Table 2. Summary of the performances for the 5- and 10-year follow-ups achieved by Model 1, Model 2, and Model 3 over the ten training sets after applying a 20 5-fold cross validation procedure.

Follow-up	Model	AUC (%)	Accuracy (%)	Sensitivity (%)	Specificity (%)
	Model 1	68.1 [66.8–69.4]	66.8 [65.9–68.1]	52.8 [50.8–55.0]	72.3 [71.1–73.6]
5 years	Model 2	94.8 [93.7–95.7]	88.8 [87.7–90.0]	81.9 [78.7–84.6]	91.2 [90.1–92.0]
	Model 3	83.8 [79.5–86.8]	75.8 [71.6–79.1]	77.1 [71.4–81.3]	73.7 [69.6–78.6]
10 years	Model 1	68.0 [66.7–69.6]	64.3 [62.9–65.3]	58.0 [55.9–60.1]	68.0 [66.3–69.4]
	Model 2	94.4 [93.9–95.5]	87.0 [85.6–88.2]	83.5 [81.5–85.2]	88.9 [87.4–90.3]
	Model 3	89.9 [86.0–92.3]	82.4 [78.5–85.1]	84.0 [79.6–87.6]	80.0 [73.5–84.4]

The performances are evaluated in percentage median values of AUC, accuracy, sensitivity, and specificity. The percentage first and third quartile values are also computed and reported in brackets.