Purpose: Sclerosing adenosis (SA), found in ¼ of benign breast disease (BBD) biopsies, is a histological feature characterized by lobulocentric proliferation of acini and stromal fibrosis and confers a two-fold increase in breast cancer risk compared to women in the general population. We evaluated a NanoString-based gene expression assay to model breast cancer risk using RNA derived from formalin-fixed, paraffin-embedded (FFPE) biopsies with SA.
Methods: The study group consisted of 151 women diagnosed with SA between 1967 and 2001 within the Mayo BBD cohort, of which 37 subsequently developed cancer within 10 years (cases) and 114 did not (controls). RNA was isolated from benign breast biopsies, and NanoString-based methods were used to assess expression levels of 61 genes, including 35 identified by previous array-based profiling experiments and 26 from biological insight. Diagonal linear discriminant analysis of these data was used to predict cancer within 10 years. Predictive performance was assessed with receiver operating characteristic area under the curve (ROC-AUC) values estimated from 5-fold cross-validation.
Results: Gene expression prediction models achieved cross-validated ROC-AUC estimates ranging from 0.66 to 0.70. Performing univariate associations within each of the five folds consistently identified genes DLK2, EXOC6, KIT, RGS12, and SORBS2 as significant; a model with only these five genes showed cross-validated ROC-AUC of 0.75, which compared favorably to risk prediction using established clinical models (Gail/BCRAT: 0.57; BBD-BC: 0.67).
Conclusions: Our results demonstrate that biomarkers of breast cancer risk can be detected in benign breast tissue years prior to cancer development in women with SA. These markers can be assessed using assay methods optimized for RNA derived from FFPE biopsy tissues which are commonly available.
Keywords: Benign breast disease; Breast cancer; Formalin-fixed paraffin-embedded; NanoString; Risk prediction; Sclerosing adenosis.