Fig. 2

Association of the nine-protein prognostic score with breast cancer-specific survival. A-B) Cumulative incidence of death due to breast cancer according to the 9-protein prognostic score in the discovery (A) and validation (B) cohorts. Maximally selected rank statistics was used in the discovery cohort to determine the optimal cut-off of the score (high vs. low), and the same cut-off was used for the validation cohort. C-D) Multivariable regression analysis of breast cancer-specific survival in the discovery cohort (n = 413) and validation cohort (n = 108), including the nine-protein prognostic score and the most powerful traditional prognostic factors of early-stage breast cancer. In the models, death from causes other than breast cancer was treated as a competing risk. ER = estrogen receptor, SHR = subdistribution hazard ratio, CI = confidence interval E-F) Time-dependent receiver operating characteristic curves in the discovery (E) and validation (F) cohorts. The area under the curve (AUC) was higher in both cohorts at all timepoints when the nine-protein signature was added to the clinical factors (tumor size, nodal status, grade, estrogen receptor)