A nomogram model based on dual-center data predicts lymph node metastasis in endometrial cancer

Abstract

Objectives

The role of lymphadenectomy in endometrial cancer management remains controversial. This study aimed to identify clinicopathological characteristics and serum tumor markers that predict lymph node metastasis (LNM), and to develop a nomogram model integrating these parameters to predict individual LNM risk and guide personalized surgical planning.

Methods

Retrospective analysis was conducted on clinicopathological data and serum tumor marker levels for endometrial cancer patients who underwent surgical staging, including total hysterectomy, bilateral salpingo-oophorectomy, pelvic lymphadenectomy with or without para-aortic lymphadenectomy, at the Third Affiliated Hospital of Nanjing Medical University and Shanghai East Hospital of Tongji University from January 2008 to December 2024. Independent predictors of LNM were determined through univariate and multivariate logistic regression analyses and optimal subset selection. A nomogram model was developed using these variables. Predictive performance and accuracy were evaluated via receiver operating characteristic (ROC) curves and calibration plots.

Results

Among 506 eligible patients, 39 (7.7%) had LNM. Multivariate logistic regression and optimal subset analysis identified ovarian invasion (OR: 33.233, 95% CI 8.322–132.708), depth of myometrial invasion (OR: 27.223, 95% CI 10.989–67.436, P < 0.001), LVSI (OR: 13.129, 95% CI 6.191–27.841, P < 0.001), histological grade (OR: 2.933, 95% CI 1.800–4.778, P < 0.001), and CA199 levels (OR: 1.005, 95% CI 1.002–1.009, P < 0.004) as independent predictors of LNM. The nomogram demonstrated an AUC of 0.9106 (95% CI 0.8643–0.9569), specificity of 0.7923, sensitivity of 0.8974, accuracy of 0.8084, and negative predictive value of 0.9893. Calibration analysis showed strong agreement between predicted and observed outcomes. Decision curve analysis (DCA) indicated that the model offered clinical benefit.

Conclusion

The nomogram provides accurate LNM risk stratification in endometrial cancer, serving as a valuable tool to optimize lymphadenectomy strategies based on individual risk profiles.

Introduction

Endometrial cancer is the sixth most common malignancy among women worldwide, with 417,000 new cases and 97,000 related deaths reported in 2020 [1, 2]. In the United States, it ranks as the second most frequent gynecologic cancer and the fourth most commonly diagnosed cancer in women, following breast, lung, and colorectal cancers. Its mortality rate has now surpassed that of ovarian cancer, making it the leading cause of death among gynecologic malignancies [3–5]. The incidence continues to rise, driven by factors such as population aging, increasing obesity rates, and advancements in diagnostic technologies. The estimated lifetime risk for women is approximately 3%, with the median age at diagnosis being 61 years [1, 2]. Most patients present with early-stage endometrioid adenocarcinoma, which generally carries a favorable prognosis. However, approximately 15% of cases are diagnosed at an advanced stage, where prognosis is poor, and the 5-year overall survival (OS) rate ranges from 15% to 17% [6].

Surgical-pathological staging remains fundamental in the management of endometrial cancer, typically involving total hysterectomy, bilateral salpingo-oophorectomy, and lymphadenectomy (pelvic ± para-aortic). In addition to its role in prognostic stratification and guiding adjuvant therapy, lymphadenectomy has been linked to improved overall survival (OS) and recurrence-free survival (RFS) in select patients [7, 8]. However, it carries risks of adverse outcomes, including lymphocyst formation, lower extremity lymphedema, and chylous ascites, all of which can significantly impact the patient’s quality of life [9, 10]. In patients with low-risk, early-stage disease, evidence suggests a minimal incidence of nodal metastasis, and lymphadenectomy does not appear to provide a survival benefit. In these cases, nodal assessment may be safely omitted [9–13]. Given the ongoing debate regarding the therapeutic value of lymphadenectomy, a risk-adapted stratification approach for nodal metastasis is essential to personalize surgical intervention. This approach allows for selective lymphadenectomy, reducing postoperative complications and optimizing quality of life.

