Article Text
Abstract
Introduction Endometrial cancer (EC) is one of the most common gynecological cancer with rising incidence and mortality worldwide. Early detection is essential to reduce the morbidity related to extensive and aggressive treatments. Herein, we developed and validated an accurate and minimally invasive screening test for EC early detection.
Methods 22 previously published EC-specific differentially methylated regions (DMRs) were validated via targeted bisulfite sequencing (TBS) with 64 endometrial brush samples (with a ‘Tao brush’) from EC patients (n=36), atypical endometrial hyperplasia patients (n=7), and benign or healthy controls (n=21). Diagnostic performances were evaluated by methylight among the combinations of 3 DMRs chosen from a set of 6 regions. Further, an EC diagnosis (EC-D) model based on logistic regression was developed using 768 Tao brush samples from women with and without EC, and validated using two different settings (retrospective case/control: n=384; prospective cohort of women with suspected EC: n=1179).
Results In the both prospective set and retrospective set, our EC-D model achieved promising accuracy. Specifically, in prospective cohort, EC-D achieved an AUC of 96.49% with a sensitivity of 95.38% at a specificity of 92.21%, which far outperformed transvaginal ultrasound (TVUS). More importantly, our EC-D model performed robustly among early and late stages of endometrial cancers or endometrial cancer in women pre- or post-menopause with the consensus cutoff.
Conclusion/Implications Our EC-D model based on Tao specimens showed fair diagnostic accuracy in a large pooled cohort with very minimally invasive procedure. This test might provide an alternative for EC early detection and has great potential to screen EC and AEH.