Background: Real-life data may help checking that public investments match closely medical needs. During the last decade, several drugs have been released on the market for MBC on the basis of a potential impact on overall survival (OS). Based on the large real-life ESME cohort, we aimed to describe the time evolution of MBC OS according to main phenotypes. Methods: ESME is a unique MBC national cohort including all consecutive patients (pts) who initiated treatment for MBC between 1/01/08 and 31/12/14 in the 18 French comprehensive cancer centres. ESME collects retrospective data using clinical trial-like methodology including quality assessments. Database lock was 8/12/2016. Primary objective was the impact of year of MBC diagnosis on OS. Multivariate Cox regressions were used with adjustment for main prognostic covariates. Results: 15170 out of 16703 pts in ESME had full IHC data allowing their classification as HR+HER2- (N=9922), HER2+ (N=2863), or HR-HER2- (N=2321) cases. Median FU and OS for the whole cohort are 4.05 yrs [95 CI: 3.98-4.12], and 3.1 yrs [95 CI: 3.03-3.18] respectively. In the adjusted multivariate analysis, year of MBC diagnosis, age at MBC, subtype (using HER2+ as reference), disease-free interval (DFI), visceral involvement, and number (nbr) of metastatic sites are significant OS predictors (table) although with low effect for the first item. Age at MBC, DFI, visceral involvement, and nbr of metastatic sites remained significant prognostic variables in subtypes. Year of diagnosis was no longer significant in HR+HER2- nor HR-HER2- cases (HR=0.997, p=0.71 and HR=0.997, p=0.84), while it was highly significant in HER2+ cases (HR=0.91, p<0.0001). Conclusions: Although OS of MBC has slightly improved over the past decade, this remains mostly confined to HER2+ cases, highlighting the need for new strategies for the luminal and triple negative populations.