This study examines the relationship between fatal breast cancer and use of estrogen replacement therapy (ERT) among women in a large prospective study in the United States. After nine years of follow-up, 1,469 breast cancer deaths were observed in a cohort of 422,373 postmenopausal women who were cancer free at study entry and who supplied information on estrogen use. Results from Cox proportional hazards modeling, adjusted for 11 other potential risk factors, showed that ever-use of ERT was associated with a significantly decreased risk of fatal breast cancer (rate ratio [RR] = 0.84, 95 percent confidence interval [CI] = 0.75-0.94). There was a moderate trend (P = 0.07) of decreasing risk with younger age at first use of ERT. This decreased risk was most pronounced in women who experienced natural menopause before the age of 40 years (RR = 0.59, CI = 0.40-0.87). There was no discernible trend of increasing risk with duration of use in estrogen users at baseline or former users, nor was there any trend in years since last use in former users. The relationship between ERT and breast cancer mortality differed by age at menarche and by a self-reported history of breast cysts. No increased risk of fatal breast cancer with ERT was observed with estrogen use status (baseline/former), age at first use, duration of use, or years since last use. These findings suggest that ever-use of ERT is associated with a 16 percent decreased risk of fatal breast cancer.