We estimated the costs and health effects of treating stage I, II, III, and IV breast cancer individually, of treating all stages, and of introducing an extensive cancer control program (treating all stages plus early stage diagnosis) in three epidemiologically different world regions--Africa, North America, and Asia. We developed a mathematical simulation model of breast cancer using the stage distribution and case fatality rates in the presence and absence of treatment as predictors of survival. Outcome measures were life-years adjusted for disability (DALYs), costs (in 2000 U.S. dollars) of treatment and follow-up, and cost-effectiveness ratios (CERs; in dollars per DALY averted). Sensitivity analyses were performed to determine the robustness of the results. Treating patients with stage I breast cancer resulted in 23.41, 12.25, and 19.25 DALYs averted per patient in Africa, North America, and Asia, respectively. The corresponding average CERs compared with no intervention were 78 U.S. dollars , 1,960 U.S. dollars, and 62 U.S. dollars per DALY averted. The number of DALYs averted per patient decreased with stage; the value was lowest for stage IV treatment (0.18-0.19), with average CERs of 4,986 U.S. dollars in Africa, 70,380 U.S. dollars in North America, and 3,510 U.S. dollars per DALY averted in Asia. An extensive breast cancer program resulted in 16.14, 12.91, and 12.58 DALYs averted per patient and average CERs of 75 U.S. dollars, 915 U.S. dollars, and 75 U.S. dollars per DALY averted. Outcomes were most sensitive to case fatality rates for untreated patients, but varying model assumptions did not change the conclusions. These findings suggest that treating stage I disease and introducing an extensive breast cancer program are the most cost-effective breast cancer interventions.