A predictive model of survival in early-stage medullary breast carcinoma: a population-based study.

INTRODUCTION This study aimed to establish a predictive model for overall survival (OS) in early-stage medullary breast carcinoma (MedBC).

Methods

Clinical data for 606 355 patients with breast cancer were extracted from the Surveillance, Epidemiology, and End Results (SEER) database, along with data for 244 patients with MedBC from electronic medical records of the National Cancer Center of China. Univariate and multivariable Cox regression analyses were conducted to identify prognostic variables. Survival outcomes and model performance were evaluated.

Results

In total, 650 and 188 patients were included in the SEER and real-world cohorts, respectively. In the real-world cohort, the median Ki-67 index was 70%. Overall survival was not significantly associated with Ki-67 index (P=0.09) or family history of breast cancer (P=0.1). A predictive model for OS was established in the training cohort, incorporating age, tumour (T) stage, nodal (N) stage, and chemotherapy. The overall C-index was 0.76. Area under the curve (AUC) values were 0.83, 0.76, 0.75, and 0.76 for 1-, 3-, 5-, and 10-year OS, respectively. In the internal validation cohort, the C-index was 0.86; AUC values were 0.78 and 0.83 for 3- and 5-year OS, respectively. In the external validation cohort, the C-index was 0.90; AUC values were 0.85, 0.85, and 0.84 for 3-, 5-, and 10-year OS, respectively. The low-risk group had significantly better OS than the high-risk group across training and validation cohorts (P<0.05).

Conclusion

A clinical model incorporating age (≥65 years), T stage (T3/T4), N stage (N2/N3), and chemotherapy (no) was developed to predict inferior OS in early-stage MedBC.