Prediction of occult axillary metastases in treatment-naïve patients with breast cancer: A transSENTINA analysis.

Abstract

566 Background: Prediction of occult axillary metastases through clinical / biological parameters may allow reduction of axillary staging. This is particularly important, as systemic therapies have become more efficient. We have conducted a systematic analysis among patients undergoing axillary sentinel lymph-node biopsy (SLNB) before initiation of primary systemic therapy as part of a clinical trial (SENTINA) with the goal to identify predictors of sentinel lymph node status in a well-defined patient cohort. Methods: Patients with a clinically negative axillary status who underwent SLNB as part of the prospective SENTINA trial were included. Univariate and multivariate analyses were carried out to identify clinical / pathological parameters associated with SLN status, using logistic regression models. Model performance was assessed by ROC analyses. Calculations were performed using R version 3.5.2. Results: Arms A and B of the SENTINA study contained 1022 patients. Among 805 cN0 patients, all parameters considered relevant for this analysis were available. 527 and 278 patients presented with negative and positive lymph nodes upon SLN biopsy, respectively. Univariate regression models identified largest tumor diameter (odds ratio (OR) 1.016, p-value 0.0041), tumor type (ductal vs. lobular, OR 2.004, p 0.00234), tumor grading (low vs. high, OR 0.537, p < 0.001), hormone receptor (HR) status (negative vs. positive, OR 2.668, p < 0.001), HER2 status (negative vs. positive, OR 1.462, p 0.0158) as being associated with SLN status with an a < 0.1. Multivariate analysis resulted in tumor diameter, HR status, HER2 status and tumor type being independently associated. These parameters were combined using stepwise (backward and forward) selection into a prediction model. This model predicted SLN status with an AUC of only 0.65. Conclusions: Using data obtained as part of the SENTINA trial we were able to build a prediction model that was able to predict SLN metastases in treatment naïve cN0 patients with limited accuracy. Additional (biological) parameters, such as response to systemic therapies (i.e. axillary conversion through PST) may be more appropriate to predict presence of occult sentinel lymph node metastases among patients with breast cancer.