157 Detection of non-metastatic non-small cell lung cancer in urine Validation by LOOCV yielded similar AUCs of 0.64, 0.72, and 0.56 for CDO1, SOX17, and TAC1, respectively. Sensitivities and specificities based on the maximal Youden’s Index (J) threshold, varied from 0.48 to 0.68 and 0.66 to 0.86, respectively (Table 2). To evaluate potential complementarity between markers, multivariate logistic regression with backward selection was used (Figure 2B). The backward selection rejected TAC1 from the final model, yielding an AUC of 0.78 (95% CI: 0.68-0.87) for CDO1 and SOX17 combined. Upon validation by LOOCV, an AUC of 0.71 was achieved, with a sensitivity of 0.55 and specificity of 0.86 based on a ‘believe-the-positive’ algorithm (Table 2). A B Specificity Sensitivity 1.0 0.8 0.6 0.4 0.2 0.0 0.0 0.2 0.4 0.6 0.8 1.0 CDO1 AUC = 0.68 SOX17 AUC = 0.72 TAC1 AUC = 0.58 Specificity Sensitivity 1.0 0.8 0.6 0.4 0.2 0.0 0.0 0.2 0.4 0.6 0.8 1.0 CDO1+SOX17 AUC = 0.78 Figure 2: Diagnostic potential of individual methylation markers and marker combination. Non-CV receiver operator characteristic (ROC) curves of CDO1, SOX17, TAC1 (A) and CDO1/SOX17 combined (B). Results of individual markers and the marker combination are quantified by the area under the curve (AUC) value. Non-CV = non-cross-validated. Table 2: Univariable logistic regression analysis on diagnostic performance of the three individual markers (CDO1, SOX17, TAC1) and multivariable logistic regression analysis on diagnostic performance of the optimal marker combination (CDO1+SOX17) for NSCLC detection. Methylation marker(s) CDO1 SOX17 TAC1 CDO1 + SOX17 AUC (non-CV; 95% CI) 0.68 (0.54-0.76) 0.72 (0.61-0.83) 0.58 (0.46-0.69) 0.78 (0.68-0.87) Sensitivity 0.68 0.57 0.48 0.57 Specificity 0.66 0.86 0.68 0.86 AUC (LOOCV) 0.64 0.72 0.56 0.71 Sensitivity 0.68 0.55 0.46 0.55 Specificity 0.64 0.82 0.58 0.86 Non–CV AUC values of individual markers and marker combination CDO1 + SOX17, including 95% CI , are reported together with sensitivity and specificity based on the Youden’s Index (J) threshold. LOOCV AUC values are reported together with sensitivity and specificity based on a ‘believe-the-positive’ algorithm. AUC = area under the receiver operating characteristic curve, CI = confidence interval, LOOCV = leaveone-out cross-validated, non-CV = non-cross-validated. 6
RkJQdWJsaXNoZXIy MjY0ODMw