The aim of this analysis was to investigate the usefulness of Ki-67 labeling index (LI) for the identification of different prognostic subgroups in primary node-negative, triple negative breast cancer (TNBC) patients. From January 1997 to December 2005, 1,053 patients operated for TNBC were identified through the institutional clinical database. The study was performed in accordance with REMARK criteria. The relationship between Ki-67LI and the risk of breast-related deaths was evaluated with a multivariable Cox regression model. Cubic splines were used to model Ki-67LI as a continuous variable. We selected 496 consecutive patients with node-negative TNBC. Median age was 52 years, median Ki-67LI 48% (range 4-95), and median follow up 6 years (range 0.5-13). Total deaths and deaths from BC were 52 (10.5%) and 38 (7.7%), respectively. Ki-67LI increased with decreasing age (P<0.01), increasing tumor size (P<0.01), and grade (P<0.01). When analyzing Ki-67LI as a continuous variable, the risk of death from BC increased steeply with increasing Ki-67LI up to about 35% and remained flat for higher values (adjusted effect of Ki-67 P=0.049; adjusted nonlinear effect P=0.021). Accordingly, when dividing patients into lower (≤35%) and higher (>35%) Ki-67LI subgroups, the 5-year cumulative incidence of breast-related deaths were 2.3 and 9.0%, respectively, with an adjusted HR(>35 vs ≤35) of 2.3 (95% CI 1.0-5.8, P=0.046). Within the group of patients with node-negative TNBC, Ki-67LI was associated with different prognoses subgroups. Ki-67LI might be useful in the design of trials of risk-adapted adjuvant therapies.