1148 - Validated Nomogram Tool Predicting Individualized Risks of Invasive Disease Recurrence for Breast Cancer Patients with N1mic Treated with and without Nodal Radiotherapy
Presenter(s)
G. E. Naoum1, L. A. Boe2, A. M. Shui3, L. Z. Braunstein2, A. J. Xu1, D. A. Roth O’Brien1, M. B. Bernstein1, Z. Abou Yehia1, B. A. Mueller1, J. J. Cuaron1, Q. LaPlant1, B. McCormick1, S. N. Powell1, A. G. Taghian4, and A. J. Khan1; 1Department of Radiation Oncology, Memorial Sloan Kettering Cancer Center, New York, NY, 2Memorial Sloan Kettering Cancer Center, New York, NY, 3Massachusetts General hospital, Boston, MA, 4Harvard Medical School, Boston, MA
Purpose/Objective(s): Breast cancer patients with nodal micrometastatic (N1mic) disease represent a rare clinical grey zone. The benefit of regional nodal irradiation (RNI) after lumpectomy or postmastectomy radiotherapy (PMRT) remains undefined, unaddressed in NCCN guidelines and EBCTCG metanalysis and largely extrapolated from higher-risk populations. Therefore, we developed and externally validated a nomogram to estimate individualized invasive disease recurrence risk with and without RNI/PMRT.
Materials/Methods: We reviewed and identified patients with isolated N1mic disease (<2 mm nodal involvement) treated with upfront lumpectomy or mastectomy at our institution between 2000 and 2025. Patients receiving neoadjuvant chemotherapy, those with any macroscopic nodal disease (>2 mm), bilateral breast cancers, or node-negative disease were excluded. Primary endpoint were locoregional recurrence (LRR) and invasive disease recurrence (IDR) (defined as incidence of any distant or local recurrence with death as competing risk). Predictor selection was performed using LASSO machine learning regression, followed by Fine–Gray competing-risk modeling and nomogram development. Model performance was internally assessed and externally validated in an independent institutional cohort.
Results: The analysis included 906 N1mic patients with a median follow-up (FU) of 6.9 years. 52% had lumpectomy, 48% mastectomy and only 10% received RNI/PMRT. The 10-year rates of LRR and IDR in the entire cohort were 11% and 20%, respectively. External Validation cohort included 225 patients with median FU of 9.1 years. Of those 225 in the external validation cohort, 50.2% had lumpectomy, 49.8% mastectomy and 44% received RNI/PMRT. The 5-year rates of LRR and IDR in the external validation cohort were 2.8% and 6.9%, respectively. The final nomogram model demonstrated good discrimination/calibration, with an area under the curve (AUC) of 0.72 on external validation. Examples of clinical translation: For a 49-year-old woman with ER/PR-positive, HER2-negative, 3.1 cm invasive ductal carcinoma grade I with lymphovascular invasion (LVI) undergoing mastectomy and sentinel node biopsy (SLNB) with N1mic disease (1/2 nodes), the model estimated a 10-years IDR risk of approximately 32% without PMRT and 17% with PMRT. Conversely, if the same patient with the same tumor biology but with a smaller tumor (1.6 cm), higher grade 3 and no LVI, the model estimated a 10-years IDR risk of approximately 15.2% without PMRT and 7.8% with PMRT.
Conclusion: We developed and externally validated the first nomogram to predict individualized IDR risks for N1mic BC patients treated with and without RNI/PMRT. This model will be translated into an online calculator tool to address a longstanding evidence gap, enabling personalized, risk-adapted radiotherapy decisions in a population where RNI/PMRT has historically relied on extrapolation and physician judgement.