrics with minimal optimism correction (< 0.02), indicating excellent internal validity. Panel D displays 5-fold cross-validation results demonstrating outstanding consistency. Panel E compares performance standard deviations across validation methods, showing superior stability for ORI-PeriOp. Panel F provides validation assessment confirming robust model performance suitable for clinical application Fig. 10 ORI Decision Curve Analysis and Clinical Utility: Six-panel decision curve analysis based on clinical data demonstrating excellent clinical utility. Panels A - B show net benefit across threshold probabilities, with ORI-PeriOp showing superior utility at both time points. Panel C displays high clinical utility at relevant clinical thresholds, particularly for ORI-PeriOp. Panel D shows clinical impact by intervention type, highlighting excellent value for surgical technique selection and postoperative monitoring. Panel E presents cost-benefit analysis supporting ORI-PeriOp implementation. Panel F summarizes clinical decision support applications. The analysis demonstrates excellent clinical utility across multiple domains, strongly supporting ORI-PeriOp for clinical decision-making ORI Decision Curve Analysis and Clinical Utility: Six-panel decision curve analysis based on clinical data demonstrating excellent clinical utility. Panels A - B show net benefit across threshold probabilities, with ORI-PeriOp showing superior utility at both time points. Panel C displays high clinical utility at relevant clinical thresholds, particularly for ORI-PeriOp. Panel D shows clinical impact by intervention type, highlighting excellent value for surgical technique selection and postoperative monitoring. Panel E presents cost-benefit analysis supporting ORI-PeriOp implementation. Panel F summarizes clinical decision support applications. The analysis demonstrates excellent clinical utility across multiple domains, strongly supporting ORI-PeriOp for clinical decision-making Despite a higher event rate at 1 month (44.6%) compared to 6 months (4.0%), model discrimination was consistently lower at the earlier time point. This discrepancy may be attributed to: (1) the greater difficulty of distinguishing acute postoperative trauma and inflammation, (2) biological differences whereby the 6-month decline reflects structural ovarian reserve damage rather than transient effects, and (3) the stronger clinical relevance of baseline ovarian reserve for long-term recovery. Preoperative AMH was the strongest predictor in both models (ORI-PreOp: OR = 2.75, P = 0.023). Incorporation of intraoperative factors, particularly the use of oxidised cellulose (OR = 0.40, P = 0.130), contributed substantially to the superior performance of the ORI-PeriOp model (Figure 8 ). Risk stratification based on model predictions showed clear separation of outcomes, with poor ovarian recovery (>30% AMH decline) at 6 months observed in 4.0% of patients overall, and and at higher rates in the bipolar coagulation group (Table 4 ). Table 4 Clinical outcomes and safety analysis Characteristic Oxidised ( n  = 37) Bipolar ( n  = 33) Suture ( n  = 31) P value Operative Outcomes Operation time, min 31.0 (25–36) 33.0 (27–37) 36.0 (33–42) 0.005b Blood loss, mL 34.7 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 7.1 35.0 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 15.1 37.9 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} 11.7 0.473a Hospital stay, days 3.0 (3–4) 3.0 (3–4) 4.0 (4–5) 0.001b Complications Any complication 0 (0.0%) 0 (0.0%) 0 (0.0%) 0.313c Data are presented as mean ± standard deviation, median (interquartile ranges) or frequencies. a: ANOVA (for normally distributed continuous variables), b: Kruskal-Wallis test (for non-normally distributed continuous variables), c: Chi-square test (for categorical variables). Complete statistical analysis details are provided in Supplementary Appendix Section M Clinical outcomes and safety analysis Data are presented as mean ± standard deviation, median (interquartile ranges) or frequencies. a: ANOVA (for normally distributed continuous variables), b: Kruskal-Wallis test (for non-normally distributed continuous variables), c: Chi-square test (for categorical variables). Complete statistical analysis details are provided in Supplementary Appendix Section M Extremely rare outcomes (e.g., AMH < 1 ng/ml, 1.0% at both 1 and 6 months) may limit model stability and inflate performance metrics. Nevertheless, the consistency of results across time points strengthens confidence in the robustness of the findings. Operative times differed significantly between groups ( P  = 0.005), with oxidised cellulose polymer requiring the least time (median 31.0 min, IQR 25.0–36.0) and suture the most (median 36.0 min, IQR 32.5–42.0). Postoperative hospital stay was shortest with oxidised cellulose polymer (median 3.0 days vs. 4.0 days for suture, P  = 0.001) (Table  1 ). Complication rates were low across all groups, with no significant differences in adverse events (Table  5 ). The most common complications were postoperative pain (2.7% in oxidised cellulose polymer group) and transient fever (6.5% in suture group). Table 5 Performance metrics of ORI models at 6 months Model AUC-ROC (95% CI) PR-AUC Improvement over Random Clinical Utility ORI-PreOp 0.794 (0.650–0.900) 0.18 260% Limited ORI-PeriOp 0.936 (0.850–0.980) 0.31 527% Excellent Random 0.50 0.05 - None Performance metrics of ORI models at 6 months

