This multicenter retrospective cohort study enrolled patients with stage I–III EC who underwent primary surgical treatment between January 2016 and January 2023 at the First Affiliated Hospital of Chongqing Medical University (training cohort, n = 545) and Liangjiang Hospital of Chongqing Medical University and Women and Children’s Hospital of Chongqing Medical University (validation cohort, n = 315). The inclusion criteria were as follows (1): receipt of standardized surgical treatment, consisting of total hysterectomy with bilateral salpingo-oophorectomy, with the performance of systematic lymph node assessment determined by pathological risk features (15); The extent of lymphadenectomy (pelvic only or combined with para-aortic) was determined by comprehensive preoperative and intraoperative evaluation. Established criteria for identifying patients at low risk of lymph node metastasis include: (a) myometrial invasion less than 50%; (b) tumor diameter less than 2 cm; and (c) histologic grade G1 or G2. However, accurately determining these parameters before final pathology can often be challenging. When possible, intraoperative frozen section evaluation by gynecologic pathologists for assessing myometrial invasion and cervical involvement may guide decision-making—for instance, omitting systematic lymphadenectomy in cases confirmed to have no myometrial invasion or cervical involvement (16, 17) (2); availability of complete preoperative baseline data; and (3) completion of standard preoperative laboratory tests (including complete blood count, liver and renal function, coagulation profile, and tumor markers such as CA125 and HE4) and imaging studies (chest and abdominal CT, and pelvic MRI). Exclusion criteria included (1): non-standard surgical treatment (2); receipt of neoadjuvant therapy prior to surgery (3); incomplete clinical data (4); loss to follow-up (5); history of other malignant tumors; and (6) significant inflammatory or immune system diseases.

Adjuvant treatment regimens were individualized based on patients’ postoperative pathological characteristics. Radiotherapy was administered to those exhibiting at least one high-risk feature (18): age ≥60 years, specific histologic types (e.g., serous or clear cell carcinoma), high-grade tumor (G3), deep myometrial invasion (≥1/2), cervical stromal involvement, or lymphovascular space invasion. The radiotherapy regimen consisted of either vaginal brachytherapy (total dose 22–24 Gy in 4 fractions) or pelvic external beam radiotherapy (total dose 45–50 Gy in conventional fractions). For patients with FIGO stage III or higher disease, specific pathologic types, or G3 tumors with deep myometrial invasion, the carboplatin-paclitaxel regimen was recommended as the first-line systemic chemotherapy, administered over six 21-day cycles (15, 19).

All patients were managed under a standardized postoperative surveillance protocol: quarterly in the first two years, semiannually during years 3–5, and annually after five years. Follow-up assessments included (1): baseline evaluations, which consisted of detailed history-taking and pelvic-rectal examination (2); biomarker tests, such as CA125, when clinically indicated; and (3) diagnostic imaging (ultrasound, CT, or MRI) for suspected recurrence or metastasis. The study follow-up concluded in January 2025. Given that approximately 70%–80% of endometrial cancer recurrences occur within the first three years postoperatively, all surviving patients had completed at least 36 months of surveillance, excluding those lost to follow-up or deceased (18).

Recurrence is strictly defined as meeting both of the following criteria (1): independent confirmation by at least two gynecologic oncologists; and (2) objective evidence from at least one of the following: ① confirmed lesions on cross-sectional imaging (CT/MRI/PET-CT), ② histopathologic confirmation of malignant cells, or ③ persistently elevated serum tumor markers (e.g., CA125) after excluding other etiologies (6). Based on anatomic site, recurrences were categorized as either locoregional (including vaginal vault recurrence and pelvic sites such as the vesicorectal space) or distant metastasis (encompassing para-aortic lymph node involvement, peritoneal dissemination, or hematogenous spread to solid organs like liver, lung, or bone) (20). Recurrence-free survival (RFS) was calculated from the date of comprehensive staging surgery to the first occurrence of either pathologically or radiologically confirmed recurrence. Patients without recurrence were censored at their last follow-up date. Overall survival (OS) was measured from the initial surgery date until death from any cause. Surviving patients were censored at the study’s cutoff date, whereas those lost to follow-up were censored at the last documented contact (21).

