Concordance among gene-expression-based predictors for ER-positive breast cancer treated with adjuvant tamoxifen.

Abstract

502 Background: ER-positive (ER+) breast cancer includes all of the intrinsic molecular subtypes, although the luminal subtypes predominate. In this study, we evaluated the prognostic ability and biologic significance of 6 independent gene-expression signatures in patients (pts) with ER+ tumors treated only with adjuvant tamoxifen. Methods: Four public microarray data sets and added clinical data were combined. Signatures evaluated: OncotypeDX, Mammaprint, PAM50, Rotterdam 76-gene, endocrine sensitivity index (SET) and the estrogen-induced gene set. We also included qRT-PCR PAM50 data from another large cohort (Nielsen et al. 2010). Distant relapse-free survival (DRFS) at 8.5 yrs was estimated by Kaplan-Meier and log-rank tests. Multivariable analyses were done using Cox regression analysis. Risk scores across the PAM50 subtypes were compared using a student's t-test. Mean Harrell's C-index after 200 repetitive cross-validations was used to evaluate the performance prediction. Results: 686 node- and 731 node+ pts were identified. PAM50 subtype distribution was LumA (46%), LumB (39%), HER2-enriched (7%), Basal-like (2%) and Normal-like (6%). Most signatures discriminated the relatively good prognosis LumA tumors from the poorer prognosis LumB/HER2-enriched/Basal-like tumors and were prognostic in all pts (p<0.001), in pts with node- (p<0.001) and node+ disease (p<0.05). In terms of performance, the majority showed similar C-Index scores in node- (∼0.73) and node+ (∼0.60) disease. A small increase in prognostic capacity in node- disease (∼0.75) was observed when these signatures were combined into a multivariable Cox model. Low-risk tumors (>90% DRFS at 8.5 yrs) were only identified within node- disease, and these were mostly (81-100%) of the LumA subtype. Conclusions: In a large ER+ patient cohort, prognostic and ER-regulated gene signatures are tracking similar biology and show significant agreement in outcome predictions; however, their performance in node+ appears minimal. From a clinical perspective, these signatures should specifically aid in identifying pts with node-negative LumA breast cancers that might be considered for adjuvant endocrine therapies alone.