Predictive Roles of Baseline Stromal Tumor-Infiltrating Lymphocytes and Ki-67 in Pathologic Complete Response in an Early-Stage Triple-Negative Breast Cancer Prospective Trial

SIMPLE SUMMARY: High stromal tumor-infiltrating lymphocytes (sTILs) are associated with improved pathologic complete response (pCR) in triple-negative breast cancer (TNBC). In this study of 408 patients enrolled in a prospective early-stage TNBC neoadjuvant chemotherapy trial, we aimed to identify clinicopathologic features that could be combined with sTILs to better predict pCR. Applying a training set and a testing set, we found that integrating high Ki-67 (cutoff > 35%) and high sTIL (cutoff ≥ 20%) in a model of computed response scores could predict a pCR rate of 65%. This model may refine the selection of early-stage TNBC patients for neoadjuvant clinical trials evaluating de-escalation strategies. ABSTRACT: High stromal tumor-infiltrating lymphocytes (sTILs) are associated with improved pathologic complete response (pCR) in triple-negative breast cancer (TNBC). We hypothesize that integrating high sTILs and additional clinicopathologic features associated with pCR could enhance our ability to predict the group of patients on whom treatment de-escalation strategies could be tested. In this prospective early-stage TNBC neoadjuvant chemotherapy study, pretreatment biopsies from 408 patients were evaluated for their clinical and demographic features, as well as biomarkers including sTILs, Ki-67, PD-L1 and androgen receptor. Multivariate logistic regression models were developed to generate a computed response score to predict pCR. The pCR rate for the entire cohort was 41%. Recursive partitioning analysis identified ≥20% as the optimal cutoff for sTILs to denote 35% (143/408) of patients as having high sTILs, with a pCR rate of 59%, and 65% (265/408) of patients as having low sTILs, with a pCR rate of 31%. High Ki-67 (cutoff > 35%) was identified as the only predictor of pCR in addition to sTILs in the training set. This finding was verified in the testing set, where the highest computed response score encompassing both high sTILa and high Ki-67 predicted a pCR rate of 65%. Integrating Ki67 and sTIL may refine the selection of early stage TNBC patients for neoadjuvant clinical trials evaluating de-escalation strategies.