Table 2.
Performances of SVM stacking models for predicting pCR to NAC in various molecular subtypes and cohorts.
| Molecular subtype | Cohort | AUC (95% CI) | ACC (%) | SEN (%) | SPE (%) | PPV (%) | NPV (%) |
|---|---|---|---|---|---|---|---|
| HR+/HER2− | PC | 0.959 (0.908–1.00) | 97.68 | 85.71 | 99.34 | 94.73 | 98.05 |
| VC-1 | 0.904 (0.817–0.991) | 93.28 | 75.00 | 95.76 | 70.58 | 96.58 | |
| VC-2 | 0.908 (0.776–1.00) | 94.54 | 71.42 | 97.91 | 83.33 | 95.91 | |
| VC-3 | 0.882 (0.779–0.985) | 82.94 | 75.00 | 83.76 | 32.14 | 97.02 | |
| HER2+ | PC | 0.974 (0.955–0.993) | 91.93 | 95.83 | 89.47 | 85.18 | 97.14 |
| VC-1 | 0.896 (0.842–0.949) | 85.62 | 85.10 | 86.36 | 89.88 | 80.28 | |
| VC-2 | 0.929 (0.881–0.978) | 86.31 | 88.13 | 83.33 | 89.65 | 81.08 | |
| VC-3 | 0.920 (0.876–0.965) | 87.66 | 83.54 | 92.00 | 91.66 | 84.14 | |
| TNBC | PC | 0.958 (0.906–1.00) | 92.00 | 91.66 | 92.30 | 91.66 | 92.30 |
| VC-1 | 0.873 (0.735–1.00) | 87.75 | 90.90 | 81.25 | 90.90 | 81.25 | |
| VC-2 | 0.901 (0.755–1.00) | 85.00 | 85.71 | 84.61 | 75.00 | 91.66 | |
| VC-3 | 0.837 (0.725–0.949) | 82.45 | 68.42 | 89.47 | 76.47 | 85.00 |
PC, primary cohort; VC, validation cohort; AUC, the area under curve; ACC, accuracy; SEN, sensitivity; SPE, specificity; NPV, negative predictive value; PPV, positive predictive value; 95% CI, 95% confidence interval; HR, hormone receptor; HER2, human epidermal growth factor receptor 2; TNBC, triple negative breast cancer; pCR, pathological complete response.