Cell-free DNA copy number variations as a marker for breast cancer in a large study cohort.

Tumor Biology
Session Type and Session Title: 
Poster Discussion Session, Tumor Biology
Abstract Number: 
J Clin Oncol 31, 2013 (suppl; abstr 11013)
Julia Beck, Ekkehard Schütz, Howard B. Urnovitz, Adel Tabchy, William M. Mitchell, Gordon B. Mills, Funda Meric-Bernstam; Chronix Biomedical, Göttingen, Germany; The University of Texas MD Anderson Cancer Center, Houston, TX; Department of Pathology, Vanderbilt University, Nashville, TN

Abstracts that were granted an exception in accordance with ASCO's Conflict of Interest Policy are designated with a caret symbol (^).

Abstract Disclosures


Background: Massive parallel sequencing provides high numbers of cell-free nucleic acid serum DNA sequences (cfDNA) that can detect trace amounts of tumor derived chromosomal imbalances and copy number variations (CNVs) in patients with cancer. The aim of this study was to determine if there is a difference between the cfDNA CNVs from patients with breast cancer (BrCa) compared to healthy controls. Methods: DNA extracted from serum samples of 225 BrCa (Stage 1 to 4) and 205 gender and age-matched healthy controls (HC) was amplified using random primers, tagged with a unique molecular identifier per sample, sequenced on an Illumina HiSeq system and aligned to the human genome (Build 37). Hits were counted in sliding 1Mbp interval regions and normalized. Using a Random-Resampling procedure, a model was established to distinguish BrCa from HC using the copy number variations (CNV) and cross validated. Results: From 1,100 rounds of random resampling (50/50), a set of 31 regions was selected, based on the frequency of occurrence in the models. Using 20 random sets of a 10-fold cross validation, the selected regions were found to be highly significant discriminators between BrCa and HC (p<10-5). When using a final linear model with 16 regions the AUC of a diagnostic ROC curve was found to be 0.895 for all samples, for Stage I and II the AUC was 0.86 compared to 0.93 for the higher stages. The final model included three regions from chromosome 8 and 1 and two regions from chromosome 15, the remaining regions were found as one per chromosome. Conclusions: Using comparative massive parallel sequencing of cfDNA from cancer patients vs. controls, we were able to show that a 16-region model based on CNV, is useful to distinguish patients with breast cancer from matched controls. Genomic instabilities that are shed into the circulation from breast cancer may play a role in screening, monitoring or as companion diagnostic tests in breast cancer.