The relationship between exposure to low-linear energy transfer ionizing radiation and subsequent breast cancer mortality risk is reported based on a further 7 years of follow-up in the Canadian fluoroscopy study. Amongst 31,917 women first treated for tuberculosis in a Canadian institution between 1930 and 1952, a total of 688 breast cancer deaths were observed between 1950 and 1987. There is a strong linear trend of increasing risk with increasing dose (P < 0.0001), with the excess relative risk per sievert decreasing with age at exposure (P = 0.0003). The excess relative risk is approximately constant between 5 and 39 years after exposure, with a suggestion of a decrease between 40 and 57 years after exposure, though this could be a chance effect (P = 0.22). Combined analyses of the Canadian fluoroscopy data and the data for the atomic bomb survivors with respect to breast cancer mortality are also reported. In general the two studies are reasonably consistent, the only distinct difference being the much greater excess relative risk per sievert amongst women exposed to very high doses in the province of Nova Scotia (P, heterogeneity <0.0001). Based on the combined data sets a simple relative risk (RR) model for the effect of a dose of D sieverts at age A years is developed: RR(D) = 1.0 + 0.52D exp[-0.10(A-15)]. This model fits the combined data well, and is used to predict excess lifetime risks of breast cancer mortality after radiation exposure from routine annual mammography.