This randomized controlled trial demonstrates that the Ovarian Reserve Index (ORI) models provide a robust framework for preoperative risk assessment in women undergoing endometriosis-associated ovarian cystectomy. The ORI-PeriOp model exhibited excellent discriminative ability (AUC 0.936 at 6 months) and consistent predictive performance across both short-term (1 month) and longer-term (6 months) outcomes, enabling precise stratification of patients at high risk for impaired ovarian recovery. By integrating preoperative and intraoperative variables, the ORI models offer a practical tool for individualized surgical planning and fertility preservation counseling. Complementing these predictive insights, our findings confirm that oxidised cellulose polymer provides superior ovarian reserve preservation compared with bipolar coagulation and suture, with a 69% relative reduction in AMH decline at 6 months postoperatively (6.0% vs. 19.1%). This benefit is likely attributable to the polymer’s mechanism of action, promoting hemostasis through platelet activation and fibrin formation without thermal injury to ovarian tissue. In contrast, bipolar coagulation may compromise follicular viability and stromal function due to thermal damage. Together, these results highlight a dual approach to optimizing ovarian outcomes: employing surgical techniques that minimize ovarian injury and leveraging predictive models such as ORI to guide individualized perioperative decision-making. Our findings extend and strengthen previous evidence regarding hemostatic method selection in ovarian surgery [ 3 , 7 – 9 , 13 , 19 ]. While earlier studies suggested potential benefits of non-thermal hemostatic methods, our trial provides comprehensive evidence with robust follow-up and detailed ovarian reserve assessment. These findings, together with the predictive capability of the ORI models, offer a framework for individualized perioperative decision-making. The superior ovarian reserve preservation observed with oxidised cellulose polymer likely reflects its mechanism of action, promoting hemostasis through platelet activation and fibrin formation without thermal injury to ovarian tissue [ 9 , 13 ]. In contrast, bipolar coagulation causes thermal injury that may compromise follicular viability and stromal function [ 20 , 21 ]. Beyond the evaluation of hemostatic techniques, our study introduces the Ovarian Reserve Index (ORI) models as a complementary innovation in the prediction of postoperative ovarian recovery. Unlike prior approaches that relied on single surrogate markers [ 16 , 22 ], these models integrate preoperative and intraoperative parameters, thereby providing a more nuanced assessment of ovarian reserve trajectories. Several features deserve emphasis. First, the inclusion of intraoperative variables enhanced predictive accuracy, as reflected by the superior performance of the ORI-PeriOp model (AUC 0.936 at 6 months). Second, the model demonstrated consistent discriminative capacity across both early and longer-term outcomes, supporting its robustness. Third, the ability to stratify patients into distinct risk categories offers clear clinical utility, enabling identification of women who may benefit from modified surgical techniques or fertility-preserving strategies. The differential model performance between 1-month and 6-month outcomes also yields biological insight. Short-term AMH decline likely reflects acute surgical trauma and inflammatory responses, which are difficult to predict from baseline characteristics. In contrast, sustained decline at 6 months appears to be more closely linked to baseline ovarian reserve and the cumulative effect of surgical factors, thus allowing stronger prediction. The 527% relative improvement in PR-AUC over random classification achieved by the ORI-PeriOp model underscores its methodological and clinical significance. Our findings suggest that incorporating intraoperative factors is crucial for accurate risk stratification. The deliberate choice of PR-AUC as the primary performance metric highlights clinical relevance in low-event-rate settings, where traditional AUC may overestimate utility. The comprehensive internal validation, including 1000 bootstrap resamples, demonstrates that meaningful prediction is achievable even under challenging data constraints. Together, these features provide a methodological framework that may inform the development and evaluation of other clinical prediction models in similarly imbalanced clinical contexts. Previous studies comparing hemostatic methods have yielded inconsistent results, largely due to methodological limitations including small sample sizes, short follow-up periods, and lack of comprehensive ovarian reserve assessment. Our trial addresses these limitations with a robust design, adequate sample size, comprehensive outcome assessment, and longer follow-up period. The