Development, verification, and comparison of a risk stratification model integrating residual cancer burden to predict individual prognosis in early-stage breast cancer treated with neoadjuvant therapy

A favorable model for predicting disease-free survival (DFS) and stratifying prognostic risk in breast cancer (BC) treated with neoadjuvant chemotherapy (NAC) is lacking. The aim of the current study was to formulate an excellent model specially for predicting prognosis in these patients.Between January 2012 and December 2015, 749 early-stage BC patients who received NAC in Xijing hospital were included. Patients were randomly assigned to a training cohort (n = 563) and an independent cohort (n = 186). A prognostic model was created and subsequently validated. Predictive performance and discrimination were further measured and compared with other models.Clinical American Joint Committee on Cancer stage, grade, estrogen receptor expression, human epidermal growth factor receptor 2 (HER2) status and treatment, Ki-67 expression, lymphovascular invasion, and residual cancer burden were identified as independent prognostic variables for BC treated with NAC. The C-index of the model consistently outperformed other available models as well as single independent factors with 0.78, 0.80, 0.75, 0.82, and 0.77 in the training cohort, independent cohort, luminal BC, HER2-positive BC, and triple-negative BC, respectively. With the optimal cut-off values (280 and 360) selected by X-tile, patients were categorized as low-risk (total points ≤280), moderate-risk (280 < total points ≤ 360), and high-risk (total points >360) groups presenting significantly different 5-year DFS of 89.9%, 56.9%, and 27.7%, respectively.In patients with BC, the first model including residual cancer burden index was demonstrated to predict the survival of individuals with favorable performance and discrimination. Furthermore, the risk stratification generated by it could determine the risk level of recurrence in whole early-stage BC cohort and subtype-specific cohorts, help tailor personalized intensive treatment, and select comparable study cohort in clinical trials.