The development of new anti-tumour drugs without clear cytoreductive activity has necessitated changes in the design of clinical trials. Defining the "time" parameter has become the essential objective of the majority of these trials. However, in breast cancer, this parameter is highly variable and, as such, difficult to quantify. We developed a useful tool that takes into account the inter-relatedness of all the variables known to have the capacity to predict the time-to-progression (TTP) in advanced breast cancer. From the Alamo database (GEICAM), we selected 1798 patients diagnosed as having metastatic breast cancer. Univariate analysis was performed using the method of Kaplan-Meier. Multivariate analysis was with the Cox regression method. The variables that were shown to have independent predictive value for the TTP were: non-visceral metastatic disease, single metastases, hormonal receptor positive N/T ratio<2 and disease-free interval (DFI) > or = 24 months. Taking into account the variables that had reached an independent predictive value, we constructed a model of scoring in which the patients were grouped according to the TTP. Using our new scoring model, it is possible to group patients with metastatic breast cancer according to the predicted TTP. This can be a useful tool at the time of selecting and stratifying patients on entry into new randomised clinical trials.