The relationship between pregnancy and the outcome of breast cancer remains controversial. The purpose of this study was to determine the prognostic value of pregnancy at the time of diagnosis of primary infiltrating breast cancer. In a retrospective multi-center study we compared a group of 154 patients presenting pregnancy-associated (PA) breast cancer with a control group of 308 patients presenting non-pregnancy-associated (non-PA) breast cancer. Classic prognostic factors, treatment modalities, disease-free survival and overall survival were compared in the 2 groups. The relative importance of pregnancy was assessed by Cox multivariate analysis. There was a significantly higher proportion of inflammatory breast cancer, large tumors and negative receptor status in the PA group. Five-year recurrence-free survival, metastasis-free survival and overall survival were lower both in the whole PA group and in the PA sub-group excluding patients with inflammatory breast cancer than in the corresponding non-PA groups. According to clinical stage, histoprognostic grade and microscopic lymph-node involvement, probability of 5-year metastasis-free survival and overall survival was lower in the PA group. Outcome was significantly poorer after chemotherapy for patients in the PA sub-group than in the non-PA sub-group. Multivariate analysis demonstrated that pregnancy was an independent and significant prognostic factor. Pregnancy has an adverse effect on the outcome of breast cancer. Concurrent or recent pregnancy should be taken into account in the development of new systemic therapies. Our findings have important implications for further research into the basic mechanisms of cancer. Int. J. Cancer 72:720–727, 1997. © 1997 Wiley-Liss, Inc.