Abstract
BACKGROUND: The increasing incidence of endometrial cancer (EC) has highlighted the need for improved early detection methods. This study aimed to develop and validate a novel DNA methylation classifier, EMPap, for EC detection using cervical scrapings. METHODS: EMPap incorporated the methylation status of BHLHE22 and CDO1, along with age and body mass index (BMI), into a logistic regression model to calculate the endometrial cancer methylation (EM) score for identifying EC in cervical scrapings. We enrolled 1297 patients with highly suspected EC, including 196 confirmed EC cases, and assessed the EMPap performance in detecting EC. RESULTS: EMPap demonstrated robust diagnostic accuracy, with an area under the curve of 0.93, sensitivity of 90.3%, and specificity of 89.3%. It effectively detected EC across various disease stages, grades, and histological subtypes, and consistently performed well across patient demographics and symptoms. EMPap correctly identified 87.5% of the type II ECs and 53.8% of premalignant lesions. Notably, compared with transvaginal ultrasonography (TVS) in patients with postmenopausal bleeding, EMPap exhibited superior sensitivity (100% vs. 82.0%) and specificity (85.2% vs. 38.5%). In asymptomatic postmenopausal women, EMPap maintained high sensitivity (89.5%) and negative predictive value (NPV) (98.3%). CONCLUSIONS: This study demonstrated the potential of EMPap as an effective tool for EC detection. Despite the limited sample size, EMPap showed promise for identifying type II EC and detecting over 50% of premalignant lesions. As a DNA methylation classifier, EMPap can reduce unnecessary uterine interventions and improve diagnosis and outcomes.