Background: This study aims to build nomograms to predict overall survival (OS) and breast cancer-specific death (BCSD) in resected nonmetastatic invasive breast cancer.
Patients and methods: Patients extracted from surveillance, epidemiology, and end results database between 2010 and 2014 were analyzed. Through multivariate Cox regression and Fine and Gray competing risks regression, independent predictive factors were identified and integrated to build nomograms for predicting OS and BCSD. The models were validated by bootstrap resampling and an independent cohort. Additionally, the models' performance was measured by the Harrell's C-index, calibrate curve, and time-dependent receiver operating characteristic (ROC) curves.
Results: In total, 110,180 cases were identified and enrolled in the analysis, with 83,450 in the training cohort and 26,730 in the validation cohort. Several independent predictive factors for OS and BCSD were identified and integrated to construct the nomograms. The C-indexes in the training cohort and validation cohort were 0.759 and 0.772 for predicting OS, and 0.857 and 0.856 for predicting BCSD, respectively. The nomogram models were well calibrated, and the time-dependent ROC curves verified the superiority of our models for clinical usefulness. Significant differences in the OS and BCSD curves were also observed when stratifying patients into several different risk groups. For convenient access, we deployed these proposed nomograms into web-based calculators.
Conclusions: We established and validated novel nomograms for individualized prediction of OS and BCSD in resected nonmetastatic invasive breast cancer. These nomograms perform better than previous models and could be easily accessed easily by clinicians.
© 2021. Society of Surgical Oncology.