Surgical Margins and the Risk of Local-Regional Recurrence After Mastectomy Without Radiation Therapy

Abstract

Although positive surgical margins are generally associated with a higher risk of local-regional recurrence (LRR) for most solid tumors, their significance after mastectomy remains unclear. We sought to clarify the influence of the mastectomy margin on the risk of LRR. The retrospective cohort consisted of 397 women who underwent mastectomy and no radiation for newly diagnosed invasive breast cancer from 1998-2005. Time to isolated LRR and time to distant metastasis (DM) were evaluated by use of cumulative-incidence analysis and competing-risks regression analysis. DM was considered a competing event for analysis of isolated LRR. The median follow-up was 6.7 years (range, 0.5-12.8 years). The superficial margin was positive in 41 patients (10%) and close (≤2 mm) in 56 (14%). The deep margin was positive in 23 patients (6%) and close in 34 (9%). The 5-year LRR and DM rates for all patients were 2.4% (95% confidence interval, 0.9-4.0) and 3.5% (95% confidence interval, 1.6-5.3) respectively. Fourteen patients had an LRR. Margin status was significantly associated with time to isolated LRR (P=.04); patients with positive margins had a 5-year LRR of 6.2%, whereas patients with close margins and negative margins had 5-year LRRs of 1.5% and 1.9%, respectively. On univariate analysis, positive margins, positive nodes, lymphovascular invasion, grade 3 histology, and triple-negative subtype were associated with significantly higher rates of LRR. When these factors were included in a multivariate analysis, only positive margins and triple-negative subtype were associated with the risk of LRR. Patients with positive mastectomy margins had a significantly higher rate of LRR than those with a close or negative margin. However, the absolute risk of LRR in patients with a positive surgical margin in this series was low, and therefore the benefit of postmastectomy radiation in this population with otherwise favorable features is likely to be small.