The nomogram model is a graphical representation of a predictive algorithm that estimates individualized event probabilities by incorporating risk variables identified through logistic regression analysis. Its use has gained widespread acceptance in oncology for predicting recurrence, survival outcomes, and response to adjuvant therapy in cancer patients [14–18]. Despite its broad clinical applications, its use in assessing lymph node metastasis (LNM) risk in endometrial cancer patients remains limited. This study aims to identify factors associated with LNM in endometrial cancer and develop a nomogram-based prediction tool to quantify metastatic risk, thereby assisting clinicians in devising individualized surgical strategies for lymphadenectomy.

Methods

Study patients

A retrospective cohort study was conducted involving 804 patients diagnosed with endometrial cancer, treated at the Third Affiliated Hospital of Nanjing Medical University and Shanghai East Hospital of Tongji University between January 2008 and December 2024. Eligibility required histopathological confirmation of primary endometrial cancer by two pathologists, surgical staging that included total hysterectomy, bilateral salpingo-oophorectomy, and pelvic lymphadenectomy with or without para-aortic lymph node dissection, availability of comprehensive clinicopathological records and serum tumor marker profiles, and classification according to the FIGO 2009 staging guidelines. Patients were excluded if they had other malignancies, did not undergo systematic lymph node dissection, received preoperative chemotherapy, radiotherapy, hormonal therapy, or immunotherapy, or had incomplete clinical or serological data. Following the application of these inclusion and exclusion criteria, 506 patients were eligible for analysis, including 422 from the Third Affiliated Hospital of Nanjing Medical University and 84 from Shanghai East Hospital of Tongji University. The cohort was stratified into LNM-negative (n = 467) and LNM-positive (n = 39) groups based on the presence or absence of lymph node metastasis (Fig. 1). As a retrospective investigation involving no intervention, informed consent was waived, and the study received approval from the Ethics Committee of the Third Affiliated Hospital with Nanjing Medical University (Approval No. [2025] KY008-01), with the IRB also waiving the need for informed consent.

Fig. 1.

Flow diagram of study population selection

Study methods

Clinical data were systematically collected, including age, height, weight, age at menarche, menopausal status, marital status, gravidity, parity, and personal histories of hypertension, diabetes, surgical trauma, and malignancy. Serum tumor markers—CA125, CA199, CEA, and AFP—were measured within one week prior to surgery. Tumor marker testing at both centers was performed using the Elecsys assay kit and Elecsys 801 analyzer, following the manufacturer’s instructions and employing the electrochemiluminescence method. Surgical pathology parameters included the operative technique (laparotomy or laparoscopy), pathological subtype, histological grade, depth of myometrial invasion, LVSI, and evidence of cervical stromal, paracervical, ovarian, or fallopian tube involvement. All specimens were fixed in 10% neutral formalin, paraffin-embedded, routinely sectioned, and stained with hematoxylin and eosin (HE). They were then examined and diagnosed microscopically by experienced pathologists. Patients were classified into LNM-negative and LNM-positive cohorts based on the presence of lymph node metastasis.

