Prediction of response to neoadjuvant dose-dense chemotherapy in luminal breast cancer using a genomic risk probability model.

Abstract

e12627 Background: Response to neoadjuvant chemotherapy (NACT) in hormone receptor-positive, HER2-negative (ER+/HER2-) breast cancer (BC) is heterogeneous. Genomic assays predict chemotherapy benefit but are not widely available. We evaluated a clinicopathologic model estimating the probability of high genomic risk as a predictor of response to dose-dense NACT (AC q2w followed by Docetaxel q3w). Methods: We retrospectively analyzed 280 patients with stage II-III ER+/HER2- BC treated with NACT (median age 45 years; 70,3% premenopausal) in a single center. Most patients had locally advanced disease (86,8%), nodal involvement (93,9%), and luminal B subtype (95,4%). Pathologic response was assessed using Residual Cancer Burden (RCB). The probability of high genomic risk (corresponding to Recurrence Score ≥26) was calculated using the model developed by S. Yamamoto et al. according to the following formula: p = 1 / {1 + exp [−(4.611 + 1.2342×HER2 − 0.0813×ER − 0.0489×PR + 0.0857×Ki-67)]}, where HER2=0-2; ER, PR, Ki-67=% positive nuclei. A cutoff of ≥50% was predefined as in the original study since the accuracy, sensitivity, specificity, positive predictive and negative predictive values are 90,5%, 72,2%, 94,8%, 76,4% and 93,5%, respectively. Univariable and multivariable logistic regression analyses were performed to identify predictors of favorable response (RCB 0-1). RCB 0-1 as an end point was chosen due comparable long-term results in ER+/HER2- BC as per meta-analysis. Results: RCB 0 was achieved in 12,9% of patients and RCB 0-1 in 24,3%. Mean probability of high genomic risk was 40,3%, with 36,8% classified as high probability (≥50%). Germline testing was performed in 83,2% of patients; mutations (BRCA1/2, ATM, PALB2, CHEK2) were identified by NGS in 18,0% of all tested cases (BRCA2 being the most common - 9%) and were slightly more frequent in patients with high vs low genomic risk probability (22,2% vs 15,2%). In univariable analysis, factors associated with RCB 0-1 included age ≤50 years (OR 2,09, 95% CI 1,07-4,07), initially operable disease (OR 2,44, 95% CI 1,19-5,04), T1–3 stage (OR 1,80, 95% CI 1,03-3,13), presence of germline mutations (OR 2,82, 95% CI 1,41-5,67), and high genomic risk probability ≥50% (OR 2,64, 95% CI 1,51-4,62; all p <0,05). In multivariable analysis, high genomic risk probability ≥50% remained an independent predictor of RCB 0-1 (OR 2,80, 95% CI 1,48-5,30; p =0,001), together with germline mutation status (OR 2,32, 95% CI 1,11-4,85; p =0,025), while age, T stage and operability were not significant. Survival analysis is ongoing. Conclusions: A clinicopathologic model estimating the probability of high genomic risk independently predicts response to NACT in luminal breast cancer and may support selection of ER+/HER2- patients most likely to benefit from NACT when genomic testing is unavailable.