Self‐Sampling for CDO1 / CELF4 Methylation Accurately Identifies Endometrial Cancer in a Resource‐Limited Region: A Prospective Study From Gansu

Timely diagnosis and risk stratification of endometrial cancer (EC) remain difficult in resource-limited regions. This prospective diagnostic accuracy and feasibility study evaluated whether self-collected cervical specimens combined with CDO1/CELF4 methylation testing, defined as the CISENDO assay, could identify histologically confirmed EC in symptomatic women from Gansu, northwest China. Of 302 screened women, 154 participants with self-collected specimens and histopathological diagnoses formed the primary analytic cohort; 35 had EC and 119 had non-EC diagnoses. An exploratory paired subset included 35 participants with both self- and physician-collected specimens. Diagnostic performance was assessed using receiver operating characteristic analysis and logistic regression, while paired sampling was evaluated using correlation, Bland-Altman analysis, and classification concordance. In self-collected specimens, CISENDO showed strong discrimination for EC at manufacturer-defined cutoffs (AUC 0.834). Bland-Altman mean differences were -1.05 ΔCt for CDO1 and -1.73 ΔCt for CELF4. CISENDO classifications were concordant in 31/35 paired specimens (88.6%), with positive agreement of 84.6%, negative agreement of 90.9%, and Cohen's κ of 0.76. CISENDO positivity remained the strongest independent predictor of EC after adjustment (adjusted OR 85.08; 95% CI 12.21-592.71; p < 0.001). Endometrial thickness showed modest discrimination (AUC 0.646) and did not improve performance when combined with CISENDO. Self-collected CISENDO testing showed strong discrimination for histologically confirmed EC and may provide a feasible molecular triage and risk-stratification approach for symptomatic women in underserved settings. This approach could support decentralized sample collection and reduce dependence on immediate specialist-based evaluation, particularly in geographically dispersed, resource-constrained populations.