Statistical methods

Statistical analyses were performed using R 4.1. Normally distributed continuous variables were expressed as mean ± standard deviation (x̄ ± S), with intergroup comparisons conducted using independent-sample t-tests. Non-normally distributed continuous variables were reported as medians with interquartile ranges [M (Q25, Q75)] and compared using the Mann–Whitney U test. Categorical variables were presented as percentages (%) and analyzed using chi-square or Fisher’s exact tests, as appropriate. Univariate and multivariate logistic regression models were used to identify predictors of LNM. The “bestglm” package in R was utilized for variable selection, identifying optimal subsets associated with LNM, which were then incorporated into a nomogram. The optimal subset selection method employed was AIC. Time-dependent ROC curves and decision curves (DCA) were generated to assess the model’s predictive accuracy, while calibration plots were constructed to evaluate the concordance between predicted and observed outcomes. Given the low incidence of LNM and sample imbalance, the SMOTE method was applied to balance the samples, refit the model, and evaluate its performance using ROC analysis. During SMOTE processing, the Random Forest method was employed for model training and evaluation, utilizing the R packages “smote family” and “randomforest.” Simultaneously, Leave Group Out Cross Validation (LGOCV) was applied for internal validation, repeated 10 times, with 70% of the data allocated to the training set in each iteration. Model performance was assessed using AUC. Variables with missing values exceeding 5% were excluded, while missing values under 5% were imputed using the Random Forest method. A P-value < 0.05 was considered statistically significant.

Results

Clinicopathological and serum tumor marker data of the two groups (Table 1)

Table 1.

Patient characteristics

	LNM negative (n = 467)	LNM positive (n = 39)	P
Age at surgery(years)	56 (49, 62)	55 (51, 62)	0.985
BMI(kg/m2)	24.98 (22.60, 27.24)	23.53 (22.59, 27.19)	0.477
Age of menarche(years)	14 (14, 15)	14 (14, 15)	0.770
Gestation	2 (2, 3)	3 (2, 4)	0.068
Production	1 (1, 2)	1 (1, 2)	0.713
Lesion diameter(cm)	1.80 (1.00, 2.80)	2.70 (1.92, 4.10)	< 0.001
AFP(ng/ml)	2.60 (1.81, 3.92)	2.10 (1.65, 4.36)	0.336
CEA(ng/mL)	1.44 (0.86, 2.33)	1.83 (1.17, 2.54)	0.178
CA199(U/mL)	13.50 (8.13, 24.73)	31.70 (12.73, 65.36)	< 0.001
CA125(U/mL)	16.07 (10.64, 25.09)	35.60 (18.88, 67.75)	< 0.001
Menopause, n (%)			0.450
No	172 (36.83)	12 (30.77)
Yes	295 (63.17)	27 (69.23)
Married, n (%)			1.000*
Yes	456 (97.64)	38 (97.44)
No	11 (2.36)	1 (2.56)
Hypertension, n (%)	180 (38.54)	16 (41.03)	0.760
Diabetes, n (%)	79 (16.92)	9 (23.08)	0.330
Operation or trauma, n (%)	301 (64.45)	24 (61.54)	0.715
Mn-fh, n (%)	16 (3.43)	0 (0.00)	0.485
Surgical approaches, n (%)	89 (19.06)	14 (35.90)	0.012
Laparoscope	378(80.9)	25(64.1)
Laparotomy	89(19.1)	14(35.9)
Histological type, n(%)			< 0.001
EEC	425 (91.01)	27 (69.23)
Non-EEC	42 (8.99)	12 (30.77)
Tumor grade, n(%)			< 0.001
1	188 (40.26)	3 (7.69)
2	175 (37.47)	17 (43.59)
3	104 (22.27)	19 (48.72)
Myometrial invasion, n(%)			< 0.001
<1/2	387 (82.87)	7 (17.95)
≥1/2	80 (17.13)	32 (82.05)
LVSI, n(%)	44 (9.42)	22 (56.41)	< 0.001
Cervical stromal invasion, n(%)	26 (5.57)	13 (33.33)	< 0.001
Parametrial invasion, n(%)	3 (0.64)	7 (17.95)	< 0.001*
Ovary involvement, n(%)	4 (0.86)	7 (17.95)	< 0.001*
Tubal involvement, n(%)	6 (1.28)	6 (15.38)	< 0.001*

LVSI lymphovascular invasion, Mn-fh family history of malignant neoplasms, EEC endometrioid carcinoma, Non-EEC non-endometrioid carcinoma, LNM lymph node metastasis, CA199 carbohydrate antigen 199, CA125 carbohydrate antigen 125, CEA carcinoembryonic antigen, AFP alpha fetoprotein

