A Validated Ultrasound-Based Scoring System to Stratify Risk of Axillary Metastasis in Breast Cancer: AX-RADS (Axillary Imaging Reporting and Data System)

Introduction: Ultrasound is the imaging modality of choice for evaluation of axillary involvement in breast cancer, but is associated with variable sensitivity and specificity. Understanding the risk of axillary lymph node metastasis (ALNM) based on ultrasonographic and clinical features will inform treatment decisions. Our group aimed to create a scoring system to quantify the risk of ALNM based on ultrasound characteristics in breast cancer patients. We validated the model and tested it among different Memorial Sloan Kettering Breast Cancer Sentinel Lymph Node Metastasis Nomogram (MSK) subgroups. Methods: The ultrasound score was developed using data collected at a single institution from 2019 to 2021 by allocating points based on the regression coefficients of variables found to significantly predict ALNM. We validated the test statistics of our score at an outside institution. The index and validation cohorts were combined: 358 pooled patients were stratified by predicted ALNM positivity according to a validated nomogram based on primary tumor characteristics. Results: Between 2019 and 2021, in the validation cohort, the NPV for low risk (0-1) scores was 87%, while the PPV for high-risk (5 +) scores was 71%. Overall, in the combined cohort, 241 (67%) patients had low-risk (0-1) axillary ultrasound scores and 33 (9%) had high risk (5 +) scores. In this combined cohort, NPV was 84% (203/241 low-risk score patients were node negative), while PPV for high-risk scores was 85% (28/33 high-risk score patients were node positive). When stratified via the Memorial Sloan Kettering Breast Cancer Nomogram: Sentinel Lymph Node Metastasis predicted ALNM rates, the NPV of low-risk scores was 87%-89% for patients with < 50% predicted ALNM positivity. For patients with > 50% predicted ALNM positivity, the PPV of high-risk scores was 82%. Conclusions: A scoring system to predict ALNM among biopsy-proven breast cancer patients undergoing upfront surgery was successfully developed from a multivariate model based on axillary ultrasound characteristics. Combining the axillary US scoring system with an additional validated nomogram based on primary tumor and patient characteristics may help foster better communication about ALNM risk to inform treatment decisions.