The role of CA 15-3 in breast cancer has been widely studied by several authors. This study was designed to describe the predictivity of CA 15-3 determinations for the metastasis or second primary tumor in patients subjected to follow-up for breast cancer. All the 1,123 determinations of CA 15-3 carried out during 1991 in a hospital laboratory were analyzed. By cross-matching anagraphic items of patients with clinical data from different hospital databases, it was possible to reconstruct the clinical history. At the cutoff point of 40 U/ml, the positive predictive value of the CA 15-3 was 99.4%. This study shows the very close association between CA 15-3 > or = 40.0 U/ml and disease, suggesting a careful restaging when the marker increases without clinical evidence of metastases. Hormonal treatment should also be considered.