91 DNA methylation testing for endometrial cancer detection in patient-friendly sample types Figure 3 visualizes the methylation levels of all individual markers per sample for all patients and controls individually using the predicted probabilities computed during the ROC curve analyses. Samples of controls showed predominantly low methylation levels (green), whereas samples of patients with endometrial cancer showed high methylation levels (red). Paired endometrial cancer cases were stratified by histological subtype. Within the non-endometrioid cancers, increased methylation was predominantly seen in CDO1 and GHSR, followed by BHLHE22. The marker panels formed by multivariable logistic regression detected more cancers than the single markers. Samples were classified as positive (black box) based on the Youden’s Index (J) thresholds of the threemarker panels calculated for urine (≥0.40), self-samples (≥0.46) and scrapes (≥0.34). Using this threshold, 90% (93/103), 89% (92/103), and 93% (96/103) of the cancers were classified as cases and 10% (10/100), 8% (9/107) and 10% (11/110) of the controls were classified as cases in the urine, self-samples and scrapes, respectively. DISCUSSION Our study presents the diagnostic potential of DNA methylation testing in minimally- and non-invasive samples for endometrial cancer detection. The diagnostic performance of nine DNA methylation markers and most optimal three-marker panels for endometrial cancer detection were evaluated and compared between the different sample types. Endometrial cancer detection in samples collected by home-based methods was excellent and comparable to the diagnostic performance of methylation testing in clinician-taken cervical scrapes. Three-marker combinations yielded an AUC value of 0.95 (95% CI: 0.92-0.98), 0.94 (0.90-0.97) and 0.97 (0.96-0.99), for endometrial cancer detection in urine, self-samples and scrapes, respectively. Virtually equal performances were obtained after cross-validation. Excellent diagnostic performances were maintained in stage I endometrial cancers, confirming the ability to detect endometrial cancer at its earliest stage. Our study emphasizes the outstanding potential of DNA methylation analysis using patient-friendly home-based sample collection methods for endometrial cancer detection. Several discovery screens identified valuable hypermethylated genes as biomarker candidates for endometrial cancer detection in minimally invasive specimens (17-19, 21, 23). The markers ADCYAP1 (18), BHLHE22 (17), CDH13 (18), CDO1 (17, 24, 32), GALR1 (21), and HAND2 (23, 24) tested in our study originate from such discovery screens carried out by different research groups. Except for CDO1 and HAND2, none of these markers have been independently validated before for endometrial cancer detection in minimally- or non-invasive samples. All genes showed a significant difference in 4
RkJQdWJsaXNoZXIy MjY0ODMw