Trends in incidence of and mortality from invasive endometrial cancer in Sweden in 1960-84 were analyzed. The study was based on virtually all 20,371 patients given this diagnosis and 4,887 patients who died of the disease in that period. Only minor changes occurred in age-standardized incidence in pre-menopausal women, in whom the rates declined consistently during the last 15 years, especially in the youngest age groups. Among post-menopausal women, an early increase was followed by stable rates in women over 60 and decreasing rates at ages 50-59 years. In contrast, mortality rates decreased consistently over the study period. Multivariate regression analyses indicated that birth cohort was a more important determinant of incidence and mortality than was time period. The relative risk of developing endometrial cancer increased by about 20 percent in women born around 1900 as compared with 1880, and by an additional 40 percent from the 1910 cohort to the maximum risk attained in those born around 1930. In successively younger birth cohorts, the risk markedly and continuously declined. These strong birth-cohort effects after 1910 may be reasonably explained by the change from the risk-increasing estrogen-only replacement therapy introduced in the 1960s to the less harmful use, starting about 10 years later, of combined estrogen-progestogen regimens; and further, by the protective exposure of a large proportion of pre-menopausal women to oral contraceptives. Mortality, however, decreased steadily in successive cohorts from those born in 1890, indicating that the increase in incidence was referable mainly to non-lethal cancers.