Development and Validation of a FIGO 2018 Staging-Based Risk Prediction Model for Lymph Node Metastasis in Cervical Cancer: A SEER Database Analysis with Internal and External Validation

Objective : To explore the risk factors for lymph node metastasis in patients with cervical cancer, construct a risk prediction model for cervical cancer lymph node metastasis based on the FIGO 2018 staging system, and validate the model using both internal and external datasets. This model aims to provide a simple and effective tool for the clinical assessment of lymph node metastasis risk in cervical cancer patients. Methods : A total of 5787 patients with pathologically confirmed cervical cancer from the SEER database, diagnosed between 2000 and 2021, were selected. The patients were randomly divided into a training group and an internal validation group in a 7:3 ratio using R software. Univariate and binary logistic regression analyses were employed to identify independent factors affecting lymph node metastasis in cervical cancer patients. A nomogram prediction model was developed based on the selected factors. The model's performance was evaluated using receiver operating characteristic curves and calculating the area under the curve , along with calibration curves. The Hosmer-Lemeshow goodness-of-fit test and Spiegelhalter Z test were used to assess model performance. Additionally, an external validation cohort consisting of 338 cervical cancer patients treated at our institution between January 2019 and December 2024 was used for receiver operating characteristic curve and calibration curve validation. Results : A total of 5787 cervical cancer patients from the SEER database were included, with 4051 patients in the training group and 1736 patients in the internal validation group. Among these, 729 patients (18.00%) in the training group and 312 patients (17.97%) in the internal validation group had lymph node metastasis. Univariate analysis revealed that histologic type, tumor stage and grade, and tumor size were significantly associated with lymph node metastasis in cervical cancer (P < 0.05). Binary logistic regression analysis identified age, histologic type, tumor stage and grade, and tumor size as factors significantly related to lymph node metastasis in cervical cancer (P < 0.001). The nomogram model based on these variables showed an AUC of 0.758 (95% CI: 0.739–0.776) in the training group, 0.753 (95% CI: 0.726–0.781) in the internal validation group, and 0.742 (95% CI: 0.678–0.805) in the external validation group, indicating good discrimination and stability of the model. The Hosmer-Lemeshow test for the training and internal validation groups both yielded P values > 0.05, and the Spiegelhalter Z test for the external validation group yielded a P value of 0.0989. Calibration curves showed good agreement between predicted probabilities and actual observed outcomes, suggesting excellent model fit. Conclusion : The FIGO 2018-based risk prediction model for lymph node metastasis in cervical cancer, developed using the SEER database, demonstrates high predictive accuracy and strong clinical applicability. This model provides an effective reference for the precise preoperative assessment of lymph node metastasis risk in cervical cancer patients.