Gene Modules and Response to Neoadjuvant Chemotherapy in Breast Cancer Subtypes: A Pooled Analysis

Abstract

To investigate the association between chemotherapy response and gene expression modules describing important biologic processes and druggable oncogenic pathways in breast cancer (BC) subtypes. We searched for publicly available gene expression studies evaluating anthracycline with or without taxane-based neoadjuvant chemotherapy and identified eight studies with 996 patients. We computed 17 gene modules and calculated odds ratios (ORs) for pathologic complete response (pCR) for one-unit increases in scaled modules with and without adjustment for clinicopathologic characteristics. Added predictive accuracy was evaluated using the area under the receiver operating characteristic curve (AUC) and integrated discrimination index (IDI). We used the false discovery rate (FDR) to adjust for multiple testing. High immune module scores were associated with increased pCR probability in all BC subtypes. High module scores of chromosomal instability, phosphatase and tensin homolog (PTEN) loss, and E2F3 transcription factor were associated with increased pCR probability in estrogen receptor (ER) -negative/human epidermal growth factor receptor 2 (HER2) -negative and ER-positive/HER2-negative but not in HER2-positive tumors (interactions between HER2 and each of these modules for their association with pCR: P < .05; FDR, 0.17; trend for interaction between HER2 and PTEN). High values of insulin-like growth factor 1 activation module were associated with increased pCR probability only in ER-positive/HER2-negative tumors (interaction between insulin-like growth factor 1 and ER: P = .002; FDR, 0.03). When adding the immune module to clinicopathologic characteristics, we observed substantial increases in predictive accuracy for pCR in the HER2-positive subtype (IDI, 0.093; P = .004; increase in AUC from 0.760 to 0.836). Different processes and pathways are associated with pCR in different BC subtypes.