Development of a predictive score of axillary lymph node dissection based on targeted axillary dissection in patients with breast cancer diagnosis, affected lymph nodes, and neoadjuvant treatment

Aim: To determine predictive factors of axillary lymph node dissection (ALND) results in breast cancer (BC) patients undergoing neoadjuvant chemotherapy (NACT), and subsequent staging using Targeted Axillary Dissection (TAD). Material and method: Case-control study between January 2016 and August 2019. Patients with BC, cN1 staging, marked with a metallic clip prior to NACT, and subsequently staged with TAD and ALND were included. They were divided into 2 groups: ALND patients with or without metastatic involvement (group 1 and group 2, respectively). We carried out a univariate analysis comparing clinical, radiological, surgical and pathological variables, and a logistic regression, (dependent variable: positive result of ALND; independent variables: number of suspicious lymph nodes in diagnostic ultrasound, positive hormone receptors, HER2 positive, complete clinical-radiological response to NACT, positive TAD, and biopsy of <= 2 nodes in TAD). A score for prediction of a metastatic ALND was proposed, with an internal validation study. Results: 60 patients were included: Group 1: 33 (55.0%); Group 2: 27 (45.0%). Tumor size (Odds Ratio (OR) = 1.67; 95%CI 1.02-2.74), number of suspected nodes in ultrasound (OR = 2.20; 95%CI 1.01-4, 77), HER2 positive (OR 0.04; 95%CI 0.003-0.54), clinical-radiological response to NACT (OR = 0.07; 95%CI 0.01-0.75), and positive TAD (OR 15.48; 95%CI 1.68-142.78) were independent predictors of a positive result in ALND. We developed a "positive ALND predictive score", with good calibration (Hosmer-Lemeshow test: p = 0.65), and discrimination (AUC = 0.93; 95% CI 0, 87-0.99), with highest Youden index (0.7) at cut-off point of 17% risk of positive ALND (sensitivity = 100%; specificity = 70%). Conclusion: Tumor size, number of suspected nodes, positive HER2, response to NACT, and metastatic TAD are independent predictors of ALND. The predictive score for positive ALND would be a good indicator to safely omit ALND.