*Fisher’s precision probability test

A total of 506 patients were included and divided into LNM-negative (n = 467) and LNM-positive (n = 39) groups. LNM was observed in 39 patients (7.7%), with a median age of 55 years. The pelvic LNM incidence was 6.9% (35/506), concurrent pelvic and para-aortic lymph node involvement occurred in 2.8% (14/506), and isolated para-aortic metastasis was found in 0.8% (4/506). No significant differences were observed between the two groups regarding age, BMI, age at menarche, gravidity, parity, menstrual status, marital status, hypertension, diabetes, surgical trauma, family history of malignancy, AFP, or CEA levels (P > 0.05). However, significant intergroup differences were identified in tumor diameter, surgical modality, pathological subtype, histological grade, depth of myometrial invasion, LVSI, cervical stromal involvement, parametrial invasion, ovarian and fallopian tube invasion, CA199, and CA125 levels (P < 0.05). Notably, the LNM rate was significantly higher in patients undergoing laparotomy (15.7%) compared to those undergoing laparoscopy (6.6%) (P = 0.012), potentially reflecting a higher lymph node metastasis rate in patients undergoing open surgery.

Univariate logistic regression analysis (Table 2)

Table 2.

Univariate logistic regression analysis of the LNM

	β	S.E	Z	P	OR (95%CI)
Surgical approaches
Laparoscope					1.00 (Reference)
Laparotomy	0.87	0.35	2.45	0.014	2.38 (1.19 ~ 4.76)
Histological type
EEC					1.00 (Reference)
Non-EEC	1.50	0.38	3.93	< 0.001	4.50 (2.12 ~ 9.52)
Tumor grade
1					1.00 (Reference)
2	1.81	0.63	2.85	0.004	6.09 (1.75 ~ 21.13)
3	2.44	0.63	3.85	< 0.001	11.45 (3.31 ~ 39.59)
Myometrial invasion
<1/2					1.00 (Reference)
≥1/2	3.10	0.43	7.12	< 0.001	22.11 (9.43 ~ 51.86)
LVSI	2.52	0.36	7.01	< 0.001	12.44 (6.15 ~ 25.18)
Cervical stromal invasion	2.14	0.40	5.41	< 0.001	8.48 (3.91 ~ 18.40)
Parametrial invasion	3.52	0.71	4.93	< 0.001	33.83 (8.35 ~ 137.08)
Ovary involvement	3.23	0.65	4.95	< 0.001	25.32 (7.04 ~ 91.04)
Tubal involvement	2.64	0.60	4.36	< 0.001	13.97 (4.27 ~ 45.71)
Lesion diameter(cm)	0.29	0.09	3.35	< 0.001	1.34 (1.13 ~ 1.59)
CA199(U/mL)	0.01	0.00	2.89	0.004	1.01 (1.01 ~ 1.01)
CA125(U/mL)	0.01	0.00	3.75	< 0.001	1.01 (1.01 ~ 1.01)

OR Odds Ratio, CI Confidence Interval, LVSI lymphovascular invasion, EEC endometrioid carcinoma, Non-EEC non-endometrioid carcinoma, CA199 carbohydrate antigen 199, CA125 carbohydrate antigen 125

Univariate logistic regression identified laparotomy, non-endometrioid carcinoma, histological grade, myometrial invasion, LVSI, cervical stromal invasion, paracervical invasion, ovarian invasion, fallopian tube invasion, tumor diameter, CA199, and CA125 as variables significantly associated with LNM in endometrial cancer (P < 0.05). No significant associations were observed for the remaining variables.

Multivariate logistic regression analysis and best subset model analysis (Table 3)

Table 3.

