Integrated Signature of Multiple Indices for Preoperative Predictive Stratification of Endometrial Hyperplasia Subtypes: A Retrospective Cohort Study.

Objective: To develop a comprehensive predictive model incorporating multi-dimensional parameters for the accurate preoperative stratification of endometrial hyperplasia subtypes. Methods: A retrospective cohort study was conducted involving 1822 patients with endometrial lesions. We gathered candidate variables spanning demographic, reproductive, hematological, hepatic and renal, coagulation, ultrasonographic, and pathological domains. Utilizing a two-step feature selection approach (variance thresholding combined with lasso regression), we constructed and compared three logistic regression models. Results: Sixteen key features were identified, encompassing pathological, clinical, and inflammatory indicators. Especially, the integrated model (Model 3) demonstrated the highest diagnostic efficacy, with an AUC of 0.923 (training set: sensitivity 87.2%, specificity 85.6%) and 0.905 (validation set: sensitivity 84.5%, specificity 83.2%), significantly outperforming the single-domain model. Conclusion: The predictive model, developed based on multidimensional clinical indicators, effectively categorized the preoperative subtypes of endometrial hyperplasia, thereby offering a practical tool for clinical decision-making and underscoring the pivotal role of inflammation in the progression of pathological lesions.