Molecular Drivers of Oncotype DX, Prosigna, EndoPredict, and the Breast Cancer Index: A TransATAC Study.

Abstract

PURPOSEThe Oncotype DX Recurrence Score (RS), Prosigna Prediction Analysis of Microarray 50 (PAM50) Risk of Recurrence (ROR), EndoPredict (EP), and Breast Cancer Index (BCI) are used clinically for estimating risk of distant recurrence for patients receiving endocrine therapy. Discordances in estimates occur between them. We aimed to identify the molecular features that drive the tests and lead to these differences.PATIENTS AND METHODSAnalyses for RS, ROR, EP, and BCI were conducted by the manufacturers in the TransATAC sample collection that consisted of the tamoxifen or anastrozole arms of the ATAC trial. Estrogen receptor–positive/human epidermal growth factor receptor 2 (HER2)–negative cases without chemotherapy treatment were included in which all four tests were available (n = 785). Clinicopathologic features included in some tests were excluded from the comparisons. Estrogen, proliferation, invasion, and HER2 module scores from RS were used to characterize the respective molecular features. Spearman correlation and analysis of variance tests were applied.RESULTSThere were moderate to strong correlations among the four molecular scores (ρ = 0.63-0.74) except for RS versus ROR (ρ = 0.32) and RS versus BCI (ρ = 0.35). RS had strong negative correlation with its estrogen module (ρ = −0.79) and moderate positive correlation with its proliferation module (ρ = 0.36). RS’s proliferation module explained 72.5% of ROR’s variance, while the estrogen module explained only 0.6%. Most of EP’s and BCI’s variation was accounted for by the proliferation module (50.0% and 54.3%, respectively) and much less by the estrogen module (20.2% and 2.7%, respectively).CONCLUSIONIn contrast to common understanding, RSs are determined more strongly by estrogen-related features and only weakly by proliferation markers. However, the EP, BCI, and particularly ROR scores are determined largely by proliferative features. These relationships help to explain the differences in the prognostic performance of the tests.