A predictive model for recurrence after hysteroscopic endometrial polypectomy: a combined analysis of clinical and ultrasound parameters.

No significant differences were observed between the recurrence and non-recurrence groups regarding age, BMI, menstrual cycle, gravidity, history of diabetes mellitus, hyperlipidemia, and endometriosis, or levels of FSH, LH, E 2, and P (all P  > 0.05). However, the proportion of patients with a history of PCOS was significantly higher in the recurrence group than in the non-recurrence group (18.6% vs. 7.87%; P  = 0.048) (Table  1 ). Table 1 Comparison of baseline clinical characteristics [ \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\overline{\mathrm x}$$\end{document} ± s, M (P 25 , P 75 ), n (%)] Characteristic Recurrence group ( n  = 43) Non-recurrence group ( n  = 127) t/z/x 2 P Age (years) 41.88 ± 3.09 40.87 ± 3.15 1.83 0.07 BMI (kg/m 2 ) 23.6 ± 2.1 23.9 ± 2.3 1.03 0.31 Menstrual cycle (days) 29.7 ± 4.8 28.9 ± 4.5 0.40 0.69 Gravidity (times) 2 (1, 3) 2 (1, 2) 1.30 0.19 History of diabetes mellitus 2 (4.65%) 11 (8.66%) 0.73 0.39 History of hyperlipidemia 2 (4.65%) 4 (3.15%) 0.21 0.65 History of PCOS 8 (18.6%) 10 (7.87%) 3.91 0.048 History of endometriosis 6 (13.95%) 14 (11.02%) 0.27 0.61 FSH (IU/L) 7.77 ± 1.19 7.63 ± 1.51 0.54 0.59 LH (IU/L) 6.59 ± 1.31 6.37 ± 1.31 1.01 0.31 E 2 (pg/mL) 80.72 ± 6.93 79.01 ± 7.83 1.28 0.20 P (ng/mL) 0.88 ± 0.20 0.85 ± 0.20 0.81 0.42 BMI Body mass index, PCOS Polycystic ovary syndrome, FSH Follicle-stimulating hormone, LH Luteinizing hormone, E 2 Estradiol, P Progesterone Comparison of baseline clinical characteristics [ \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\overline{\mathrm x}$$\end{document} ± s, M (P 25 , P 75 ), n (%)] BMI Body mass index, PCOS Polycystic ovary syndrome, FSH Follicle-stimulating hormone, LH Luteinizing hormone, E 2 Estradiol, P Progesterone Analysis of ultrasound parameters revealed that endometrial thickness ( P  < 0.001) and the proportion of patients with multiple polyps ( P  = 0.006) were significantly greater in the recurrence group. No significant differences were observed between the two groups concerning the maximum diameter of polyps or vascularity grades (Table  2 ). Table 2 Comparison of ultrasound parameters [ \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\overline{\mathrm x}$$\end{document} ± s, n (%)] Parameter Recurrence group ( n  = 43) Non-recurrence group ( n  = 127) t/x 2 P Endometrial thickness (mm) 11.35 ± 1.71 9.43 ± 1.90 5.87 < 0.001 Maximum diameter of polyps (mm) 15.11 ± 4.04 13.91 ± 4.15 1.65 0.10 Number of polyps 7.53 0.006  Single 19 (44.2%) 86 (67.7%)  Multiple 24 (55.8%) 41 (32.3%) Blood flow signals 2.92 0.09  Grade 0–Ⅰ 20 (46.51%) 78 (61.42%)  Grade Ⅱ–III 23 (53.49%) 49 (38.58%) Comparison of ultrasound parameters [ \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\overline{\mathrm x}$$\end{document} ± s, n (%)] No significant differences were observed between the two groups in WBC count, NEUT, PLT count, CRP level, or PLR (all P  > 0.05). In contrast, the NLR was significantly higher in the recurrence group ( P  = 0.01) (Table  3 ). Table 3 Comparison of inflammatory markers [ \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\overline{\mathrm x}$$\end{document} ± s] Marker Recurrence group ( n  = 43) Non-recurrence group ( n  = 127) t P WBC count (× 10 9 /L) 6.45 ± 1.59 6.07 ± 1.31 1.54 0.13 NEUT (%) 62.87 ± 5.17 61.11 ± 5.50 1.85 0.07 PLT count (× 10 9 /L) 242.02 ± 21.68 239.26 ± 17.07 0.85 0.39 CRP (mg/L) 2.77 ± 1.26 2.48 ± 1.30 1.27 0.21 NLR 2.33 ± 0.69 2.02 ± 0.65 2.61 0.01 PLR 130.93 ± 16.24 126.57 ± 17.51 1.44 0.15 WBC White blood cell, NEUT Neutrophil percentage, PLT Platelet, CRP C-reactive protein, NLR Neutrophil-to-lymphocyte ratio, PLR Platelet-to-lymphocyte ratio Comparison of inflammatory markers [ \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\overline{\mathrm x}$$\end{document} ± s] WBC White blood cell, NEUT Neutrophil percentage, PLT Platelet, CRP C-reactive protein, NLR Neutrophil-to-lymphocyte ratio, PLR Platelet-to-lymphocyte ratio Variables that were statistically significant in univariate analysis were included in the multivariate logistic regression model. The results demonstrated that a history of PCOS (OR = 3.883, 95% CI: 1.183–12.747, P  = 0.025), increased endometrial thickness (OR = 1.757, 95% CI: 1.384–2.232, P  < 0.001), and multiple polyps (OR = 2.818, 95% CI: 1.240–6.406, P  = 0.013) were independent risk factors for postoperative recurrence. NLR did not retain statistical significance in the multivariate model ( P  = 0.144) (Table  4 ). Collinearity diagnostics indicated