C-erbB-2 serum levels were studied in the sera of 50 healthy subjects, 56 patients with benign breast diseases and 412 patients with breast cancer. Using 15 U/ml as the cut-off, no healthy subjects, patients with benign disease and only 2.4% of patients with no-evidence of disease had serum levels higher than this cut-off point. Abnormal c-erbB-2 levels were found in 9.2% of the patients with locoregional breast carcinoma and in 45.4% of those with advanced disease. C-erbB-2 serum levels in patients with locoregional breast cancer were not related to tumor size or nodal involvement. By contrast, significantly higher c-erbB-2 serum levels were found in ER- or PgR- tumors than in those ER+ or PgR+ tumors, in both locoregional or metastatic tumors. The correlation between serum and tissue levels of C-erbB-2 was studied in the tumors of 161 patients. Significantly higher c-erbB-2 serum levels were found in patients with overexpression in tissue by immunohistochemistry, in both locoregional and advanced disease (p = 0.0001). In patients with C-erbB-2 overexpression in tissue, c-erbB-2 serum levels were related to tumor size and nodes, with higher values in tumors greater than 5 cm or in those with more than 3 nodes involved. When the prognostic value of this oncoprotein was evaluated, patients with abnormally high presurgical c-erbB-2 had a worse prognosis than those patients with normal values, in both node-negative and node-positive patients. Serum concentrations in patients with advanced disease, were related to the site of recurrence with significantly higher values in patients with metastases (mainly in those with liver metastases) than in those with locoregional recurrence. In summary, c-erbB-2 serum level seem to be a useful tumor marker in the prognosis of patients with breast cancer.