Clinicopathological Model for Predicting Endometrial Cancer and Atypical Hyperplasia in Women Aged >40 Years: Development and Evaluation in a Single-Institution Retrospective Cohort

Background: Endometrial cancer poses a significant global health burden with rising mortality. Current diagnostics for women ≥40 with abnormal uterine bleeding or imaging abnormalities detect malignancy in <10% of biopsies, subjecting over 90% to unnecessary invasive procedures. Existing prediction models have suboptimal accuracy.To develop and validate a clinically practical nomogram incorporating the novel biomarker cumulative menstrual years, quantifying estrogen exposure, for predicting atypical endometrial hyperplasia or endometrial cancer risk. Methods: This retrospective cohort study included 1,490 women (aged >40 years) who underwent ≥ 2 endometrial biopsies at the International Peace Maternity and Child Health Hospital between 2014- 2023. Univariable and multivariable logistic regression were used to identify potential independent predictors of atypical endometrial hyperplasia or endometrial carcinoma ( AEH/EC ). A nomogram prediction model was developed using significant predictors, with its performance internally validated through AUC analysis (discrimination) and decision curve analysis (clinical utility). Results: Independt Risk factors were postmenopausal bleeding ≥5 years postmenopause (OR=14.55, 95% CI: 7.67–27.04), cumulative menstrual years>40 years (OR=7.28, 95% CI: 2.50–24.01), menstrual irregularity (OR=3.93, 95% CI: 1.74–7.99), abnormal endometrial thickness (OR=2.92, 95% CI: 1.70–5.27), and diabetes mellitus (paradoxical OR=0.40, 95% CI: 0.24–0.66). The nomogram demonstrated robust performance (training AUC=0.82; validation AUC=0.83), excellent calibration (slope=1.000), and clinical utility across thresholds (10–50%). Risk stratification thresholds: low (<40 points), medium (40–70 points), high (>70 points). Conclusion: This cumulative menstrual years integrated nomogram provides a practical, high-performance tool for dynamic AEH/EC risk stratification using routine parameters, while maintaining high sensitivity, particularly in resource-limited settings. The paradoxical protective association of diabetes (OR=0.40) requires cautious interpretation owing to incomplete BMI adjustment (dichotomized at 23 kg/m² without obesity stratification); prospective validation with granular metabolic profiling is warranted.