that the VIF values for all variables were close to 1, suggesting no significant multicollinearity. Table 4 Multivariate logistic regression analysis of independent predictors of postoperative recurrence Variable β SE Wald χ² P OR 95% CI VIF Constant -8.557 1.488 33.061 < 0.001 – – – PCOS history 1.357 0.607 5.002 0.025 3.883 1.183–12.747 1.013 Endometrial thickness 0.564 0.122 21.378 < 0.001 1.757 1.384–2.232 1.017 Polyps number (multiple vs. single) 1.036 0.419 6.113 0.013 2.818 1.240–6.406 1.027 NLR 0.452 0.31 2.133 0.144 1.572 0.857–2.883 1.038 PCOS Polycystic ovary syndrome, NLR Neutrophil-to-lymphocyte ratio, SE Standard error, OR Odds ratio, CI Confidence interval, VIF Variance inflation factor Multivariate logistic regression analysis of independent predictors of postoperative recurrence PCOS Polycystic ovary syndrome, NLR Neutrophil-to-lymphocyte ratio, SE Standard error, OR Odds ratio, CI Confidence interval, VIF Variance inflation factor Among the individual predictors, endometrial thickness demonstrated the best predictive value for recurrence (AUC = 0.7793), superior to a PCOS history or polyp number alone. The combined model, based on PCOS history, endometrial thickness, and polyp number, demonstrated superior predictive performance. The combined model achieved an AUC of 0.8216 (95% CI: 0.7541–0.8890), with a sensitivity of 83.72% and a specificity of 67.72%. The optimal cut-off probability (Youden index) was 0.2038 (Table  5 ; Fig.  1 ). Table 5 ROC curve analysis of the combined model for predicting recurrence Variable AUC Sensitivity (%) Specificity (%) Optimal cut-off 95% CI Lower limit Upper limit PCOS history 0.5537 18.60 92.13 ≥ 0.5 0.4503 0.6570 Endometrial thickness 0.7793 81.39 66.14 ≥ 10.22 0.7015 0.8570 Polyp number 0.6177 55.81 67.72 ≥ 1.5 0.5189 0.7164 Combined diagnosis 0.8216 83.72 67.72 ≥ 0.2038 0.7541 0.8890 PCOS history was coded as 0 = no and 1 = yes. Polyp number was coded as 1 = single and 2 = multiple. For ROC curve analysis, values greater than or equal to the optimal cut-off were considered predictive of postoperative recurrence. For binary variables, ROC cut-off reflects classification threshold. ROC Receiver operating characteristic, PCOS Polycystic ovary syndrome, AUC Area under the curve, CI Confidence interval ROC curve analysis of the combined model for predicting recurrence PCOS history was coded as 0 = no and 1 = yes. Polyp number was coded as 1 = single and 2 = multiple. For ROC curve analysis, values greater than or equal to the optimal cut-off were considered predictive of postoperative recurrence. For binary variables, ROC cut-off reflects classification threshold. ROC Receiver operating characteristic, PCOS Polycystic ovary syndrome, AUC Area under the curve, CI Confidence interval Fig. 1 ROC curve analysis and predictive performance of the combined model. Note: ROC, receiver operating characteristic; PCOS, polycystic ovary syndrome ROC curve analysis and predictive performance of the combined model. Note: ROC, receiver operating characteristic; PCOS, polycystic ovary syndrome The predictive probability was calculated as: \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P\;=\;1/\lbrack1\;+\;e^{-\left(-8.557\;+\;1.357\;\times\;PCOS\;history\;+\;0.564\;\times\;endometrial\;thickness\;+\;1.036\;\times\;number\;of\;polyps\right)}\rbrack$$\end{document} , where PCOS history was coded as 0 = no and 1 = yes, and number of polyps was coded as 1 = single polyp and 2 = multiple polyps. A nomogram prediction model was constructed based on the independent risk factors: PCOS history, endometrial thickness, and polyp number (Fig.  2 ). The nomogram demonstrated that endometrial thickness contributed the greatest proportion to the total score, followed by polyp number and PCOS history. Fig. 2 Nomogram for predicting postoperative recurrence of endometrial polyps. The nomogram was developed based on three independent risk factors identified in multivariate analysis, including PCOS history, endometrial thickness, and number of polyps. Each variable corresponds to a specific point score, and the total points are used to estimate the probability of postoperative recurrence for individual patients. Higher total scores indicate a higher predicted risk of recurrence Nomogram for predicting postoperative recurrence of endometrial polyps. The nomogram was developed based on three independent risk factors identified in multivariate analysis, including PCOS history, endometrial thickness, and number of polyps. Each variable corresponds to a specific point score, and the total points are used to estimate the probability of postoperative