Research Article

Comparison of Different Machine Learning Models in Prediction of Postirradiation Recurrence in Prostate Carcinoma Patients

Table 3

Results of the ROC analysis for nadir PSA, time to nadir PSA, PSA doubling time, PSA velocity, PSA in the moment of disease reevaluation, and relevant events.

CharacteristicsNadir PSATime to nadir PSAPSA doubling timePSA velocityPSA in the moment of disease reevaluation

AUC ROCa (95% CI)64.3% (51.7-76.9%)74.2% (62.6-85.8%)87.2% (78.8-95.5%)91.8% (86.0-97.6%)97.1% (93.3-100%)
Likelihood ratio testb
ROC cut-off valuec0.30956.55.050.556.49
Sensitivity (95% CI)63.0% (50.0-75.9%)66.7% (53.7-77.8%)79.6% (68.5-90.7%)72.2% (59.2-83.3%)94.4% (89.0-100%)
Specificity (95% CI)68.0% (48.0-84.0%)80.0% (64.0-96.0%)84.0% (68.0-96.0%)92.0% (80.0-100%)92.0% (80.0-100%)

aArea under the ROC curve (DeLong’s method); bLikelihood ratio test for AUC ROC; cValue with maximum sensitivity and specificity.