Multivariate logistic regression analysis of LNM

	OR	95%CI	P值
Tumor grade	2.933	1.800-4.778	< 0.001
Myometrial invasion	27.223	10.989–67.436	< 0.001
LVSI	13.129	6.191–27.841	< 0.001
Ovary involvement	33.233	8.322-132.708	< 0.001
CA199(U/mL)	1.005	1.002–1.009	0.004

LNM lymph node metastasis, LVSI lymphovascular invasion, CA199 carbohydrate antigen 199

Multivariate logistic regression and best subset model analyses identified histological grade, myometrial invasion, LVSI, ovarian invasion, and CA199 as independent predictors of LNM in endometrial cancer. Variables found significant in the univariate analysis were incorporated into the optimal subset model using the “bestglm” package in R. In alignment with the multivariate results, the subset model also recognized histological grade, myometrial invasion, LVSI, ovarian invasion, and CA199 as independent risk factors. Ovarian invasion emerged as the most significant predictor, with a 33.233-fold increase in LNM risk (95% CI 8.322–132.708), followed by myometrial invasion (OR: 27.223, 95% CI 10.989–67.436, P < 0.001), LVSI (OR: 13.129, 95% CI 6.191–27.841, P < 0.001), histological grade (OR: 2.933, 95% CI 1.800–4.778, P < 0.001), and CA199 (OR: 1.005, 95% CI 1.002–1.009, P < 0.004).

Development of the nomogram prediction model

The nomogram was developed based on variables identified through multivariate logistic regression and optimal subset analysis, including histological grade, depth of myometrial invasion, LVSI, ovarian invasion, and CA199 (Fig. 2)

Fig. 2.

The nomogram prediction model of LNM. CA199 carbohydrate antigen 199, measured in U/mL, LVSI lymphovascular invasion

Assessment of predictive performance and calibration

The model exhibited strong discriminative ability, with an AUC of 0.9106 (95% CI 0.8643–0.9569), specificity of 0.7923, sensitivity of 0.8974, accuracy of 0.8084, negative predictive value of 0.9893, and positive predictive value of 0.2651 (Fig. 3). The clinical DCA revealed that the predictive model provided a higher clinical net benefit compared to both the “all” and “none” strategies (Fig. 4).For calibration evaluation, 1000 bootstrap replicates were applied to the modeling dataset. The resulting calibration curve plotted the nomogram-estimated LNM probability on the X-axis against the observed probability on the Y-axis. The dashed diagonal line represented perfect agreement between predicted and actual outcomes. The calibration plot indicated strong concordance between predicted and observed LNM probabilities (Fig. 5), confirming robust model calibration and predictive reliability. The model was re-evaluated using the SMOTE method, yielding excellent predictive performance (AUC 0.887, specificity 0.957, accuracy 0.901). Furthermore, internal cross-validation with LGOCV resulted in outstanding predictive performance (AUC 0.917, sensitivity 0.984).

Fig. 3.

ROC curves of nomogram

Fig. 4.

Decision curve analysis of nomogram

Fig. 5.

Internal calibration of the nomogram

Discussion

Endometrial cancer is a commonly encountered gynecologic malignancy, with most cases presenting early due to irregular vaginal bleeding. As a result, tumors are typically confined to the uterine corpus at diagnosis, leading to favorable outcomes, with a 5-year DFS approaching 90% and a 10-year survival rate around 75% [19, 20]. In contrast, the presence of LNM significantly worsens prognosis, reducing the 5-year DFS to approximately 54% and increasing the overall recurrence rate to nearly 48% [19, 20]. Systematic lymphadenectomy plays a key role in accurate staging, guiding adjuvant therapy, and providing prognostic information. However, LNM occurs in only a minority of early-stage cases, with a reported incidence of 7.7% in the present cohort—6.9% involving pelvic nodes and just 0.8% limited to isolated para-aortic nodes. These findings suggest that the majority of patients do not exhibit lymphatic spread. For most, systematic lymphadenectomy serves as a transitional intervention, introducing additional surgical risks such as extended operative time, increased intraoperative blood loss, prolonged hospitalization, lymphocyst formation, and lower extremity lymphedema, all of which adversely affect postoperative recovery and long-term quality of life. Risk-adapted strategies are therefore essential. Tailoring the extent of lymphadenectomy based on individualized risk stratification allows selective intervention in high-risk LNM populations, while sparing low-risk patients from unnecessary procedures, potentially improving clinical outcomes and reducing treatment-related morbidity.

