Diagnostic and triage performance of CDO1 and CELF4 methylation in cervical exfoliated cells compared with serum biomarkers for endometrial cancer in abnormal bleeding women

OBJECTIVE: To validate CDO1 and CELF4 methylation analysis in cervical scrapings as a non-invasive diagnostic tool for endometrial cancer (EC) in women with abnormal uterine bleeding (AUB), aiming to develop an evidence-based triage strategy that reduces unnecessary invasive procedures. METHODS: In this prospective cross-sectional study (January 2024-October 2025), 492 women with AUB underwent diagnostic evaluation before histopathology, including CDO1/CELF4 methylation testing in cervical scrapings, serum biomarker measurements, and transvaginal sonography (TVS)-based endometrial thickness (ET). Diagnostic performance was assessed by the area under the receiver operating characteristic curve (AUC). Multivariate logistic regression, incorporating multiple imputation for missing data, identified independent EC predictors, and screening-triage strategies were compared. RESULTS: Methylation-based assays outperformed conventional methods, with the combined CDO1/CELF4 test achieving the highest AUC (0.913) versus ET (0.608) and the best serum biomarker, human epididymis protein 4 (HE4; 0.752). After multiple imputation, CDO1 (odds ratio [OR]=30.0; 95% confidence interval [CI]=7.4-122.1) and CELF4 (OR=30.6; 95% CI=7.7-121.9) remained independent EC predictors, while HE4 lost significance. Methylation testing provided the most effective risk stratification, indicating a 75% EC risk when positive versus 4.5% when negative; in women with abnormal ET, risk was 76.7% if positive versus 7.6% if negative. CONCLUSION: CDO1 and CELF4 methylation analysis in cervical scrapings offers superior diagnostic and triage performance for EC compared with serum biomarkers and ET, enabling accurate risk identification and minimizing invasive procedures in low-risk women. As this strategy was derived in a high-prevalence, referral-based cohort, further validation in lower-prevalence primary care populations is needed before clinical implementation.