Purpose: Purpose of this study was to determine the accuracy of prediction of non-sentinel lymph node (NSLN) involvement in sentinel node (SLN)-positive breast cancer patients based on protein concentrations measured in lysates from initially taken breast biopsies.
Methods: Data on protein expression, previously generated by multiplexed bead-based immunoassays, were analysed by multivariate logistic regression to define parameter sets of value to predict NSLN involvement. Receiver-operator characteristics (ROCs) were calculated as indicators of diagnostic significance.
Results: Analyses of data from all patients (n = 99) resulted in parameter sets that allowed direct prediction of the NSLN status with a ROC area under the curve (AUC) of 0.83. The clinically most relevant prediction of NSLN status in SLN-positive patients (n = 37) based on only seven parameters (including TIMP-2 as the most relevant single value) was possible with high accuracy indicated by an AUC of 0.89.
Conclusions: Parallel assessment of protein concentrations in breast biopsies is a highly promising approach to predict nodal involvement and even the NSLN status in SLN-negative breast cancer patients. Such diagnostic information could substantially reduce the number of completion axillary lymph node dissections in clinical practice.