We retrospectively collected clinicopathological data, including: 1) baseline characteristics: age and body mass index; 2) tumor features: FIGO stage, maximum tumor diameter, histologic subtype and grade, cervical stromal invasion, myometrial invasion depth (<50% or ≥50%), and lymphovascular space invasion. Formalin-fixed paraffin-embedded tissue specimens were obtained from the pathology department for subsequent immunohistochemical analysis of IGF2BP2 protein expression.

Our pre-processing and quality control pipeline comprised the following systematic steps to ensure data integrity and reproducibility: (1) Harmonization and standardization of multi-center data: To standardize the data collected from multiple sites, a centralized harmonization procedure was implemented. All categorical variables—such as FIGO stage and histological type—were recoded using standardized international classification systems (the FIGO 2009 staging system and the WHO 2020 classification for histology). Continuous variables, including age and BMI, were retained in their original numerical units without categorization at the pre-processing stage. (2) Exclusion of cases with missing critical variables: As detailed in the study flowchart (Figure 1), patients with missing data for any of the key variables ultimately included in the final nomogram model were excluded prior to model development. (3) Quality control and inter-rater reliability assessment for immunohistochemical scoring: Two expert gynecological pathologists independently assessed all IGF2BP2 and p53 slides while blinded to clinical outcomes. Inter-rater consistency was formally evaluated using the intraclass correlation coefficient (ICC). Discrepancies exceeding 10% were resolved through a consensus review using a multi-headed microscope. (4) Appropriate transformation of variables for clinical application: To align with established clinical practice and enhance the utility of the nomogram, specific continuous variables were dichotomized using validated or commonly accepted thresholds. Age was categorized as ≥60 vs. <60 years. Serum CA125 was dichotomized at 35 U/mL. The semi-quantitative IGF2BP2 immunohistochemical score was converted to a binary variable (high vs. low expression) using the median score of the training cohort as the optimal threshold, as determined by ROC analysis. (5) Verification of cohort balance: Baseline characteristics between the training and validation cohorts were compared using Chi-square tests (categorical) and t-tests (continuous) to ensure comparability.

Postoperative tissue specimens were fixed in formalin and uniformly processed for paraffin embedding, sectioning, and H&E staining by the pathology departments across all participating centers to identify carcinoma regions.

Pathological evaluation encompassed tumor location, maximum diameter, histologic type and grade, depth of myometrial invasion, lymphovascular space invasion (LVSI), and cervical stromal involvement. Histologic classification followed established criteria (21, 22), categorizing tumors as G1/G2 endometrioid adenocarcinoma, G3 endometrioid adenocarcinoma, or non-endometrioid subtypes (including serous and clear cell carcinomas) (21, 22).

Tissue sections were processed following standard Immunohistochemistry (IHC) protocols including antigen retrieval in citrate buffer (pH 6.0) and peroxidase blocking. Primary antibody incubations were performed with anti-IGF2BP2 (1:300, 4 °C overnight) and anti-p53 (clone DO7, 1:200) using an automated stainer. Detection was completed with HRP-conjugated secondary antibody and DAB visualization (21).Two pathologists independently assessed protein expression. P53 nuclear expression was quantified as percentage of positive cells (0-100%). Based on established molecular classification criteria for endometrial carcinoma, cases were categorized into three patterns with defined biological and clinical implications: p53-abnormal (overexpression pattern)-≥75% of tumor nuclei exhibiting strong, diffuse staining; p53-abnormal (null pattern)-complete absence (0%) of nuclear staining with appropriate internal positive controls; or p53-wild-type—1-75% of tumor nuclei showing heterogeneous staining intensity. This classification correlates with TP53 mutation status and informs molecular subtyping. IGF2BP2 cytoplasmic staining was semi-quantitatively scored based on intensity and distribution. Immunohistochemical assessment of IGF2BP2 employed a semi-quantitative scoring system (range 1–4), derived by multiplying the percentage of positively stained tumor cells by the staining intensity (graded on a 1–3 scale). This yielded final scores categorized as follows: 1 = negative, 2 = weak positivity, 3 = moderate positivity, and 4 = strong positivity. Using the median score of the training cohort as the optimal threshold, specimens with scores ≥2.5 were classified into the high-expression group for subsequent nomogram incorporation. IGF2BP2 expression was markedly higher in recurrent tumors compared to recurrence-free cases, as evidenced by representative immunohistochemical staining (Figure 2). Inter-observer differences in H-scores ≤10% were resolved by averaging the two independent assessments. For cases with discrepancies >10% (n=37, 4.3%), a formal consensus review was conducted: both pathologists jointly re-examined the slides using a multi-headed microscope and, through unblinded discussion guided by the standardized criteria, reached a definitive classification. This two-tier approach ensured consistent application of the scoring system across all specimens (23).