observed AMH decline rates in our study are consistent with previous reports, ranging from 6 to 20% depending on the hemostatic method used [ 11 , 23 ]. However, our study provides more precise estimates due to the larger sample size and standardized outcome assessment. This study was performed by a single senior surgeon with nearly a decade of experience in single-port laparoscopic surgery and approximately 500 annual cases of single-port gynecological reproductive system surgery. Our median durations of 31.0 (25–36) minutes for oxidized cellulose polymer, 33.0 (27–37) minutes for bipolar coagulation, and 36.0 (33–42) minutes for suturing reflect complete surgical management and align with efficiency trends reported in modern ovarian surgery literature. Recent comparative studies demonstrate that conventional laparoscopy typically requires 59–88 min, while robotic-assisted procedures extend to 86.5–106.6 min [ 24 – 26 ]. The superior efficiency of oxidized cellulose polymer in our cohort mirrors findings from other surgical specialties, where hemostatic agents have shown significant time savings compared to traditional methods. For instance, electrothermal bipolar vessel sealing systems have reduced operating time by approximately one quarter compared to conventional hemostasis techniques [ 27 ]. These temporal advantages hold important clinical implications for ovarian reserve preservation. The modest but consistent time differences between hemostatic methods in our study likely reflect the technical efficiency of each approach—oxidized cellulose polymer provides rapid hemostasis through platelet activation and fibrin formation without the precision demands of suture placement or the controlled application required for bipolar coagulation [ 28 ]. Recent systematic reviews emphasize that surgical technique critically impacts ovarian reserve preservation during cystectomy [ 29 , 30 ], and our findings suggest that hemostatic efficiency may contribute to this protective effect by minimizing overall surgical stress and thermal exposure. This balance between operative efficiency and technical thoroughness becomes particularly crucial when the primary goal is ovarian reserve preservation for future fertility, as shorter, precise procedures may reduce ovarian ischemia time and inflammatory responses while maintaining comprehensive surgical management. An interesting finding in our analysis was the discrepancy between median and mean values in AMH decline rates, particularly when comparing oxidised cellulose polymer and suture groups (Fig.  2 ). This represents a classic case of how distribution shape affects statistical interpretation. The boxplot (which displays medians) showed that oxidised cellulose polymer had a higher median AMH decline (6.70%) compared to suture (5.40%). However, the mean values showed the opposite pattern: oxidised cellulose polymer (5.97%) vs. suture (7.22%). This discrepancy can be explained by the different distribution characteristics between groups: Oxidised cellulose polymer group: The distribution was symmetric (skewness = 0.012), with median (6.70%) closely approximating the mean (5.97%). This indicates consistent treatment response across patients with minimal extreme values. Suture group: The distribution was right-skewed (skewness = 0.516), with mean (7.22%) substantially higher than median (5.40%). Several patients experienced much higher AMH decline rates (up to 18.20%), which pulled the mean upward. Oxidised cellulose polymer group: The distribution was symmetric (skewness = 0.012), with median (6.70%) closely approximating the mean (5.97%). This indicates consistent treatment response across patients with minimal extreme values. Suture group: The distribution was right-skewed (skewness = 0.516), with mean (7.22%) substantially higher than median (5.40%). Several patients experienced much higher AMH decline rates (up to 18.20%), which pulled the mean upward. This distributional difference has important clinical implications. The symmetric distribution in the oxidised cellulose polymer group suggests consistent and predictable treatment outcomes, with most patients experiencing AMH decline around 6%. In contrast, the right-skewed distribution in the suture group indicates more variable treatment response, where while most patients had moderate decline (around 5.4%), a subset experienced substantially worse outcomes [ 31 ]. The discrepancy between boxplot visualization (showing medians) and mean values highlights the importance of considering both measures when interpreting clinical trial data. Boxplots are resistant to outliers and show the typical patient experience, while means reflect the average effect across all patients and can be influenced by extreme values. When distributions are skewed, these two measures can tell different but equally important stories about treatment effectiveness. Our findings have important implications for clinical practice. The superior ovarian reserve preservation with oxidised cellulose polymer suggests that this method should be considered the preferred hemostatic approach in endometriosis-associated ovarian cystectomy, particularly for women