Objective:

To characterize clinical and metabolic profiles across pathological subgroups and develop a risk-stratification model for endometrial neoplasia in women with insulin-resistant PCOS and suspected endometrial polyps, focusing on insulin resistance as a core pathophysiological driver.

Approach:

• Study Design: Retrospective cohort study involving 185 PCOS-IR patients with ultrasound-detected endometrial polyps, stratified into benign, atypical hyperplasia, and endometrial carcinoma groups.
• Model Development: A two-stage strategy combining LASSO regression and stepwise logistic regression was used for variable selection and model construction.
• Model Assessment: Model performance was evaluated using ROC curve, calibration curve, decision curve analysis, and multicollinearity diagnostics.

Key Findings:

• Endometrial neoplasia group showed significantly higher HOMA-IR, fasting plasma glucose, 2-hour OGTT glucose, and fasting insulin, and lower HDL-C compared to the benign group (P<0.05).
• The final model included age, HDL-C, free androgen index (FAI), and HOMA-IR, achieving an AUC of 0.767 with high sensitivity (0.913) and specificity (0.500).
• All variables had VIF values <5, indicating no significant multicollinearity.

Interpretation:

Insulin resistance is linked to early metabolic alterations in PCOS-IR patients with endometrial polyps and is a core component of the predictive model.

Limitations:

• The model lacks external validation.
• The combined neoplasia endpoint shows heterogeneity.

Conclusion:

The four-variable model demonstrates moderate discriminatory performance as an exploratory tool for pre-hysteroscopy risk stratification, requiring further external validation.