A prognostic model for predicting breast cancer (BC)-related survival in operable triple-negative (TN) patients (pts).

Abstract

1049 Background: TNBC represent a heterogeneous disease in terms of biology, prognosis, and treatment response. We propose a prognostic model to identify homogeneous subgroups of patients and tailor risk-adapted adjuvant therapies indications. Methods: We analyzed 1,049 pts operated in our institute from 1997 to 2007 for early TNBC. Pts who received neoadjuvant chemotherapy (CT), with T4 tumors or previous history of cancer were excluded. Death from BC was the primary endpoint of the study. We calculated an individual predicted risk using a multivariable Cox regression model, with age, tumor size, number of positive lymph nodes and Ki-67 analyzed as continuous covariates, and tumor grade and perivascular invasion as categorical covariates. Results: Median age was 52 years, 562 (53.4%) and 670 (65.1%) pts had a pT1 and pN0 TNBC, respectively. Median Ki-67 was 48%. Adjuvant CT regimens were distributed as follows: classical CMF 388 (37.0%), anthracycline containing regimens 455 (43.4%), taxanes 12 (1.1%), other regimens 66 (6.3%) and no CT 128 (12.2%). After a median follow-up of 6 years, 131 deaths from BC were observed (5-year cumulative incidence 11.9%). At multivariable analysis, age, tumor size, number of positive lymph nodes, Ki-67, tumor grade and perivascular invasion were associated with the risk of death and were included in the prognostic model. Its predictive accuracy was good (C-index 0.73). We subsequently identified three homogeneous prognostic subgroups - low, medium and high-risk - according to the tertiles values of the predicted risk. The outcomes are shown in the table. Conclusions: We could identify homogeneous prognostic subgroups of TNBC pts according to clinical-pathological features. This prognostic model suggests that the use of CT in TN low-risk pts might be questionable. We are currently externally validating this model on a different series of pts. [Table: see text]