Summary The association between breast and endometrial cancer was investigated in a cohort consisting of 60,065 subjects (99% of all women in whom a first breast cancer was diagnosed in Sweden in 1960-63 and 1968-81 (Dunn, 1975;Cole & Cramer, 1977;Howe et al., 1984;Henderson et al., 1982;Willett & MacMahon, 1984). Descriptive epidemiologic studies have revealed high correlations between breast and endometrial cancer with regard to incidence and mortality rates. Such observations have been reported from international investigations in both low-and high-risk countries and in migrant populations (Dunn, 1975) and also from different parts of high-risk countries such as Canada (Howe et al., 1984) and the United States (Hoover et al., 1975;Winkelstein et al., 1977).The possibility that these demographic correlations reflect common genetic or environmental aetiologic factors has gained some support from cohort studies in which an increased risk of endometrial cancer has been found in breast cancer patients (Schoenberg et al., 1969) and vice versa (Schottenfeld & Berg, 1971;Bailar, 1963;MacMahon & Austin, 1969). The knowledge that can be derived from these analytic studies is generally uncertain, however, because of small numbers of cases, contradictory findings in blacks and whites (Newell et al., 1974), and limited information as to the possible dependence of an association between breast and endometrial cancer on age at diagnosis of the first primary (Bailar, 1963;MacMahon & Austin, 1969). The aim of this investigation was to confirm and extend in a larger cohort our recent finding that the association between breast and endometrial cancer is seemingly restricted to older women (Adami et al., 1984). The availability of a national cancer registry, reliable incidence figures for the studied population and opportunities for complete long-term follow-up facilitated our analysis. The overall frequency of underreporting has been shown to be -5% (Mattsson, 1977) and the completeness of registration of breast and endometrial cancer has been assessed from death certificates to be 98 and 95% respectively (Mattsson & Wallgren, 1984).The cohort was based on all women reported as having a first breast cancer diagnosed in 1960 through 1963 and 1968 through 1981. The four-year period of our previous study (Adami et al., 1984), 1964 through 1967, was excluded, since one major aim was to rule out the possibility that the excess risk previously found was due to chance (type I error) i.e. to a random high relative risk among the large number of tumour sites and subgroups that were analysed. Patients were included irrespective of whether they had had any other malignant disease reported prior to the breast cancer.A total of 61,341 women with breast cancer were notified to the Cancer Registry during the period of study. We excluded 583 women with an incomplete national registration number (a number which permits exclusive identification), precluding computerized linkages and followup in the Cancer Registry and other registers. The b...