Risk of endometrial malignancy in women treated for breast cancer: the BLUSH prediction model - evidence from a comprehensive multicentric retrospective cohort study.

Abstract

Objective: This study aimed to evaluate characteristics of endometrial surveillance in women treated for breast cancer to build a clinical prediction model.

Design: A multicentric retrospective cohort study was conducted at two tertiary-care university hospitals from January 2020 to June 2023. Perimenopausal and postmenopausal women treated for breast cancer were categorized into two groups: patients with and without diagnosis of endometrial malignancy (endometrial carcinoma) or premalignancy (atypical endometrial hyperplasia). Characteristics of breast cancer and ultrasonographic and hysteroscopic examinations were compared. A prediction model for endometrial malignancy was built using logistic regression. Predictive accuracy was assessed using the receiver operating characteristic (ROC) curve and goodness of fit using the Hosmer-Lemeshow test.

Results: One hundred and thirty-two (28 with premalignancy or malignancy and 104 without malignancy) women were analyzed. A nomogram was produced for prediction model development utilizing the presence and duration in months of abnormal uterine (BL)eeding, ultrasound (US) vascular pattern and echogenicity and (H)ysteroscopic appearance of endometrium (BLUSH) as determined by logistic regression. Sensitivity and specificity were 79.17% and 95.19%, respectively, with an area under ROC curve of 0.965, indicating good accuracy. Good goodness of fit and prediction stability were indicated by the calibration curve and Hosmer-Lemeshow test (χ² = 26.36; p = 0.999).

Conclusions: Breast cancer survivors undergoing endometrial surveillance might benefit from a potentially useful prediction model based on hysteroscopic appearance, ultrasonographic uniformity of endometrium, Doppler flow and presence of abnormal uterine bleeding.