Response and survival of breast cancer intrinsic subtypes following multi-agent neoadjuvant chemotherapy

Aleix Prat, Cheng Fan, Aranzazu Fernández, Katherine A. Hoadley, Rossella Martinello, Maria Vidal, Margarita Viladot, Estela Pineda, Ana Arance, Montserrat Muñoz, Laia Paré, Maggie C. U. Cheang, Barbara Adamo, Charles M. Perou

Predicting treatment benefit and/or outcome before any therapeutic intervention has taken place would be clinically very useful. Herein, we evaluate the ability of the intrinsic subtypes and the risk of relapse score at diagnosis to predict survival and response following neoadjuvant chemotherapy. In addition, we evaluated the ability of the Claudin-low and 7-TNBCtype classifications to predict response within triple-negative breast cancer (TNBC). Gene expression and clinical-pathological data were evaluated in a combined dataset of 957 breast cancer patients, including 350 with TNBC, treated with sequential anthracycline and anti-microtubule-based neoadjuvant regimens. Intrinsic subtype, risk of relapse score based on subtype and proliferation (ROR-P), the Claudin-low subtype and the 7-TNBCtype subtype classification were evaluated. Logistic regression models for pathological complete response (pCR) and Cox models for distant relapse-free survival (DRFS) were used. Basal-like, Luminal A, Luminal B, and HER2-enriched subtypes represented 32.7 %, 30.6 %, 18.2 %, and 10.3 % of cases, respectively. Intrinsic subtype was independently associated with pCR in all patients, in hormone receptor-positive/HER2-negative disease, in HER2-positive disease, and in TNBC. The pCR rate of Basal-like disease was >35 % across all clinical cohorts. Neither the Claudin-low nor the 7-TNBCtype subtype classifications predicted pCR within TNBCs after accounting for intrinsic subtype. Finally, intrinsic subtype and ROR-P provided independent prognostic information beyond clinicopathological variables and type of pathological response. A 5-year DRFS of 97.5 % (92.8-100.0 %) was observed in these neoadjuvant-treated and clinically node-negative patients predicted to be low risk by ROR-P (i.e. 57.4 % of Luminal A tumors with clinically node-negative disease). Intrinsic subtyping at diagnosis provides prognostic and predictive information for patients receiving neoadjuvant chemotherapy. Although we could not exclude a survival benefit of neoadjuvant chemotherapy in patients with early breast cancer with clinically node-negative and ROR-low disease at diagnosis, the absolute benefit of cytotoxic therapy in this group might be rather small (if any).