A risk stratification model to predict chemotherapy benefit in medullary carcinoma of the breast: a population-based SEER database.

Whether patients with medullary breast carcinoma (MBC) receive chemotherapy is controversial. Therefore, the aim of our study was to screen out patients with MBC who benefit from chemotherapy. We enrolled 618 consecutive patients with MBC from The Surveillance, Epidemiology, and End Results (SEER) database (2010-2018). Cox regression analysis was used to identify independent prognostic factors. Next, a nomogram was constructed and evaluated using calibration plots and the area under the curve (AUC) of receiver operating characteristic (ROC) curves. Kaplan‒Meier curves were used to evaluate the overall survival (OS) benefit of chemotherapy in different risk groups. A total of 618 MBC patients were involved in our study, and an 8:2 ratio was used to randomly split them into a training cohort (n = 545) and a validation cohort (n = 136). Next, a nomogram predicting 3- and 5-year OS rates was constructed based on the five independent factors (age at diagnosis, T stage, N status, subtype and radiation). The nomogram AUCs for 3- and 5-year OS (training set: 0.793 and 0.797; validation set: 0.781 and 0.823) and calibration plots exhibited good discriminative and predictive ability. Additionally, a novel risk classification system for MBC patients demonstrated that we do not have enough evidence to support the benefit effect of chemotherapy for the high-risk group as the result is not statistically significant (total population: p = 0.180; training set: p = 0.340) but could improve OS in the low-risk group (total population: p = 0.001; training set: p = 0.001). Our results suggested that chemotherapy should be selected more carefully for high-risk groups based on a combination of factors and that the possibility of exemption from chemotherapy should be confirmed by more clinical trials in the future.