190MO Association of 18-Gene expression profile (GEP) with clinical outcomes in patients with metastatic triple-negative breast cancer (mTNBC) treated with pembrolizumab (Pembro) or chemotherapy (Chemo) in KEYNOTE-119

Abstract

In the phase III, randomized, open-label KEYNOTE-119 (NCT02555657) study, second- or third-line treatment with pembro monotherapy did not significantly improve OS vs single-agent chemo treatment of physician’s choice (TPC) in patients (pts) with mTNBC, although the pembro treatment effect increased with increasing PD-L1 enrichment. In this exploratory analysis, we evaluated the association of T-cell-inflamed gene expression profile (GEP) with clinical outcomes from KEYNOTE-119. Pts with centrally confirmed mTNBC, ECOG PS ≤1, and 1-2 prior systemic treatments for metastatic breast cancer were randomized 1:1 to pembro 200 mg Q3W for up to 35 cycles or TPC (capecitabine, eribulin, gemcitabine, or vinorelbine) per local guidelines. Tumor RNA profiling was conducted on the RNA-seq platform (Illumina, San Diego, CA). Association of GEP with clinical outcomes was evaluated by logistic regression (best objective response [BOR] per RECIST v1.1) and Cox proportional hazards models (PFS and OS) with baseline adjustment for ECOG PS; 95% CIs for BOR were estimated using the Clopper and Pearson method. As of April 11, 2019, GEP data for 333 pts (pembro, n=177; TPC, n=156) were available for analysis. GEP was significantly associated with improved BOR, PFS, and OS in pts treated with pembro but not TPC (Table). HR for OS was 0.77 (95% CI, 0.58–1.04) in pts with GEP non-low (≥1st tertile) and 1.72 (1.15–2.55) in pts with GEP low (<1st tertile). No association was seen between BRCA/HRD status and treatment response. These findings suggest a positive association between GEP and clinical outcomes in pts with mTNBC who are treated with pembro, with a more favorable treatment effect in the GEP-enriched population. Results may help inform future analyses of outcomes by biomarker status in this population.