The Mayo criteria are widely used in clinical practice to assess the risk of lymph node metastasis (LNM) in endometrial cancer [21]. However, Vargas R et al. found that over 70% of patients may have undergone unnecessary lymph node resection based on this criterion, which raises concerns about overtreatment [22]. To enhance predictive accuracy, Li et al. developed a nomogram model using the SEER database, which achieved an AUC of 0.848, surpassing the Mayo criteria [23]. Nevertheless, this model did not incorporate lymphovascular space invasion (LVSI), limiting its predictive scope. In the current study, LVSI emerged as a significant predictor of LNM, and the AUC of the constructed model was improved to 0.9106, with both sensitivity and negative predictive value outperforming Li et al.‘s model. Zhang et al. developed a preoperative prediction model combining depth of myometrial invasion assessed by MRI, CA125, CA199, and the platelet/lymphocyte ratio, yielding an AUC of 0.841. However, the small sample size and myometrial invasion assessed by MRI may have compromised accuracy [24]. In contrast, this study incorporated additional tumor markers, such as CEA and AFP, alongside CA125 and CA199, with myometrial and cervical invasion confirmed by intraoperative frozen pathology, enhancing result reliability. The model by Wang et al. included variables such as histological type, grade, myometrial invasion, LVSI, cervical invasion, parametrial invasion, and hemoglobin, achieving an AUC of 0.916 [25]. However, it did not specify the location of cervical invasion (mucosa or stroma), which impacts FIGO staging. This study, in addition to including information on cervical stromal invasion and parametrial invasion, also incorporated ovarian and fallopian tube invasion, providing an advantage in terms of comprehensiveness, accuracy, and predictive performance.

This study employed optimal subset regression to identify variables associated with LNM, ultimately selecting five key predictors: histological grade, myometrial invasion, LVSI, ovarian invasion, and CA199. Multivariate logistic regression analysis confirmed these variables as independent risk factors for LNM. A nomogram model was constructed based on these five factors to predict the risk of LNM. The model demonstrated robust discriminatory capacity, with an AUC of 0.9106 (95% CI 0.8643–0.9569), specificity of 0.7923, sensitivity of 0.8974, accuracy of 0.8084, and a negative predictive value of 0.9893. DCA indicated a favorable clinical net benefit rate for the predictive model. Calibration curves showed strong concordance between predicted probabilities and observed outcomes, further validating the reliability of the nomogram.

The likelihood of LNM increases with higher histological grade and deeper myometrial infiltration. Grade 3 tumors with the depth of invasion show a 34% incidence of LNM [26]. Two additional studies have also found a significant correlation between muscle layer infiltration, histological grade, and LNM [27, 28]. Mandić A et al. reported that while ≥ 50% myometrial invasion was associated with a higher LNM risk, an elevated histological grade did not correlate with increased LNM incidence [29]. Previous research has shown that deep myometrial invasion significantly increases the risk of LVSI (RR 3.07, 95% CI 2.17–4.35) and LNM (RR 4.45, 95% CI 3.29–6.01) in endometrial cancer, potentially due to increased microvascular density at the tumor site [30, 31]. Research suggested that intraoperative frozen sectioning exhibit high diagnostic accuracy for evaluating myometrial invasion, with reported accuracies of 94.4% [32]. This study also confirmed that muscle layer infiltration (OR 27.223, 95% CI 10.989–67.436) and histological grade (OR 2.933, 95% CI 1.800–4.778) were independent risk factors for LNM, consistent with previous studies.

