The aim of this retrospective cohort study was to investigate whether survival of patients with breast cancer has changed over the period 1975-89. A total of 2604 women diagnosed as having invasive breast cancer at a clinical oncology unit in London were followed up for between 5 and 20 years. Patients were divided into four groups according to menstrual status (pre or post) and the staging of cancer (operable or inoperable). For each group, survival from diagnosis was compared between three consecutive 5-year cohorts, both with and without adjustments made for relevant prognostic factors. No temporal patterns were found in patients with inoperable cancer, in whom the survival rate was consistently low. Of women with operable cancers, differences were seen only among post-menopausal women, for whom the best survival patterns were seen in patients diagnosed between 1985-89. This is probably due to tamoxifen being commonly prescribed as adjuvant treatment for this cohort of patients. We cannot explain an apparently worse survival in the group of patients presenting in the early 1980s compared with that observed in the late 1970s.