The purpose of this study was to determine the prognostic significance of quantitative CD31 immunohistochemical assays. CD31 assays were performed on a series of 167 breast carcinoma specimens under optimal technical conditions that involved frozen sections, an automated immunoperoxidase technique, and computer-assisted analysis of digitized colored microscopic images. Results of automated quantitative immunohistochemical assays were correlated with patient follow-up (9.6 years). Patients were divided into two subgroups: those who had axillary lymph node-positive (N+) disease and those who had lymph node-negative (N-) disease. The marked immunocytochemical expression of CD31 in tumors (cutoff point, 20%) was significantly (P = .033) associated with a poor overall survival rate (Kaplan-Meier, log rank test); however, a significant association was not observed in the N+ and N- subgroups. CD31-immunostained tumor cell surfaces larger than 20% correlated with the metastasis-free survival rate (P = .004) in all patients and in the N+ subgroup (P = .005) but not in the N- subgroup. In addition, marked immunocytochemical expression of CD31 correlated with the short-term disease-free survival rate (P = .04) in the N+ subgroup but not in the N- subgroup. In multivariate analysis (proportional hazards regression, Cox model) the prognostic significance of CD31 was independent of tumor size and histologic type but not of grade. The results suggest that, under optimal technical conditions (automated and quantitative immunohistochemical assays on frozen sections), immunohistochemical expression of CD31 is a significant prognostic indicator of overall and metastasis-free survival rates. CD31 has limited prognostic value, however, and is not a completely independent prognostic indicator because it is related to nodal status.