Nomograms predicting all-cause death and cancer-specific death in patients with bilateral primary breast cancer: a study based on Surveillance, Epidemiology, and End Results

Abstract

ABSTRACT Bilateral primary breast cancer (BPBC) patients have a worse prognosis. Tools for accurately predicting mortality risk in patients with BPBC are lacking in clinical practice. We aimed to develop a clinically useful prediction model for the death of BPBC patients. A total of 19,245 BPBC patients from the Surveillance, Epidemiology, and End Results (SEER) database between 2004 and 2015 were randomly divided into the training set (n = 13,471) and test set (5,774). Models for predicting the 1-, 3- and 5-year death risk of BPBC patients were developed. Multivariate Cox regression analysis was used to develop the all-cause death prediction model, and competitive risk analysis was used to establish the cancer-specific death prediction model. The performance of the model was assessed by calculating the area under the receiver operating characteristic curve (AUC) with 95% confidence interval (CI), sensitivity, specificity and accuracy. Age, married status, interval time and first and second tumor’s status were associated with both all-cause death and cancer-specific death (all P < 0.05). The AUC of Cox regression models predicted 1-, 3- and 5-year all-cause death was 0.854 (95% CI, 0.835–0.874), 0.838 (95% CI, 0.823–0.852) and 0.799 (95% CI, 0.785–0.812), respectively. The AUC of competitive risk models to predict 1-, 3- and 5-year cancer-specific death was 0.878 (95% CI, 0.859–0.897), 0.866 (95% CI, 0.852–0.879) and 0.854 (95% CI, 0.841–0.867), respectively. Nomograms were developed to predict all-cause death and cancer-specific death in BPBC patients, which may provide tools for clinicians to predict the death risk of BPBC patients.