The antigen levels of components of the urokinase-type plasminogen activator (uPA) system of plasminogen activation are correlated with prognosis in several types of cancers, including breast cancer. In the present study involving 2780 patients with primary invasive breast cancer, we have evaluated the prognostic importance of the four major components of the uPA system [uPA, the receptor uPAR (CD87), and the inhibitors PAI-1 and PAI-2]. The antigen levels were determined by ELISA in cytosols prepared from primary breast tumors. The levels of the four factors significantly correlated with each other; the Spearman rank correlation coefficients (r(s)) ranged from 0.32 (between PAI-2 and PAI-1 or uPAR) to 0.59 (between uPA and PAI-1). The median duration of follow-up of patients still alive was 88 months. In the multivariate analyses for relapse-free survival (RFS) and overall survival (OS), we defined a basic model including age, menopausal status, tumor size and grade, lymph node status, adjuvant therapy, and steroid hormone receptor status. uPA, uPAR, PAI-1, and PAI-2 were considered as categorical variables, each with two cut points that were established by isotonic regression analysis. Compared with tumors with low levels, those with intermediate and high levels showed a relative hazard rate (RHR) and 95% confidence interval (95% CI) of 1.22 (1.02-1.45) and 1.69 (1.39-2.05) for uPA, and 1.32 (1.14-1.54) and 2.17 (1.74-2.70) for PAI-1, respectively, in multivariate analysis for RFS in all patients. Compared with tumors with high PAI-2 levels, those with intermediate and low levels showed a poor RFS with a RHR (95% CI) of 1.30 (1.14-1.48) and 1.76 (1.38-2.24), respectively. Similar results were obtained in the multivariate analysis for OS in all patients. Furthermore, uPA and PAI-1 were independent predictive factors of a poor RFS and OS in node-negative and node-positive patients. PAI-2 also added to the multivariate models for RFS in node-negative and node-positive patients, and in the analysis for OS in node-negative patients. uPAR did not further contribute to any of the multivariate models. A prognostic score was calculated based on the estimates from the final multivariate model for RFS. Using this score, the difference between the highest and lowest 10% risk groups was 66% in the analysis for RFS at 10 years and 61% in the analysis for OS. Moreover, separate prognostic scores were calculated for node-negative and node-positive patients. In the 10% highest risk groups, the proportion of disease-free patients was only 27 +/- 6% and 9 +/- 3% at 10 years for node-negative and node-positive patients, respectively. These proportions were 86 +/- 4% and 61 +/- 6% for the corresponding 10% lowest risk groups of relapse. We conclude that several components of the uPA system are potential predictors of RFS and OS in patients with primary invasive breast cancer. Knowledge of these factors could be helpful to assess the individual risk of patients, to select various types of adjuvant treatment and to identify patients who may benefit from targeted therapies that are currently being developed.