Background: Patients with metastases in four or more axillary lymph nodes (≥4+ALN) represent a subset of patients with breast cancer who are at increased risk of local recurrence and who benefit from postmastectomy radiation. Risk prediction models designed to identify such patients have been published by Rivers et al., Chagpar et al., and Katz et al. We sought to evaluate and compare the performance of these models in an independent patient population. Methods: We reviewed 454 patients with breast cancer with one to three positive sentinel lymph nodes who underwent completion axillary lymph node dissection at our institution. Each of the three published models was applied to our sample as described in the respective publications. The models' performances were analyzed with the Hosmer-Lemeshow goodness-of-fit test and with the area under the curve (AUC). Sensitivity, specificity, and false-negative percentages were calculated for clinically meaningful cutoff points of each score. Results: Of 454 eligible patients, 87 (19.2%) had four or more positive axillary nodes. The Rivers, Chagpar, and Katz models demonstrated good calibration in our population based on the Hosmer-Lemeshow test (p = 0.82, p = 0.73, p = 0.71, respectively). Assessment of discriminatory ability for the models resulted in AUCs of 0.81, 0.73, and 0.81, respectively. Conclusions: The Rivers and Katz models performed well in our patient population and may be clinically useful to predict patients with ≥4+ALN. However, their clinical utility is limited by the current controversy surrounding the use of postmastectomy radiation for all node-positive patients.
ASJC Scopus subject areas