P2-12-15: GAINS: A Breast Cancer Prognostic Index Utilizing Lymph Node Ratio.

Abstract

Abstract Introduction: Axillary lymph node status is an important predictor of overall survival (OS), hence its inclusion in clinical prognostic tools. The Nottingham Prognostic Index (NPI) which incorporates Lymph Node (LN) stage, tumor size and grade generates a score which predicts a percentage 10-year survival. Despite its status as a benchmark model for breast cancer prognosis, newer prognostic factors do exist. Lymph Node Ratio (LNR) is a superior prognostic indicator compared to absolute positive lymph node number, warranting re-evaluation of breast cancer prognostication. The aims of this study were threefold: identify the strength of LNR as a prognostic indicator compared to LN stage and NPI; establish a new prognostic index (Galway Index of Survival [GAINS]), taking into account the effect of LNR and breast cancer subtype on traditional clinicopathological features in breast cancer prognostication; and evaluate the prognostic efficacy of the new index compared to NPI. Methods: Two cohorts were used: Galway Cohort-a prospectively compiled cohort of 1668 cases with histologically proven Stage 1, 2 and 3 primary operable breast cancer treated between 1990–2010 in a single institution; and ONCOPOOL-a retrospectively compiled database of 16944 cases treated across 12 European breast cancer units between 1990–1999. A Cox Proportional Hazards model was fitted to evaluate the strength of LNR compared to LN stage (within the NPI model) in both cohorts. The effect of clinicopathological variables on OS was analyzed using multivariable analysis in the Galway cohort. Three models were created (Model 1: Traditional prognostic variables excluding NPI and LNR; Model 2: Model 1 and LNR; Model 3: Model 1 and NPI) and compared using Likelihood-ratio tests. Stepwise variable selection was used to identify the best model to create a prognostic index and performance of the two indices was evaluated using Receiver Operating Curves (ROC). Results: Controlling for tumor size and histological grade, LNR is a stronger prognostic factor than LN Stage in both the Galway (β values of 1.2 and 0.3 respectively) and ONCOPOOL (b values of 1.3 and 0.3 respectively) cohorts, with LNR rendering LN stage non significant (p=0.135) in the former. In the Galway cohort, separate comparisons of Model 2 and 3 with Model 1 demonstrated that traditional clinicopathological variables in addition to LNR (Model 2) best predicted OS (p=0.019). Within Model 2, LNR was a significant predictor of survival (p=0.014, Hazard Ratio 7.4). Age, LNR, stage and molecular subtype were significant prognostic factors, and corresponded to distinct survival patterns when included in the new prognostic index. GAINS performed almost identically to NPI, with similar areas under the curve (AUC) (GAINS AUC=0.745 [95%CI0.67−0.82]; NPI AUC=0.742 [95%CI0.67−0.82]). Conclusion: LNR is a better prognostic indicator compared to LN Stage. GAINS performs just as well as NPI as a prognostic index and has the potential for clinical utility given further validation. Citation Information: Cancer Res 2011;71(24 Suppl):Abstract nr P2-12-15.