who wish to preserve future fertility. The ORI model provides clinicians with a valuable tool for preoperative counseling and treatment planning. Patients identified as high-risk based on ORI scores may benefit from alternative treatment strategies, including fertility preservation techniques or referral to specialized centers. Despite the challenging low event rate of 4.0%, our ORI-PeriOp model achieved clinically meaningful performance with a PR-AUC of 0.31. This suggests that the model can effectively identify high-risk patients while avoiding unnecessary interventions for the 96% of patients at low risk. The substantial improvement in performance with the addition of intraoperative variables highlights the clinical importance of surgical technique selection in ovarian function preservation. The model provides a quantitative tool for precision medicine, enabling clinicians to focus resources on patients who would benefit most from close monitoring and potential fertility preservation interventions. Key strengths of our study include the randomized design, comprehensive assessment of ovarian reserve using multiple biomarkers, longer follow-up than most prior studies, and the development of a validated clinical prediction model. The incorporation of both preoperative and intraoperative variables allowed for a nuanced evaluation of surgical impact, and the use of multiple ovarian reserve markers provided a more complete picture of ovarian function than single-marker approaches. Several limitations warrant consideration. First, the study was conducted at a single center, which may limit generalizability. Second, although our follow-up extended to 6 months, longer-term effects on fertility outcomes remain unknown. Due to incomplete follow-up beyond 6 months, long-term trajectories could not be assessed in this cohort; extended follow-up is ongoing [ 32 – 34 ]. Third, we did not systematically evaluate postoperative adhesion formation or perform structural imaging assessments, as current clinical guidelines do not routinely recommend advanced imaging for adhesion evaluation after ovarian cystectomy. While ultrasound examination has limited specificity for detecting pelvic adhesions, future studies incorporating postoperative imaging or intraoperative second-look evaluations could provide valuable data on structural outcomes and adhesion risk. Fourth, regarding the safety profile of oxidized cellulose polymer, we observed no differences in postoperative complications, inflammation, or adhesion formation between treatment groups (Table  1 and Supplementary Appendix M). It is important to note that oxidized regenerated cellulose is typically absorbed within 6–8 weeks postoperatively, and our findings are consistent with previous reports demonstrating its favorable safety profile [ 35 , 36 ]. However, further clarification on postoperative handling, absorption timeline, and any observed inflammatory responses would strengthen the safety profile and guide clinical application. Fifth, the predictive models were internally validated only; external validation across independent cohorts and surgical contexts is essential before widespread adoption. Finally, the relatively low event rate (AMH decline > 30% at 6 months occurred in only 4.0% of cases) posed challenges for predictive modeling. While this low event rate is clinically favorable, it poses statistical challenges for model development and stability [ 37 , 38 ]. We addressed this through the application of Firth’s penalized logistic regression to mitigate small-sample bias, the use of PR-AUC metrics suitable for imbalanced data, and rigorous internal validation with 1000 bootstrap resamples. Nonetheless, external validation and testing in larger, more diverse cohorts are essential to confirm model stability and generalizability. Taken together, these strengths and limitations highlight both the promise and the challenges of applying predictive modeling in reproductive surgery, underscoring the need for multicenter studies with longer-term follow-up and external validation to refine and generalize the ORI models for clinical use. Future research should prioritize external validation of the ORI models in independent populations to confirm generalizability across diverse clinical settings and surgical teams. Further studies are warranted to evaluate the models’ predictive performance for long-term reproductive outcomes, including natural conception, assisted reproductive technology success, and live birth rates. Integration of the ORI models into prospective clinical workflows could also be explored to guide individualized surgical decision-making and fertility preservation strategies. In parallel, additional research should investigate cost-effectiveness of different hemostatic methods from a societal perspective, as well as the potential benefits of combined approaches, such as oxidised cellulose polymer with minimal bipolar coagulation, to optimize ovarian reserve preservation. Long-term follow-up studies incorporating both clinical outcomes and patient-reported reproductive quality of life are needed to fully assess the translational impact of these predictive tools.

