Table 3.
Metrics to assess the performance of the proposed methods evaluated in 219 endometrioid EC patients preoperatively classified as early stage (clinical stage).
| Metrics | CA125 | HE4 | RERT with CV | Logistic Regression | RT with CV |
|---|---|---|---|---|---|
| ROC-AUC | 0.58*** | 0.72* | 0.84 | 0.57*** | 0.66** |
| Threshold (Youden Index) | 13.55 | 80.70 | 0.29 | 0.15 | 0.18 |
| Specificity | 0.39 | 0.66 | 0.67 | 0.76 | 0.69 |
| Sensitivity | 0.76 | 0.75 | 0.84 | 0.41 | 0.67 |
| Accuracy | 0.48 | 0.67 | 0.71 | 0.68 | 0.68 |
| PPV | 0.28 | 0.39 | 0.44 | 0.34 | 0.40 |
| NPV | 0.85 | 0.89 | 0.93 | 0.81 | 0.87 |
*p-value = 0.02, **p-value < 0.01, ***p-values < 0.0001: these are p-values of the DeLong’s test or Bootstrap test (underlined) for the comparison of two AUCs (RERT with CV vs other methods). For major details see Table S7 in Supplementary Tables. RT with CV stands for Regression Tree with Cross-Validation.