Abstract P5-12-01: A novel gene expression signature prognostic for both locoregional and distant failure and predictive for adjuvant radiotherapy

Abstract Background: Most patients with early stage breast cancer (BC) are treated with adjuvant radiotherapy (RT) following breast conserving surgery (BCS) to prevent locoregional recurrences (LRR). No predictive tools are currently available to select patients for RT, resulting in considerable over- and under treatment. We aimed to create and validate a gene expression-based classifier to prognosticate for LRR and to stratify patients for treatment with RT. Patients and methods: A 27-gene expression signature was developed using three publicly available early stage BC gene expression datasets where patients were treated with RT and had detailed local recurrence information. The largest of the datasets was used to train the signature, and the other two datasets were used for signature refinement. As age was the strongest clinical factor for the endpoint in the training dataset, it was included in the model, resulting in a final clinical-genomic classifier of 27 genes and age. The classifier was locked before external validation in the SweBCG91-RT trial. This phase III clinical trial included primary tumors from 765 patients and for which gene expression data was available. The trial randomized node-negative BC patients to +/- RT following BCS, with sparse use of adjuvant systemic treatment (9%) and a median follow-up of 14.0 years for LRR in patients free from event. The classifier was validated using Cox regression with LRR as the primary endpoint, and hazard ratios (HRs) were calculated using the raw continuous classifier score (range: 0.5 to 2.5). Results: The novel classifier was highly prognostic for LRR in SweBCG91-RT patients treated with RT (HR=7.5[3.3-16.9], p<0.001), and remained prognostic in multivariate analysis (MVA) that included systemic treatment, subtype and grade (HR=7.2[3.1-16.4], p<0.001). To a lesser extent, the classifier was also prognostic for LRR in patients not treated with RT (HR=1.9[1.0-3.5], p=0.03; MVA HR=1.9[1.0-3.3], p=0.05). Patients at high risk of LRR had a smaller effect of RT, and the treatment predictive potential was confirmed by testing for interaction (pinteraction=0.008). In patients treated with RT, age and the genomic component of the model were both prognostic for LRR (p<0.01) as well as predictive for RT response (pinteraction<0.05) and provided independent information (p<0.01). The combined classifier has increased performance over its individual components (10-year AUC=0.72, 0.67, 0.65 for the classifier, age, and genomic component, respectively). While the novel signature was prognostic for metastasis (HR=4.3[2.3-7.8], p<0.0001), calculated scores from previously published signatures to the metastasis endpoint, including the Oncotype-like score, were not prognostic for LRR. Conclusions: This novel gene expression signature is highly prognostic for LRR, can identify patients at risk of LRR despite RT, and appears to be treatment predictive for adjuvant RT. Furthermore, the current signature is highly prognostic for metastasis. In contrast, calculated scores of previously published signatures modeled for the metastasis endpoint had inferior performance for LRR. These results underscore both the importance of signatures prognostic for LRR and the similarities in the biology of LRR and distant failure. Citation Format: Sjöström M, Chang SL, Fishbane N, Davicioni E, Zhao SG, Hartman L, Holmberg E, Feng FY, Speers CW, Pierce LJ, Malmström P, Fernö M, Karlsson P. A novel gene expression signature prognostic for both locoregional and distant failure and predictive for adjuvant radiotherapy [abstract]. In: Proceedings of the 2018 San Antonio Breast Cancer Symposium; 2018 Dec 4-8; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2019;79(4 Suppl):Abstract nr P5-12-01.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.