Breast cancer metastasis gene signature: The search for the holy grail.

Abstract

1075 Background: Lymph node status remains one of the most useful prognostic indicators in breast cancer, however, current methods to assess nodal status disrupt the lymphatic system, lead to secondary complications, and 10-33% of women diagnosed with negative lymph nodes will develop metastases. Identification of molecular signatures discriminating primary tumors from patients with the propensity to metastasize from those without would allow for stratification of patients requiring surgical assessment of lymph node status and identification of node negative patients at high-risk of developing metastatic disease. Methods: Frozen breast specimens were collected from women with negative (n=39) and positive (n=35) lymph node status. Patients with positive BRCA1/BRCA2 status, neoadjuvant therapy or negative lymph node status who later developed distant metastasis were not included in this study. RNA was isolated from pure tumor cell populations after laser microdissection and gene expression data generated using HG U133A 2.0 arrays (Affymetrix). Data was randomly split into a training set (29 negative and 25 positive) and a validation set (10 negative and 10 positive) and a classifier was built using a forward-building Bayesian naïve classifier. Results: Neither age at diagnosis nor pathological characteristics including ER, PR, or HER2 status differed significantly between the two groups. A median prediction probability >95% was achieved with the Bayesian classifier using 17 markers (mean probability 94.31%, median 95.05%, min 74.62%, and max 99.91%) in the training set. The 17 marker panel was, however, only able to correctly classify 9/20 samples in validation set, resulting in approximately random classification. Conclusions: The inability to derive molecular profiles of metastasis may reflect tumor heterogeneity (e.g., different methods of metastasis for each tumor subtype), the presence of only a few cells within the primary tumor that have metastatic potential, or an inherited host susceptibility to metastasis, rather than molecular properties of the tumor itself. Together, these data suggest that a molecular signature that can discriminate between primary breast tumors with and without lymph node metastases does not exist. No significant financial relationships to disclose.