All Times ET
Keywords: Recursive Partitioning, Bootstrap, Metastatic Breast Cancer, Cancer Staging
Currently, the American Joint Commission on Cancer (AJCC) staging guidelines subdivide non-metastatic breast cancer (MBC) patients into multiple groups, yet those with MBC are currently staged as a single group (stage IV). However, there is evidence that the prognosis of MBC patients can vary significantly. Therefore, we conducted a recursive partitioning analysis (RPA) among de novo MBC patients to determine if distinct prognostic groups exist. Adults diagnosed with de novo MBC (2010-2013) were selected from the National Cancer Database. RPA was then used to group patients with similar overall survival (OS) based on select clinical characteristics similar to those used to stage non-MBC: clinical T/N stage, tumor grade, receptor status, number of metastatic sites, and presence of bone-only metastases. P=0.10 was used as the primary splitting criteria. After RPA, 3-year OS rates were estimated for each terminal node, and nodes were amalgamated into 3 groups based on 3-year OS: (A) >50%, (B) 30-50%, and (C) <30%. A bootstrap simulation with 1000 iterations was then conducted around this RPA to better approximate the population distribution. Characteristic profiles were then assigned to final groupings based on the substage (A, B, or C) to which they were assigned most often. Kaplan-Meier analyses and log-rank tests were used to estimate differences in OS between substages. A single-institution sample was then used as a validation cohort.
RPA with bootstrapping resulted in identification of distinct stage IVA, IVB, and IVC groupings [5-year OS (95% CI): (A) 0.39 (0.37-0.41) vs (B) 0.23 (0.21-0.25) vs (C) 0.09 (0.07-0.11), log-rank p<0.001]. The validation cohort confirmed that application of these staging criteria resulted in distinct prognostic groups [5-year OS (95% CI): (A) 0.91 (0.76-0.96) vs (B) 0.91 (0.51-0.99) vs (C) 0.49 (0.27-0.68), log-rank p<0.001].
These findings have motivated the AJCC to reevaluate the MBC staging guidelines.