This retrospective cohort study included 545 endometrial cancer patients from the First Affiliated Hospital of Chongqing Medical University (January 2016-January 2023) as the training cohort, and 315 patients from the Women and Children’s Hospital and Liangjiang Hospital of Chongqing Medical University as the validation cohort. Categorical variables were presented as frequencies (percentages) and compared using χ² or Fisher’s exact tests; continuous variables were expressed as mean ± standard deviation and analyzed with independent samples t-tests. Statistical significance was declared for findings with p-values below 0.05.

Univariable Cox regression was first employed to identify clinicopathological parameters and IGF2BP2 expression levels associated with RFS (P < 0.05). Significant variables were subsequently incorporated into a multivariable Cox model to determine independent prognostic factors. A nomogram was constructed to predict 1-, 3-, and 5-year RFS probabilities. The optimal cutoff value for IGF2BP2 expression was determined by receiver operating characteristic (ROC) curve analysis, with the maximum Youden index (sensitivity + specificity - 1) as the criterion (24). Patients were stratified into high- and low-risk groups based on the predicted 3-year recurrence-free survival probability derived from the model. Kaplan-Meier survival curves were generated, and between-group differences were assessed using the log-rank test.

Calibration curves were plotted in both the training and validation cohorts to evaluate the agreement between predicted and observed outcomes. The concordance index (C-index) was calculated to assess model discrimination, with values of 0.5–0.6 indicating limited, 0.6–0.7 moderate, and >0.8 strong discriminatory ability. All statistical analyses were performed using SPSS (version 25.0, IBM Statistics, Chicago, IL, USA) and R software (version 4.0.3, http://www.r-project.org), with a two-sided P < 0.05 considered statistically significant (25).

This study enrolled a total of 860 EC patients, with 545 cases from The First Affiliated Hospital of Chongqing Medical University comprising the training cohort and 315 cases from the Liangjiang Hospital of Chongqing Medical University and Women and Children’s Hospital of Chongqing Medical University forming the validation cohort (Figure 1). As summarized in Table 1, the two cohorts demonstrated comparable distributions across all documented clinicopathological features-including age, body mass index(BMI), FIGO stage, histologic type, grade, cervical invasion, myometrial invasion depth, LVSI, adjuvant treatment, CA125 and IGF2BP2 expression level (all P > 0.05)-indicating well-balanced baseline characteristics between the groups.

As indicated in Table 2, the multivariable Cox regression analysis revealed FIGO stage (p = 0.001), depth of myometrial invasion (p = 0.004), histologic type (p = 0.001), LVSI (p = 0.007), CA125 level (p = 0.001), p53 expression status (p = 0.013), and IGF2BP2 expression (p < 0.001) as independent prognostic factors for RFS in EC patients.

In the training cohort, the nomogram demonstrated superior predictive performance, achieving an area under the ROC curve (AUC) of 0.884 (95% CI: 0.841–0.927), which outperformed both the clinicopathological parameter-based model (AUC = 0.840, 95% CI: 0.794–0.886) and IGF2BP2 alone (AUC = 0.671, 95% CI: 0.603–0.738). This discriminative ability was consistently maintained in the validation cohort, with AUCs of 0.865 (95% CI: 0.808–0.922) for the nomogram, 0.812 (95% CI: 0.776–0.889) for the clinical model, and 0.633 (95% CI: 0.543–0.724) for IGF2BP2 alone (Figure 3). It yielded a C-index of 0.884 (95% CI: 0.841–0.927) in the training cohort, which further increased to 0.865 (95% CI: 0.808–0.922) in the validation cohort, indicating robust performance.

