Abstract B09: Multivariable Models for Predicting Likely Metastatic Sites for Triple Negative Breast Cancers

Background: Unique organ microenvironment may preferentially support growth of specific tumor clones because of which different breast cancer subtypes show distinct tropisms for sites of metastasis. While a few gene expressionbased signatures are known to predict sitespecific metastasis of breast cancer, little work has focused on identification of clinically facile immunohistochemical predictors of metastasis to specific sites, especially for triple negative breast cancers (TNBCs).

Methods: Primary tumor samples from 322 TNBC patients were stained for 133 biomarkers and assessed by immunohistochemistry. Differences in average levels of these biomarkers were compared between patients with or without metastasis to specific sites (brain, bone, lungs, liver, lymph nodes). Significantly different biomarkers were then analyzed within a Cox regression model to evaluate their prognostic value when patients with metastasis to the site of interest were compared to patients with no metastasis. Ideal thresholds, based on maximizing model fit, stratified cohorts that show high and low expression of each biomarker. A combination of a biomarker found high for each site, low for each site, and the Nottingham Prognostic Index (NPI) was used to stratify patients.

Results: Our analysis uncovered several biomarkers whose expression levels in primary tumors can predict the site of future metastasis in TNBCs. Our models for brain (PARP1 & BRCA2), bone (MTA1 &Tumor-Adjacent CD8), liver (TFF1 & N-Cadherin), and lung (ROR) were able to identify patients who had at least a 500% risk of site specific distant metastasis.

Conclusion: Relatively simple and inexpensive immunohistochemical analyses of biomarkers, combined with logical model building, in primary tumors of TNBCs may allow prediction of the site of future metastasis. Highrisk patients may benefit from increased surveillance of such sites.

Citation Format: Sergey Klimov, Andrew Green, Mohammed Aleskandarany, Emad Rakha, Ian Ellis, Michelle Reid, Rida C. G. Padmashree, Ritu Aneja. Multivariable Models for Predicting Likely Metastatic Sites for Triple Negative Breast Cancers. [abstract]. In: Proceedings of the Ninth AACR Conference on the Science of Cancer Health Disparities in Racial/Ethnic Minorities and the Medically Underserved; 2016 Sep 25-28; Fort Lauderdale, FL. Philadelphia (PA): AACR; Cancer Epidemiol Biomarkers Prev 2017;26(2 Suppl):Abstract nr B09.