In 2001 approximately 40,000 deaths from breast cancer will occur in the United States. Although some estimates suggest possible state-to-state variations in breast cancer mortality rates the reasons for such differences remain unknown. Our objective was to confirm whether breast cancer mortality rates are significantly different by state and to identify predictors for such variation. Administrative data from the National Center for Health Statistics (NCHS) report were used to determine statewide death rates. Analyses were similarly performed with the Surveillance, Epidemiology, and End Results (SEER) cancer database to determine predictors of high versus low mortality rates. State-level variation in breast cancer mortality rates was demonstrated in the NCHS database. A subsequent analysis of high versus low mortality states in the SEER cancer registry demonstrates that stage at presentation was a significant predictor of mortality, as "high" mortality states had more patients presenting with later-stage disease. We conclude that variations in the breast cancer mortality rates exist between states. A nearly 50 per cent increase is observed between the states with the highest and lowest mortality rates. Adjusted analyses demonstrate that stage at presentation is a more important predictor of mortality variation than treatment differences. As such breast cancer mortality rates may be best improved by targeting screening and access-to-care issues rather than treatment.