Based on multivariable Cox regression analysis, we developed a nomogram (Figure 4) integrating clinicopathological parameters-including FIGO stage, histologic type and grade, cervical stromal invasion, myometrial invasion, LVSI, and adjuvant therapy-along with IGF2BP2 expression to predict recurrence-free survival (RFS) in endometrial cancer patients. By summing the points assigned to each predictor in the nomogram, individualized recurrence risk can be quantitatively assessed.

The nomogram exhibited exceptional discriminative capacity for forecasting 1-, 3-, and 5-year RFS, achieving AUCs of 0.920, 0.879, and 0.879 in the training cohort and 0.914, 0.866, and 0.838 in the validation cohort, respectively. These outcomes affirm its robust performance across various postoperative intervals, particularly sustaining high predictive accuracy during the 3-year peak recurrence period, thereby offering a quantitative foundation for customizing individualized follow-up strategies. The nomogram-predicted RFS rates exhibited excellent concordance with actual observations in both the training and validation cohorts (Figure 5), suggesting favorable calibration and clinical predictive accuracy of the model. The calibration curves demonstrated exceptional agreement between the model-predicted probabilities and the observed outcomes in both the training and validation cohorts (Figure 6), indicating that the model is well-calibrated.

ROC analysis identified an optimal IGF2BP2 cutoff value of 0.878 for predicting endometrial cancer recurrence, with an area under the curve (AUC) of 0.884, sensitivity of 81.8%, and specificity of 80.3% (Figure 7). This threshold was determined by maximizing the Youden index (sensitivity + specificity - 1), reflecting the model’s strong discriminative capacity in identifying high-risk patients.

Based on ROC analysis and the maximum Youden index, the optimal risk threshold for predicting EC recurrence in the training cohort was determined to be 0.878 (Figure 7). Patients were stratified into low-risk (3-year RFS ≥ 0.878) and high-risk (3-year RFS < 0.878) groups. Kaplan-Meier analysis revealed significantly lower 3- and 5-year recurrence-free survival rates in high-risk patients (79.80% [95% CI: 73.72–85.88%] and 74.00% [95% CI: 67.53–80.47%], respectively) compared to low-risk patients (both 91.90% [95% CI: 89.16–94.64%]) (P < 0.001). Similarly, overall survival was significantly reduced in the high-risk group (3-year: 80.90% [95% CI: 75.02–86.78%]; 5-year: 79.70% [95% CI: 79.64–85.78%]) versus the low-risk group (3-year: 95.40% [95% CI: 93.24–97.56%]; 5-year: 95.10% [95% CI: 92.94–97.26%]) (Tables 3, 4). This survival disparity was consistently validated in the independent cohort.

Utilizing the optimal cutoff value of 0.878 obtained from the model, patients in both the training and validation cohorts were categorized into high-risk (predicted risk probability ≥ 0.878) and low-risk (predicted risk probability < 0.878) groups. Kaplan-Meier analysis revealed significantly shorter RFS and OS in the high-risk group compared to the low-risk group (Figure 8).

To investigate the association between IGF2BP2 expression and key pathological markers in endometrial carcinoma, we compared p53, ER, and PR status between IGF2BP2 low- and high-expression groups. As shown in Supplementary Table S1, tumors with high IGF2BP2 expression exhibited a significantly higher frequency of abnormal p53 compared to the low-expression group 45.7% vs 30.4%, p=0.001). Furthermore, high IGF2BP2 expression was strongly associated with ER-negative (27.7% vs 14.2%, p<0.001) and PR-negative (33.5% vs 17.2%, p<0.001) status. These findings suggest that elevated IGF2BP2 expression may be linked to a hormone receptor-negative and p53-aberrant molecular phenotype in endometrial carcinoma.