Abstract P1-10-02: A signature predictive of early vs. late recurrence after radiation treatment (RT) for breast cancer that may inform the biology of early, aggressive recurrences

Abstract

Abstract Purpose: Unmet clinical needs in breast cancer (BC) management include the identification of patients (pts) at high risk to fail locally despite standard local therapy including RT and understanding the biology of these recurrences. We previously reported1 a RT response signature and here extend those studies to identify a signature predictive of timing of recurrence after completion of RT (before or after 3 years). Methods: Two independent patient cohorts (treated with BCS) from non-randomized clinical trials were used for training and validation. The training cohort included 119 pts with in-breast tumor recurrence and the validation cohort had 25 pts with recurrences. Initial feature selection used Spearman's rank correlation correlating gene expression (14,806 genes) to recurrence time. Genes with sig. correlation (FDR <0.1) and large expression range (fold change >2) were used to train an elastic net penalized Poisson regression model. This model was locked and then applied to the validation dataset. Cox regression was used for both univariate and multivariable analyses (UVA and MVA). To identify biological-related concepts, Spearman's corr. coefficients of recurrence time to gene expression within the training cohort were used to generate a pre-ranked list upon which GSEA pathway analysis was performed. Results: Spearman's correlation identified 485 genes whose expression was significantly associated with recurrence time (early vs. late). Feature reduction further refined the gene list to 41 genes, which were retained within the signature and locked for further validation. In the training dataset the Spearman's correlation of the continuous score to recurrence time was 0.852 with a P-value of 1.3x10-34 and an AUC of 0.92. Application of this early vs late signature to an independent BC validation set accurately identifies pts with early vs. late recurrences (Spearman's corr.=0.537, p-value<0.007, AUC=0.74, sensitivity=0.71, specificity=0.73, PPV=0.77, NPV=0.67). In UVA and MVA the early vs. late recurrence signature remained the most significant factor associated with recurrence time. Although independent of intrinsic subtype, GSEA analysis of the 41 genes retained within the signature identifies proliferation and EGFR concepts associated with early recurrences and luminal and ER-signaling pathways associated with late recurrences. Knockdown of genes associated with the early and late recurrences is currently underway to assess phenotypic changes (proliferation and clonogenic survival as a measure of early and durable RT response) associated with the early and late recurrence-associated genes. Conclusion: In this study we derive a BC-specific RT signature predictive of early vs. late recurrence with biologic relevance and validate this signature for prediction of timing of recurrence in an independent clinical dataset. By identifying pts with tumors likely to recur sooner vs. later this signature has the potential to allow for a furthered understanding of the biology underlying early and late recurrences and has a potential to personalize RT, particularly in patients for whom treatment intensification is needed. 1. Clin Cancer Res. 2015 Aug 15;21(16):3667-77. Citation Format: Speers C, Chang L, Santola A, Liu M, Zhao SG, Chandler B, Olsen E, Bartelink H, Feng FY, Pierce LJ. A signature predictive of early vs. late recurrence after radiation treatment (RT) for breast cancer that may inform the biology of early, aggressive recurrences [abstract]. In: Proceedings of the 2016 San Antonio Breast Cancer Symposium; 2016 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2017;77(4 Suppl):Abstract nr P1-10-02.