In breast cancer, several investigations have demonstrated that the tumour biological factors uPA urokinase-type plasminogen activator) and its inhibitor PAI-1 are statistically independent, strong prognostic factors for disease-free (DFS) and overall survival (OS). However, statistical analyses performed for varying follow-up periods suggested a time variation of prognostic strength. We therefore investigated the time-dependent prognostic power of uPA, PAI-1 and steroid hormone receptor status applying the time-varying coefficient model of Gray. uPA and PAI-1 were analysed by enzyme-linked immunosorbent assay in tumour tissue extracts from 314 breast cancer patients. Hormone receptors (oestrogen and progesterone) were determined by radioligand binding or by immunohistochemistry. Univariate and multivariate analyses (Cox proportional hazards model) of DFS and OS were performed for all patients, including 147 node-negative patients. Median follow-up of patients still alive at time of analysis (n = 232) was 58 months. Although initially of high prognostic impact, a continuous decrease over time in the prognostic power of hormone receptor status and uPA was observed. In contrast, the prognostic impact of PAI-1 increased over time and reached similar strength as the lymph node status. The time-dependent risk profile of prognostic factors may have important clinical implications in regard to follow-up and patients' individual risk situation. Evaluation of time dependency of prognostic factors may also give a more profound insight into the dynamics of breast cancer metastasis.