Table 2.
Overall performance of risk models 1.
| Performance Metric | All Sites (N = 1734 Cases & 120,822 Non-Cases) | |||||||
|---|---|---|---|---|---|---|---|---|
| BCRAT | BCSC | BRCAPRO | BRCAPRO+BCRAT | |||||
| Estimate | 95% CI | Estimate | 95% CI | Estimate | 95% CI | Estimate | 95% CI | |
| O/E | 1.036 | 0.989, 1.084 | 1.185 | 1.130, 1.239 | 1.076 | 1.027, 1.125 | 1.063 | 1.014, 1.112 |
| AUC 1 | 0.604 | 0.590, 0.618 | 0.617 | 0.603, 0.630 | 0.590 | 0.578, 0.603 | 0.608 | 0.594, 0.621 |
| TPR | 0.378 | 0.354, 0.402 | 0.307 | 0.284, 0.328 | 0.358 | 0.338, 0.378 | 0.347 | 0.324, 0.368 |
| FPR | 0.238 | 0.235, 0.240 | 0.184 | 0.181, 0.186 | 0.243 | 0.241, 0.245 | 0.217 | 0.215, 0.219 |
| Patients with high 5-year risk, N (%) |
29,396 (23.99) | - | 22,751 (18.56) | - | 30,016 (24.49) | - | 26,830 (21.89) | - |
1 p-values for difference in AUC: BCRAT vs. BCSC p = 0.010, BCRAT vs. BRCAPRO p = 0.007, BCRAT vs. BRCAPRO+BCRAT p = 0.040, BCSC vs. BRCAPRO p < 0.001, BCSC vs. BRCAPRO+BCRAT p = 0.071, BRCAPRO vs. BRCAPRO+BCRAT p < 0.001. Bonferroni corrected p-values remain significant for BCSC vs. BRCAPRO and BRCAPRO vs. BRCAPRO+BCRAT.