PURPOSE: c-erbB-2 (HER-2, c-neu) might play a role as a predictive factor in breast cancer. However, the clinical utility of the marker in this disease is still not established. We conducted a critical analysis of the literature, in which we reviewed the factors that contribute to the lack of acceptance of c-erbB-2 for clinical use and attempted to determine the predictive role of c-erbB-2 for response to specific therapies. METHODS: We conducted a MEDLINE literature search using the keywords c-erbB-2, HER2, neu, and breast cancer, reviewed the references included in each publication, and reviewed abstracts that have been reported in the 1997-2000 proceedings to the American Association of Cancer Research and American Society for Clinical Oncology annual meetings. RESULTS: The preclinical and clinical data reported to date suggest that amplification or overexpression of c-erbB-2 is a weak to moderate negative pure prognostic factor. c-erbB-2 seems to be a weak to moderate negative predictive factor for response to endocrine therapy. The marker is also a moderate negative predictive factor for response to alkylating agents and a moderate positive predictive factor for response to anthracyclines. The data regarding response to taxanes or radiotherapy are not sufficient to make recommendations regarding treatment decision making. Finally, c-erbB-2 is a strong predictive factor for response to trastuzumab. CONCLUSION: We conclude that, in the adjuvant setting, c-erbB-2 status should not be used to determine whether a woman should receive adjuvant systemic therapy (weak prognostic factor). In addition, c-erbB-2 status should not be used to determine whether a patient should receive endocrine therapy. When adjuvant chemotherapy is recommended, anthracycline-based therapy should be the preferred regimen for c-erbB-2-positive patients. However, when anthracyclines are contraindicated, alkylating agent-based therapy should not be withheld. To determine the true predictive role and strength of the marker for response to each therapy, prospective randomized clinical trials or formal meta-analyses are required.