Data currently available are insufficient to demonstrate a real utility for CA 15-3 in the diagnosis, staging or surveillance of breast cancer patients following primary treatment. The aim of this study was to determine if there was a correlation between supranormal CA 15-3 serum levels and clinical and biological variables in breast cancer patients at first disease relapse. From October 1988 to March 1998, 430 consecutive patients entered the study. Overall CA 15-3 sensitivity was 60.7%. Elevated CA 15-3 levels were found more frequently in patients with liver metastases (74.6%) and in those with pleural effusion (75.7%). CA 15-3 sensitivity was 70.4% in patients with estrogen-receptor-positive (ER+) primary tumors and 45.9% in those with estrogen-receptor-negative (ER–) tumors (p < 0.0001). In patients with a limited extent of disease, marker sensitivity was 57.7% in ER+ tumors and 25.7% in ER– tumors (p < 0.0001). Logistic regression analysis showed ER status, disease extent and pleural effusion as independent variables associated with CA 15-3 positivity. The multivariate Cox analysis showed ER and disease extent as independent variables predicting overall survival, whereas CA 15-3 failed to be statistically significant. CA 15-3 was an independent variable only when the disease extent variable was removed. This study suggests that CA 15-3 in advanced breast cancer patients is a marker of both disease extent and ER status. The direct relationship with ER status indicates that CA 15-3 diagnostic sensitivity in the early detection of disease recurrence could be greater in ER+ patients than in ER– ones. Furthermore, this suggests that patients with elevated CA 15-3 levels could have disease that is more sensitive to hormone manipulation than those with normal CA 15-3 values.