recurrence for individual patients. Higher total scores indicate a higher predicted risk of recurrence The calibration curve showed good agreement between the bias-corrected curve and the ideal curve, indicating strong consistency between predicted and observed probabilities (Fig.  3 ). The Hosmer-Lemeshow test demonstrated satisfactory model fit (χ² = 2.023, P  = 0.980). Internal validation with 1,000 bootstrap resamples yielded a corrected Dxy value of 0.618 (corresponding to an AUC of ~ 0.81), a mean absolute error of 0.02, and a calibration slope of 0.932. These findings indicate good discriminative ability and calibration stability. Fig. 3 Calibration curve of the nomogram for predicting postoperative recurrence. The calibration curve illustrates the agreement between the predicted probability of recurrence and the observed outcomes. The dashed diagonal line represents the ideal reference line indicating perfect prediction. The solid line represents the bias-corrected performance of the model, while the dotted line represents the apparent performance. The light grey bands indicate the 95% confidence intervals estimated by bootstrap resampling (1,000 iterations). A closer alignment between the bias-corrected curve and the ideal line indicates good calibration of the predictive model Calibration curve of the nomogram for predicting postoperative recurrence. The calibration curve illustrates the agreement between the predicted probability of recurrence and the observed outcomes. The dashed diagonal line represents the ideal reference line indicating perfect prediction. The solid line represents the bias-corrected performance of the model, while the dotted line represents the apparent performance. The light grey bands indicate the 95% confidence intervals estimated by bootstrap resampling (1,000 iterations). A closer alignment between the bias-corrected curve and the ideal line indicates good calibration of the predictive model DCA demonstrated that, across a wide range of threshold probabilities, the combined model provided a greater net clinical benefit than the “treat-all” or “treat-none” strategies (Fig.  4 ), suggesting potential clinical utility for individualized decision-making. Fig. 4 Decision curve analysis (DCA) of the nomogram for predicting postoperative recurrence. The x-axis represents the threshold probability, and the y-axis represents net benefit. The blue line represents the proposed model, while the red and green lines represent the “treat-all” and “treat-none” strategies, respectively. A higher net benefit of the model across a range of threshold probabilities suggests potential usefulness in decision-making under certain clinical contexts Decision curve analysis (DCA) of the nomogram for predicting postoperative recurrence. The x-axis represents the threshold probability, and the y-axis represents net benefit. The blue line represents the proposed model, while the red and green lines represent the “treat-all” and “treat-none” strategies, respectively. A higher net benefit of the model across a range of threshold probabilities suggests potential usefulness in decision-making under certain clinical contexts Using the optimal cut-off value (0.2038), patients were stratified into low-risk and high-risk groups. The recurrence rate in the high-risk group was markedly higher than in the low-risk group [46.8% (36/77) vs. 7.5% (7/93)], a statistically significant difference ( χ² = 34.30, P  < 0.001; Table  6 ). Table 6 Comparison of recurrence rates based on risk stratification by the combined model [n (%)] Group Recurrence Non-recurrence Recurrence rate x 2 P Low-risk group 7 (16.28%) 86 (67.72%) 7.5% – – High-risk group 36 (83.72%) 41 (32.28%) 46.8% 34.30 < 0.001 Comparison of recurrence rates based on risk stratification by the combined model [n (%)]

The sample size was estimated based on the recurrence rate of endometrial polyps after hysteroscopic polypectomy reported in previous literature [ 15 ]. Using G*Power software, the minimum required sample size was calculated as 150 patients. To enhance the robustness of the study and account for potential loss to follow-up, a total of 170 patients were ultimately enrolled. This single-center retrospective observational study enrolled consecutive patients who underwent hysteroscopic endometrial polypectomy at our hospital between March 2023 and December 2024. All patients were followed postoperatively for at least 12 months. Based on the occurrence of endometrial polyp recurrence during the follow-up period, patients were divided into a recurrence group ( n  = 43) and a non-recurrence group ( n  = 127). Eligible