Purpose: The Eastern Cooperative Oncology Group (ECOG) entered 766 patients onto two prospectively randomized surgical adjuvant clinical trials for lymph node-positive breast cancer (T1-3N1M0). Ninety-five percent (n = 728) of eligible patients have complete information on the prognostic covariables under study (tumor necrosis [TN], tumor size, number of positive lymph nodes, age) and a median follow-up duration of 10.3 years.
Methods: TN was defined as confluent cell death in invasive areas of primary cancers, visible at 4 x objective lens magnification. Cox proportional hazards models were used to estimate presence versus absence of TN effects on clinical outcomes over full cross-stratification of variables, including delivery of chemotherapy versus observation only. Time-varying effects were modeled using spline functions of time, and by fixing proportional hazards models separately in the time periods 0 to 2 and 2+ years.
Results: Presence of TN was an independent predictor for time to recurrence (TTR) (P = .007) and for survival (P = .0003) in the overall 10-year follow-up period. Presence of TN was also an independent predictor for TTR and for survival (each P < .0001) in the period 0 to 2 years after diagnosis. Spline function time-modeling calculations showed different hazard ratios in the TN-present (TN+) versus TN-absent (TN-) groups for both TTR and survival (each P < .0001). This difference is changing over time (P = .0001 for TTR, P = .0005 for survival). Once a patient has been disease-free beyond 2 years after diagnosis, presence or absence of TN is irrelevant to future prognosis.
Conclusion: Confluent TN of any dimension in invasive areas of lymph node-positive breast cancer is an independent predictor for early recurrence and death from the disease.