Abstract P4-09-08: A targeted breast cancer radiosensitivity gene expression panel

Background: A majority of patients with early breast cancer is operated with breast conserving surgery (BCS) and adjuvant radiotherapy (RT) is administered to prevent ipsilateral breast tumor recurrence (IBTR), including a new ipsilateral cancer. The EBCTCG meta-analysis showed a majority of patients treated with surgery only to be recurrence free at 10 years, and more than 10% to suffer an IBTR despite RT, thus implying considerable over- and under treatment. A wide range of prognosticators, including multigene tests, are well established, but we lack predictive factors for RT, which is the aim in the present study. Patients and methods: Fresh frozen tissue from 340 patients operated with BCS with or without RT and with or without IBTR was collected (without IBTR N=196, with IBTR n=144). Patients were stratified according to estrogen receptor (ER) status and RT, and divided into a training cohort (N=172) and a validation cohort (N=168). The training cohort was analyzed with whole transcriptome analysis (Illumina HT12 v4) and top discriminating genes for IBTR (N=155) were selected based on a random forest machine learning algorithm with recursive feature elimination and cross-validation. Further, genes described in the literature as associated with radioresistance were included in the panel to a total of 248 genes. A custom nCounter (Nanostring Technologies) gene expression panel was designed and both the training and validation cohorts were analyzed with the custom panel. Single-sample classifiers using a k-top scoring pairs algorithm were trained in the training cohort and validated in the validation cohort. Area under the curve (AUC) with a receiver operator characteristics (ROC) analysis were calculated and p-values were calculated with a log-rank test. All calculations were done using the R statistical environment. Results: Our classifiers were prognostic for IBTR in the validation cohort among ER+ patients given RT (AUC 0.67, p=0.005), ER+ patients not given RT (AUC=0.89, p=0.015) and ER- patients given RT (AUC=0.78, p Conclusions: Our targeted radiosensitivity gene expression panel could identify patients of high or low risk of LR, with or without RT. The most promising was however that it seems as the panel could be used as a predictive marker, i.e., finding patients that do, or do not, respond to RT. Further refinement and testing of the panel and models is ongoing. Citation Format: Sjostrom M, Staaf J, Eden P, Warnberg F, Bergh J, Malmstrom P, Ferno M, Nimeus E, Fredriksson I. A targeted breast cancer radiosensitivity gene expression panel [abstract]. In: Proceedings of the 2017 San Antonio Breast Cancer Symposium; 2017 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2018;78(4 Suppl):Abstract nr P4-09-08.