participants met the following inclusion criteria: (1) age ≥ 18 years; (2) endometrial polyps confirmed by hysteroscopy and histopathological examination; (3) hysteroscopic endometrial polypectomy; (4) preoperative transvaginal ultrasonography; (5) preoperative laboratory assessment of inflammation-related indicators; and (6) at least 12 months of postoperative follow-up. The exclusion criteria were: (1) concurrent endometrial atypical hyperplasia or malignancy; (2) adenomyosis or submucosal uterine fibroids; (3) a history of endometrial ablation or long-term hormonal therapy; (4) pregnancy within 3 months after surgery; or (5) incomplete follow-up data. Baseline clinical data were collected for all patients, including age, body mass index (BMI), menstrual cycle characteristics, gravidity, and medical history of diabetes mellitus, hyperlipidemia, PCOS, and endometriosis. Additionally, preoperative serum levels of sex hormones, including follicle-stimulating hormone (FSH), luteinizing hormone (LH), estradiol (E 2 ), and progesterone (P), were documented. All patients underwent preoperative transvaginal ultrasonography performed by experienced sonographers. Ultrasound parameters included endometrial thickness, maximum diameter of the polyp(s), and number of polyps. Endometrial thickness was measured as the maximal double-layer thickness on the midsagittal uterine plane. Patients were classified as having either a single polyp or multiple polyps. Polyp vascularity was assessed using color Doppler flow imaging and graded as 0–I or II–III according to blood flow intensity. In premenopausal patients, ultrasound examinations were performed during the early proliferative phase of the menstrual cycle whenever possible. All ultrasound measurements followed the routine gynecologic ultrasound protocol at our hospital. Interobserver agreement between two sonographers for endometrial thickness measurements showed good consistency (intraclass correlation coefficient = 0.786). Peripheral venous blood samples were collected preoperatively for laboratory analysis. Recorded values included white blood cell (WBC) count, neutrophil percentage (NEUT), platelet (PLT) count, and C-reactive protein (CRP) level. Subsequently, the NLR and platelet-to-lymphocyte ratio (PLR) were calculated. The primary outcome of this study was 12-month postoperative recurrence of endometrial polyps, defined as the detection of new polypoid lesions within the uterine cavity on transvaginal ultrasound during follow-up, subsequently confirmed as endometrial polyps by repeat hysteroscopy and histopathological examination. Follow-up consisted of outpatient clinic visits and imaging assessments for at least 12 months, ending in December 2025. The date of the initial hysteroscopic endometrial polypectomy was defined as the starting point of follow-up. Statistical analyses were performed using SPSS software (version 26.0; IBM Corp., Armonk, NY, USA) and R software (version 4.4.2; R Foundation for Statistical Computing, Vienna, Austria). The Shapiro-Wilk test was applied to assess the normality of continuous variables. Normally distributed variables were presented as mean ± standard deviation ( \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\overline{\mathrm x}$$\end{document} ± s) and compared using the independent-samples t -test. Non-normally distributed variables were expressed as median [M (P 25 , P 75 )] and compared using the Mann-Whitney U test. Categorical variables were presented as counts and percentages [n (%)] and compared using the chi-square ( χ² ) test. Variables with P  < 0.05 in univariate analysis were included in the multivariate logistic regression analysis. Collinearity diagnostics were performed using variance inflation factor (VIF) analysis. Independent risk factors were identified by calculating odds ratios (ORs) and 95% confidence intervals (CIs). Receiver operating characteristic (ROC) curve analysis was performed to evaluate the predictive efficacy of single indicators and the combined model. Discriminative ability was assessed using the area under the curve (AUC), and the optimal cut-off point was determined using the Youden index. A nomogram predictive model was constructed based on the multivariate logistic regression results. Bootstrap resampling (1,000 repetitions) was performed for internal validation, and calibration curves and decision curve analysis (DCA) were used to evaluate model performance. Patients were stratified into risk groups based on the optimal cut-off point, and recurrence rates were compared between groups. A two-sided P -value < 0.05 was considered statistically significant.

