| Stage |
Number |
Diagnostic parameters for CRC detection |
Median probability |
p-value*1 |
| |
CRC |
Control |
AUC (95%CI) |
sensitivity
(%) |
specificity
(%) |
cutoff |
CRC |
Control |
CRC/
Control |
|
| II+IIIa / IIIb+IV |
72 |
63 |
0.924 (0.876-0.962) |
83 |
92 |
0.567 |
0.898 |
0.132 |
6.80 |
2.40 × 10-17 |
| II+IIIa |
50 |
63 |
0.917 (0.861-0.962) |
80 |
92 |
0.567 |
0.906 |
0.132 |
6.86 |
2.92 × 10-14 |
| IIIb+IV |
22 |
63 |
0.937 (0.878-0.993) |
91 |
92 |
0.567 |
0.890 |
0.132 |
6.74 |
1.20 × 10-9 |
*1 Mann-Whitney's U-test. p-value< 0.05 was considered statistically significant.
The multiple logistic regression model was constructed with 2–5 peptides randomly selected from 14 candidate peptides. After examination of 15,302 combinations of the
peptide sets, an optimized model (equation 1 in Results) was obtained with the highest AUC value (0.924), suggesting it was sufficient to discriminate between CRC patient
and control groups. Using the same model, the diagnostic parameters for the subsets of CRC groups were also calculated. |
