Background: AR is expressed in BC and may be a discrete subtype of triple negative (TN) BC. Recent trials showed modest efficacy of AR antagonists in AR+ TNBC. We investigated associations between primary BC AR gene expression (ARe), clinical characteristics, and outcomes in publically available databases. Methods: ARe was evaluated with microarray data from the neoadjuvant I-SPY 1 study (n = 149). Associations with clinical features and chemotherapy response were assessed with Kruskal-Wallis and Wilcoxon rank sum tests and recurrence free survival (RFS) by the Cox proportional hazard model. Pearson correlations between AR and selected genes were determined in I-SPY 1, METABRIC (n = 1992), and TCGA (n = 817). Results: In I-SPY1, ARe was lower in TN than hormone receptor+/HER2- and HER2+ BC (p < 0.0001). ARe was lower in basal BC (p < 0.0001) and was higher in patients > age 50 (p = 0.05). ARe correlated with grade I/II histology (p < 0.0001) and node negativity at diagnosis (p = 0.0058) but not with stage, lymphovascular invasion, or pathologic complete response. Higher ARe remained associated with better RFS upon adjustment for receptor subtype (p = 0.01). Table 1 below shows genes with significant (p < 0.05) correlations with AR in ≥ 2 datasets. Conclusions: ARe is lowest in TN and basal BC. Consistent with this finding, AR correlates with luminal genes FOXA1, ESR1, MUC1, and inversely with basal genes MYC, CRYAB, EGFR, DNA damage repair genes, mesenchymal genes KIT and CDK 6, and immune genes PD-1 and STAT5A. ARe appears to provide independent prognostic information when receptor subtypes were considered.

NS: non-significant