In summary, this study constructed and validated a predictive model for postoperative recurrence risk after endometrial polyp surgery based on routinely available clinical and ultrasound parameters, while additionally evaluating the potential association of inflammation-related variables with recurrence. The model demonstrated good discriminative performance and may provide a quantitative reference for postoperative follow-up and risk-oriented management.

This retrospective cohort study addressed the common clinical issue of recurrence after hysteroscopic endometrial polypectomy by developing and internally validating a preliminary model based on readily available clinical and ultrasound parameters. The findings suggest that routinely obtainable variables may provide value for postoperative risk stratification. However, the clinical utility of the model should be interpreted with caution, and external validation in independent cohorts is needed before routine application. Therefore, the present model should be regarded as a preliminary risk assessment tool rather than a definitive clinical decision-making instrument. A history of PCOS, increased endometrial thickness, and a higher number of polyps were identified as independent predictors of postoperative recurrence. These findings suggest that postoperative recurrence may be associated not only with residual lesions but also with broader abnormalities in endometrial proliferative activity and tissue homeostasis [ 11 ]. Endometrial thickness, an easily obtainable imaging parameter, demonstrated favorable predictive performance in both multivariate and ROC curve analyses. This supports previous evidence linking the overall proliferative state of the endometrium to polyp recurrence [ 16 ]. Rather than reflecting a simple structural change, increased endometrial thickness may indicate a persistently proliferative or hormonally imbalanced endometrial milieu, serving as a surrogate marker of broader endometrial biological activity [ 17 ]. The number of polyps was also significantly associated with recurrence risk. Compared with maximum polyp diameter, polyp number may better reflect the diffuse nature of the disease, indicating more extensive structural or functional abnormalities within the endometrium [ 18 ]. Mechanistically, patients with multiple polyps may retain an underlying pathological predisposition even after surgical resection, thereby increasing the likelihood of new polyp formation [ 19 ]. From a biological perspective, recurrent lesions identified during follow-up may not necessarily represent true recurrence from residual microscopic disease but may instead reflect the persistence of an abnormal endometrial microenvironment or the de novo formation of new polyps. This consideration may be particularly relevant when interpreting predictors such as multiple polyps and increased endometrial thickness, both of which may reflect a sustained proliferative tendency of the endometrium. Notably, patients with PCOS often exhibit chronic anovulation, relative estrogen excess, and metabolic abnormalities. These factors may collectively affect cyclic shedding and repair processes, thereby increasing the risk of recurrence [ 20 ]. This finding corroborates previous research on the relationship between PCOS and endometrial proliferative disorders and suggests that PCOS may be an important factor associated with postoperative recurrence risk in this specific patient population. Recent studies exploring metabolic and endocrine-targeted interventions, including inositol supplementation, highlight the importance of metabolic regulation in long-term disease management for women with endocrine-related gynecological disorders [ 21 , 22 ]. Regarding inflammatory markers, NLR was associated with recurrence in the univariate analysis but did not retain independent significance in the multivariate model after adjustment for clinical and ultrasound variables. This finding suggests that systemic inflammation may play a relatively limited role in recurrence prediction compared with local endometrial characteristics and endocrine-related factors. ROC curve analysis demonstrated that the predictive model achieved superior performance compared with any single indicator and may help improve the discrimination between high- and low-risk patients, suggesting that multi-factor integration may contribute to more accurate characterization of postoperative recurrence risk. The nomogram, calibration analysis, and DCA further supported the discrimination ability and potential clinical utility of the model. The calibration curve demonstrated good agreement between predicted and observed probabilities, while DCA suggested potential net clinical benefit across a wide range of threshold probabilities. Nevertheless, these findings are based on internal validation only and should not be interpreted as sufficient evidence for immediate clinical implementation. Based on risk stratification, clinicians could potentially identify high-risk patients early and consider more intensive follow-up strategies, including more frequent transvaginal ultrasound surveillance and shorter intervals, to facilitate earlier detection of recurrence. In selected high-risk patients, particularly those with PCOS or recurrent proliferative conditions, closer endocrine or metabolic management and individualized postoperative hormonal interventions may also be considered as supportive strategies. In contrast, low-risk patients could avoid unnecessary intensive surveillance, helping optimize healthcare resource allocation and