Recursive partitioning analysis of locoregional recurrence patterns following mastectomy: Implications for adjuvant irradiation

Purpose: Postmastectomy irradiation improves overall survival for breast ca ncer patients at high risk for locoregional recurrence (LRR). The objective of this study was to use recursive partitioning analysis (RPA) to define p atient subgroups at high risk for LRR following mastectomy. Patients and Methods: A cohort of 1031 patients treated on prospective tria ls with mastectomy and doxorubicin-based chemotherapy without irradiation w as analyzed. The variables considered in the RPA were tumor size, number of involved nodes, number of nodes examined, and percentage of nodes involved (nodes involved/nodes examined). The endpoint was LRR +/- distant metastas is. Only patients with complete data were analyzed (n = 913). Median follow -up was 8 years (range, 0.7-22 years). Results: Involvement of 20% or more of the lymph nodes examined was the mos t significant variable predicting LRR. Three risk categories were defined. Patients with 20% or more involved nodes and tumors of 3.5 cm or more were at greatest risk for LRR (41% at 8 years). An intermediate-risk group inclu ded patients,vith 20% or more involved nodes and tumors of less than 3.5 cm as well as those with less than 20% involved nodes and tumor size of 5 cm or greater (18% at 8 years). Patients with less than 20% involved nodes and tumor size of less than 5 cm were at lowest risk for LRR (10% at 8 years). Conclusion: Tumor size and extent of nodal involvement play interrelated ro les in predicting LRR following mastectomy and systemic therapy. Patients w ith 20% or greater involved nodes and those with less than 20% nodes and tu mors of 5.0 cm or greater are at significant risk of LRR and should be cons idered for postoperative irradiation. (C) 2001 Elsevier Science Inc.