PURPOSE: A systematic review of randomized clinical trials (RCTs) was undertaken to assess the effectiveness of medical treatment for metastatic breast cancer. METHODS: RCTs published between 1975 and 1997 have been classified according to 12 therapeutic comparisons: (1) polychemotherapy (PCHT) agents versus single agent; (2) PCHT regimens with anthracycline versus PCHT without anthracycline; (3) other PCHT versus cyclophosphamide, methotrexate, and fluorouracil (CMF); (4) chemotherapy (CHT) with epirubicin versus CHT with doxorubicin; (5) CHT versus same CHT delivered with less intensive schedules; (6) other endocrine therapy (OET) versus tamoxifen; (7) OET plus tamoxifen versus tamoxifen alone; (8) OET versus medroxyprogesterone; (9) OET versus aromatase inhibitors; (10) OET versus megestrol; (11) endocrine therapy (ET) versus same ET at lower doses; and (12) CHT plus ET versus CHT. Tumor response rates, mortality hazards ratio (HR) and frequency of severe side effects were the outcome measures. RESULTS: A total of 189 eligible trials (31,510 patients) were identified. All provided response rates and 133 (70%) data or survival curves needed for calculation of the HR. In eight of 12 comparisons, statistically significant differences for response emerged (1, 2, 3, 5, 7, 8, 11, 12); all but no. 8 favored the first term of the comparison. Overall survival analysis showed better results of (a) PCHT versus single-agent CHT (HR=0.82; 95% confidence interval [CI], 0.75 to 0.90); (b) CHT with doxorubicin versus CHT with epirubicin (HR=1.13; 95% CI, 1.00 to 1.27); (c) CHT versus the same CHT delivered with less intensive schedules (HR=0.90; 95% CI, 0.83 to 0.97); (d) ET versus the same ET at lower doses (HR=0.86; 95% CI, 0.77 to 0.97). Quality of life was measured in only 2,995 of 31,510 patients (9.5%). CONCLUSION: Despite some evidence of effectiveness of specific regimens, the relevance of these findings is limited by the modest survival benefit and the lack of evaluation of the quality-of-life impact of these treatments.