This study was to determine whether the proportion of death due to breast cancer changed over time in different cohorts of women diagnosed with breast cancer. We identified 316,149 women with breast cancer at age 20 or older during 1975-2003 from the Surveillance, Epidemiology, and End Results 9 tumor registries in the United States. Logistic regression models were used to assess the effects of time period on the likelihood of dying because of breast cancer as underlying cause of death, adjusting for other factors. Overall, underlying cause of death was 52.8% due to breast cancer, 17.8% due to heart disease, and 4.9% due to stroke. Percentage of death due to breast cancer did not change significantly from 1975 to 2003 in those who died < 12 months after diagnosis, but decreased significantly in women who died between 1 and 15 years. Risk of death due to breast cancer in women diagnosed during 1995-1998 was significantly lower than those in 1975-1979 (odds ratio = 0.79, 95% confidence interval = 0.70-0.89), after adjusting for age, race, ethnicity, and tumor stage. Percentage of death due to breast cancer decreased significantly with age from 87.5% in women < 40% to 30.7% in those 80 or older, which was not significantly affected by year of diagnosis. Proportion of death due to breast cancer increased with advanced tumor stage and was similar in various racial/ethnic groups of population. The findings demonstrated that the impact of breast cancer on overall death was reduced after 1 year of diagnosis, but suggested the need for continued cancer surveillance.