Nomogram for predicting the risk of locoregional recurrence in patients treated with accelerated partial-breast irradiation.

Abstract

59 Background: There are limited tools to guide clinicians regarding the risk of locoregional recurrence (LRR) in patients wishing to pursue APBI. The purpose of this study was to develop a nomogram taking into account clinicopathologic features to predict LRR in patients treated with APBI for early stage breast cancer. Methods: A total of 2,000 breasts (1,990 women) were treated with APBI at William Beaumont Hospital (N=551) or on the ASBrS MammoSite Registry Trial (N=1,449). Techniques included multiplanar interstitial catheters (N=98), balloon-based brachytherapy (N=1,689), and 3D conformal radiotherapy (N=213). Clinicopathologic variables were gathered prospectively. A nomogram was formulated utilizing the Cox Proportional Hazards Regression model to predict for LRR. This was validated by generating a bias-corrected index and cross-validated with a C-index. Results: Median follow-up was 5.5 years (0.9 to 18.3). Of the 2,000 cases, 435 were excluded due to missing data. Univariate analysis found that age <50, pre/perimenopausal status, close/positive margins, ER negativity, and high grade were associated with a higher frequency of LRR. These five independent covariates were used to create adjusted estimates, weighting each on a scale of 0 to 100. The total score is identified on a points scale to obtain the probability of an LRR over the study period. The model demonstrated good concordance for predicting LRR with a C-index of 0.641. Conclusions: The formulation of a practical, easy-to-use nomogram for calculating the risk of LRR in patients undergoing APBI will help guide the appropriate selection of patients for off-protocol utilization of APBI.