Limited Clinical Utility of Prognostic Gene Expression Profiles in Grade 3 Node-Negative Early Stage Breast Cancer.

Abstract

Abstract Background: There are a variety of commercially available prognostic gene expression profiles (GEPs) for early stage breast cancer. These have been shown to be highly concordant in their risk classification, as their prognostic performance is largely driven by an improved quantification of cellular proliferation in the ER+/HER2- (luminal) subtype. We hypothesized that the clinical utility of prognostic GEPs would be limited to the subset of tumors with an indeterminate assessment of proliferation by histological grading (grade 2).Methods: We computed five gene-expression signatures – genomic grade index (GGI), 70-gene signature (GENE70), 76-gene signature (GENE76), 21-gene recurrence score, and wound response signature – in 999 lymph node negative, untreated breast tumors (median follow-up of 7.8 years) from six publicly available datasets. We compared their performance against traditional clinicopathological risk classification using Adjuvant!Online (AOL) for each histological grading category.Results: All five GEPs demonstrated superior prognostication of 10-year distant metastasis free survival (DMFS) in grade 1 (range of area under the ROC curves (AUC)=0.63-0.75) and grade 2 (AUC=0.64-0.71) tumors than in grade 3 (AUC=0.48-0.53) tumors. For the GGI, GENE70, and GENE76 profiles, discordant risk classification with AOL was more frequent in grade 2 (53%) than grade 1 (38%) and grade 3 (26%) tumors (p<.001). In the event of discordant risk classification, GEP identified a low risk subgroup that might be spared chemotherapy in grade 2 (10-yr DFMS: 80% vs 66% (GGI); 81% vs 65% (GENE70); 80% vs 67% (GENE76)) and grade 1 (10-yr DFMS: 92% vs 69% (GGI); 84% vs 79% (GENE70); and 87% vs 70% (GENE76)) but not grade 3 (10-yr DFMS: 67% vs 47% (GGI); 75% vs 47% (GENE70); and 71% vs 52% (GENE76)). Even within the subset of all small (T<2cm), ER-positive, node-negative grade 3 tumors, low genomic risk patients did not experience a sufficiently favorable long-term outcome (10-yr DMFS: 60% (GGI), 69% (GENE70) and 64% (GENE76)) to justify the omission of adjuvant chemotherapy.Conclusion: There is limited clinical utility of prognostic GEPs in grade 3 early breast cancer. Histological grading should guide decision-making for assessment of genomic risk. Citation Information: Cancer Res 2009;69(24 Suppl):Abstract nr 103.