Prediction of axillary lymph node metastasis in breast cancer using an ultrasonic feature- and clinical data-based model.

Abstract

The involvement of axillary lymph nodes (ALNs) is a critical prognostic factor affecting patient management and outcomes in breast cancer (BC). This study aims to comprehensively analyze the clinical data of BC patients, evaluate ultrasonic signs of ALNs, and explore the implications of a prediction model for ALN metastasis (ALNM) in early-stage BC patients based on ultrasonic features and clinical data. This study retrospectively analyzed ultrasonic features and clinical data from 216 patients diagnosed with unilateral invasive BC. The dataset was divided into a training (n = 162) and a validation set (n = 54) in a 3:1 ratio. Patients were then assigned into metastasis and non-metastasis groups depending on ALNM determined by pathological findings. Univariate analysis of various indicators followed by multivariate Logistic regression analysis was performed on the training set. A prediction model for ALNM in BC was established using binary logistic regression analysis, with its prediction performance evaluated by receiver operating characteristic curves (ROC) and area under the curve (AUC), and its reproducibility verified by the validation set. The pathological findings identified 57 (35.2%) cases of ALNM among 162 BC patients in the training set. Risk factors for ALNM included poorly differentiated type, high Ki-67 expression, lymph node (LN) aspect ratio ≥2, LN cortical thickness ≥1/2 of lymphatic hilum diameter, and mixed or peripheral LN blood flow. Protective factors included mass location in the outer upper quadrant and LN size >1 cm. A prediction model was established based on risk factors, with the equation being Logit (P) = -4.881 - 1.285 * differentiation degree + 1.485 * Ki-67 - 1.090 * lump quadrant - 0.956 * lymph node size + 1.244 * lymph aspect ratio + 1.032 * LN cortical thickness + 1.454 * LN medullary disappearance + 1.266 * LN blood flow. ROC analysis of the model yielded an AUC of 0.866, with a sensitivity of 80.7% and a specificity of 80.0%. The prediction model was validated using the validation set, producing an AUC of 0.809. These results demonstrate that color Doppler ultrasound effectively evaluates ALN status in BC patients. The prediction model for ALNM in BC shows strong accuracy and has potential clinical application.