Comprehensive molecular analysis of several prognostic signatures using molecular indices related to hallmarks of breast cancer: proliferation index appears to be the most significant component of all signatures 


Sotiriou C, Haibe-Kains B, Desmedt C, Wirapati P, Durbecq V, Harris A, Larsimont D, Bontempi G, Buyse M, Delorenzi M, and Piccart M.

Background. Although the development of high-throughput gene expression technologies has led to the identification of several “molecular signatures” predicting clinical outcome, no attempt has yet been made to perform a comprehensive analysis integrating well characterized biological processes and gene expression data. Here we aim to elucidate the relationship of gene expression patterns defined by several biologically relevant indices with previously reported prognostic signatures and their interaction with prognosis.

Patients and Methods. We first selected prototype genes for several biological processes in breast cancer such as basal/luminal phenotype, ERBB2, proliferation, fully captured by the gene expression grade index, stroma/invasion, angiogenesis, apoptosis and immune response (Sotiriou et al ASCO 2006). We used a multivariate linear regression model to generate significant gene lists associated with each prototype and applied them to several previously reported prognostic signatures (70-gene, 76-gene, wound healing, recurrence score, p53 and intrinsic gene-list). Multivariable analyses were used to characterize the dependency patterns between these indices for each prognostic signature and their impact on survival using several microarray datasets.

Results. All signatures comprised mainly proliferation and estrogen receptor (ER) index genes (30%-78% of all genes).  In both uni- and multivariate analysis, proliferation index was the most significant component predicting clinical outcome, achieving nearly similar performance when compared to the original signatures in the original patient series (i.e. 70-gene-signature: HR 5.5, 95% CI [3.09-9.97]; proliferation index (N=20 genes) HR 3.6, 95% CI [2.15-6.0]. Interestingly, its prognostic value far more pronounced in ER-positive tumours, which was also observed when we considered all original signatures. Proliferation index genes performance was equivalent using an external independent validation series.

Conclusion. This is the first comprehensive gene expression analysis of many existing prognostic signatures using molecular indices underlying several hallmarks of breast cancer. Proliferation seems to be the common denominator of many existing prognostic gene signatures, recapitulating their prognostic power. Furthermore, its weight seems to be far more significant in ER-positive disease.