LVSI not only serves as a prognostic marker in endometrial cancer but also exhibits a strong association with LNM. Restaino S et al. categorized LVSI into absent, focal, and diffuse forms, which corresponded to LNM rates of 5%, 15%, and 33%, respectively [33]. Evidence indicates that LVSI presence increases the likelihood of LNM by a factor of 11.04 and is associated with reduced PFS and OS [34]. In alignment with this, the present study observed a 13.129-fold increase in LNM risk among patients with LVSI.

Sun Y et al. reported that adnexal invasion increased the likelihood of para-aortic LNM in non-endometrioid carcinoma cases [35]. Similarly, Turan T et al. found that adnexal metastasis correlated with para-aortic dissemination [36], suggesting a potential link between adnexal spread and distant nodal involvement. Li M et al. further demonstrated that ovarian metastasis increased the risk of LNM by 6.236-fold [37]. Consistent with these findings, this study showed that the risk of LNM was highest when ovarian involvement occurred. Intraoperative frozen section pathology can confirm ovarian involvement. Given this evidence, intraoperative confirmation of ovarian involvement may justify comprehensive lymphadenectomy, including both pelvic and para-aortic nodes, to optimize surgical management.

Studies have demonstrated that elevated CA199 levels are significantly associated with advanced tumor stages, higher histological grades, and lymph node metastasis [38]. The present study identified CA199 as an independent risk factor for LNM. The potential mechanism by which CA199 induces LNM may involve CA199-induced macrophage polarization towards the M2 phenotype in the tumor microenvironment, which promotes the formation of an immunosuppressive environment and facilitates tumor cell immune escape, warranting further investigation [39]. Due to its accessibility and routine availability, preoperative CA199 measurement may serve as a practical adjunct for assessing preoperative lymph node status.

Several limitations must be acknowledged. First, the retrospective design and limited sample size may introduce selection and recall biases. Second, this study lacks molecular and imaging data; future research will incorporate these factors to explore their impact on LNM and improve prediction models. Third, although internal cross-validation demonstrated strong predictive performance, the absence of external validation remains a limitation. External validation in larger, independent cohorts is necessary to more robustly assess the model’s predictive accuracy.

Conclusions

In conclusion, the nomogram model developed in this study, incorporating histological grade, depth of myometrial invasion, LVSI, ovarian invasion, and CA199 levels, demonstrates strong predictive capacity for assessing LNM risk in endometrial cancer. Its application provides a foundation for personalized lymphadenectomy strategies aimed at minimizing surgical-related complications and improving postoperative quality of life.

Author contributions

LiuJunling-liujunling0181@163.com: data collection and management, data analysis, manuscript writing and editing. Chen Jiming-cjming@126.com: data collection and management, manuscript review and editing. Wang Wenjie-wwjinwy@126.com: data collection and management. Zheng Yafeng-zyf99010111@Outlook.com: data collection, methodology, manuscript review and editing, project administration. Li Fang-fang_li@ tongji.edu.cn: data collection, methodology, manuscript review and editing, project administration.

Funding

This study was supported by the Changzhou Municipal Health Commission Science and Technology Project (Project No. QN202363).

Data availability

The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The study received approval from the Ethics Committee of the Third Affiliated Hospital with Nanjing Medical University (Approval No. [2025] KY008-01, date: March 2025). All procedures were performed in accordance with the ethical standards of the institutional and national research committees and with the principles of the Declaration of Helsinki. Patient consent for participation was waived by the ethics committee of the Third Affiliated Hospital with Nanjing Medical University owing to the retrospective design of the study and the use of anonymized data.

Consent for publication

Not applicable, as this study did not include identifiable individual patient data and involved only anonymized retrospective analysis.

Competing interests

The authors declare no competing interests.