PurposeRecent data have shown that the expression levels of long noncoding RNAs (lncRNAs) are associated with tamoxifen sensitivity in estrogen receptor (ER)-positive breast cancer. Herein, we constructed an lncRNA-based model to predict disease outcomes of ER-positive breast cancer patients treated with tamoxifen.
MethodsLncRNA expression information was acquired from Gene Expression Omnibus by re-mapping pre-existing microarrays of patients with ER-positive breast cancer treated with tamoxifen. The distant metastasis-free survival (DMFS) predictive signature was subsequently built based on a Cox proportional hazard regression model in discover cohort patients, which was further evaluated in another independent validation dataset.
ResultsSix lncRNAs were found to be associated with DMFS in the discover cohort, which were used to construct a tamoxifen efficacy-related lncRNA signature (TLS). There were 133 and 362 patients with TLS high- and low-risk signatures in the discover cohort. Both univariate and multivariate analysis demonstrated that TLS was associated with DMFS. TLS high-risk patients had worse outcomes than low-risk patients, with a hazard ratio of 4.04 (95% confidence interval, 2.83-5.77; p<0.001). Both subgroup analysis and receiver operating characteristic analysis indicated that TLS performed better in lymph node-negative, luminal B, 21-gene recurrence score high-risk, and 70-gene prognosis signature high-risk patients. Moreover, in a comparison of the 21-gene recurrence score and 70-gene prognosis signature, TLS showed a similar area under receiver operating characteristic curve in all patients. Gene Set Enrichment Analysis indicated that TLS high-risk patients showed different gene expression patterns related to the cell cycle and nucleotide metabolism from those of low-risk patients.
ConclusionThis six-lncRNA signature was associated with disease outcome in ER-positive breast cancer patients treated with tamoxifen, which is comparable to previous messenger RNA signatures and requires further clinical evaluation.
SUBMITTER: Wang G