reduce patient burden. However, these management implications require further confirmation in independent external cohorts before translation into routine practice. From a clinical perspective, the predictive indicators included in this study were derived from routine examinations, offering good accessibility and reproducibility, which may facilitate future evaluation in different healthcare settings. The development of this combined model may help support risk-oriented postoperative management and contribute to the optimization of follow-up strategies for patients with endometrial polyps. Additionally, the standardized follow-up process and incorporation of internal validation and DCA further support the internal performance of the model. Despite these findings, several limitations should be acknowledged. First, the single-center design and relatively small cohort may restrict the broader applicability of the model. Although bootstrap internal validation demonstrated good model stability, external validation using independent cohorts was not performed. The relatively limited number of recurrence events may introduce a risk of residual overfitting, and the reported predictive performance may remain somewhat optimistic without external validation. Alternative penalized regression approaches, such as LASSO regression, may further improve model robustness and should be considered in future studies. Second, although all participants were followed for at least 12 months, this duration may be insufficient to fully capture long-term recurrence patterns. Longer follow-up studies are warranted to further evaluate the long-term stability of the model. Third, several potentially important confounding factors were not comprehensively evaluated. In particular, postoperative hormonal management (e.g., hormonal therapy, levonorgestrel-releasing intrauterine system use) may substantially influence recurrence risk. Differences between premenopausal and perimenopausal patients may also contribute to heterogeneity in endocrine status and recurrence dynamics. The absence of detailed stratified analyses regarding these factors may have partially affected the predictive performance and generalizability of the model. Furthermore, local endometrial microenvironment-related factors and molecular biomarkers were not evaluated, limiting further interpretation of the biological mechanisms underlying recurrence. Future multicenter studies with larger sample sizes should evaluate these potential confounding factors and explore additional biomarkers to further optimize the predictive performance and clinical applicability of the model.

Endometrial polyps commonly present with abnormal uterine bleeding, infertility, and recurrent miscarriage, which significantly affect the reproductive health and quality of life of reproductive-aged and perimenopausal women [ 1 – 3 ]. With the widespread application of hysteroscopic techniques, endometrial polypectomy has become the gold-standard surgical treatment, owing to its minimal invasiveness, excellent visualization, and reliable efficacy [ 4 , 5 ]. However, despite effective surgical removal, postoperative recurrence remains relatively common. Reported recurrence rates vary across studies [ 6 – 8 ], posing challenges for long-term patient management and clinical decision-making. Recent research has increasingly focused on factors associated with postoperative recurrence of endometrial polyps, particularly endocrine status, metabolic factors, and pathological characteristics [ 9 , 10 ]. For instance, polycystic ovary syndrome (PCOS), prolonged estrogen exposure, and the number or morphological features of polyps have been suggested as potential risk factors [ 11 ]. Preoperative transvaginal ultrasound also holds significant value for assessing endometrial thickness, polyp size, and vascularity [ 12 ]. Emerging evidence indicates that chronic low-grade inflammation may be associated with endometrial proliferative disorders, leading to the exploration of systemic inflammation-related markers, such as the neutrophil-to-lymphocyte ratio (NLR), as potential indicators of recurrence risk [ 13 ]. Although previous studies have explored various factors associated with endometrial polyp recurrence, several limitations remain. Most have focused on isolated clinical or imaging indicators, lacking an integrated evaluation of multidimensional factors, which limits predictive performance [ 6 , 11 ]. Furthermore, the independent predictive value of inflammation-related markers for recurrence remains uncertain and requires validation [ 14 ]. Limitations such as relatively small sample sizes, heterogeneous study populations, and insufficient follow-up duration may also affect the stability and generalizability of findings [ 7 ]. Therefore, constructing a clinically applicable recurrence risk assessment model based on routinely available clinical, imaging, and laboratory indicators is of considerable importance for postoperative risk stratification and individualized patient management. Accordingly, this retrospective cohort study integrated transvaginal ultrasound parameters with clinical variables and inflammation-related laboratory indicators to comprehensively evaluate factors associated with postoperative recurrence of endometrial polyps. A combined predictive model was subsequently developed and internally validated based on variables identified through multivariate analysis. The aim of this study was to provide a practical assessment tool for the early identification of recurrence risk after endometrial polyp surgery.