This randomized controlled trial confirms that oxidised cellulose polymer provides superior ovarian reserve preservation compared with bipolar coagulation and suture in endometriosis-associated ovarian cystectomy, with a clinically meaningful 69% reduction in AMH decline relative to bipolar coagulation. Importantly, the development and validation of the Ovarian Reserve Index (ORI) models represents a major advance in preoperative risk assessment. The ORI-PeriOp model demonstrated excellent discriminative ability (AUC 0.936 at 6 months) and consistent predictive performance across both short-term and longer-term outcomes. By integrating preoperative and intraoperative variables, the ORI models allow precise risk stratification, enabling clinicians to identify patients at high risk for impaired ovarian recovery and tailor surgical or fertility-preserving strategies accordingly. Taken together, these findings highlight the dual contributions of surgical technique and predictive modeling. While oxidised cellulose polymer remains the preferred hemostatic method for fertility preservation, the ORI models provide a practical, individualized framework to optimize perioperative decision-making and patient counseling in women undergoing ovarian cystectomy for endometriosis. The complete study protocol, CONSORT flow diagram, and prediction model specification are provided in the Supplementary Appendix for transparency and reproducibility.

Endometriosis-associated ovarian cysts represent a significant clinical challenge, affecting approximately 17–44% of women with endometriosis and potentially compromising ovarian reserve [ 1 , 2 ]. Ovarian cystectomy, while effective for symptom relief and histological diagnosis, carries inherent risks of ovarian tissue damage and subsequent decline in ovarian function [ 3 – 5 ]. The choice of hemostatic method during cystectomy has emerged as a critical factor influencing postoperative ovarian reserve preservation, yet optimal practice remains controversial [ 6 , 7 ]. Traditional hemostatic methods include bipolar coagulation and suturing, both of which have been associated with varying degrees of ovarian reserve impairment [ 8 – 10 ]. Bipolar coagulation, while effective for hemostasis, may cause thermal damage to ovarian follicles and stromal tissue [ 11 , 12 ]. Suturing, though avoiding thermal injury, may cause mechanical trauma and require additional operative time [ 13 ]. Recently, oxidised cellulose polymer has emerged as a promising alternative, promoting hemostasis through platelet activation and fibrin formation without thermal effects [ 9 , 14 , 15 ]. The assessment of ovarian reserve has evolved significantly, with anti-Müllerian hormone (AMH) recognized as the most sensitive biochemical marker [ 16 – 18 ]. However, comprehensive evaluation requires integration of multiple parameters including antral follicle count (AFC) and ovarian blood flow parameters such as peak systolic velocity (PSV). Previous studies comparing hemostatic methods have yielded inconsistent results, largely due to methodological limitations including small sample sizes, short follow-up periods, and incomplete ovarian reserve assessment. The development of predictive models for ovarian reserve outcomes represents an important advance in personalized surgical planning. Such models could enable clinicians to identify high-risk patients and tailor surgical approaches accordingly. However, existing models have focused primarily on baseline characteristics without incorporating intraoperative factors or comprehensive postoperative assessments. We conducted this randomized controlled trial to compare the effects of three hemostatic methods on ovarian reserve preservation in endometriosis-associated ovarian cystectomy. Our primary objectives were to determine the optimal hemostatic approach and to develop a comprehensive predictive model for postoperative ovarian recovery. This study aims to: (1) compare the effectiveness of three hemostatic methods for ovarian reserve preservation in endometriosis cystectomy, and (2) develop and internally validate a predictive model for postoperative ovarian function decline. The study design and reporting follow CONSORT 2010 and TRIPOD guidelines (Fig.  1 and Supplementary Appendix J). Fig. 1 Study Flow Diagram: Patient flow diagram showing enrollment and allocation. Complete CONSORT flow diagram with detailed exclusion reasons is provided in Supplementary Appendix Section J.1 Study Flow Diagram: Patient flow diagram showing enrollment and allocation. Complete CONSORT flow diagram with detailed exclusion reasons is provided in Supplementary Appendix Section J.1

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