Table 3.

Estimation of tree performances over several groups of patients, obtained from the colorectal and breast cancer cohorts

Specificity		Sensitivity
Estimation	95% CI	Estimation	95% CI
CRC decision tree
Training set—Patients used for tree building N = 527 (CRC + healthy)
0.87	[0.79–0.92]	0.76	[0.67–0.85]
Testing set—Patients not used for tree building N = 424 (CRC + healthy)
0.89	[0.84–0.94]	0.77	[0.73–0.80]
CRC stages 0/I/II versus healthy
0.87	[0.83–0.91]	0.72	[0.67–0.76]
Breast cancer patients N = 278 versus healthy
0.90	[0.84–0.95]	0.58	[0.50–0.65]
Breast cancer decision tree
Training set—Patients used for tree building N = 236 (breast cancer + healthy)
0.72	[0.65–0.84]	0.86	[0.80–0.91]
Testing set—Patients not used for tree building N = 46 (breast cancer + healthy)
0.80	[0.64–0.95]	0.95	[0.85–1]
CRC patients N = 951 versus healthy
0.66	[0.62–0.72]	0.85	[0.82–0.87]

a, The confidence intervals were obtained using the empirical bootstrap method.