Biologically motivated mathematical models are important for understanding the mechanisms of radiation-induced carcinogenesis. Existing models fall into two categories: (1) short-term formalisms, which focus on the processes taking place during and shortly after irradiation (effects of dose, radiation quality, dose rate and fractionation), and (2) long-term formalisms, which track background cancer risks throughout the entire lifetime (effects of age at exposure and time since exposure) but make relatively simplistic assumptions about radiation effects. Grafting long-term mechanisms on to short-term models is badly needed for modelling radiogenic cancer. A combined formalism was developed and applied to cancer risk data in atomic bomb survivors and radiotherapy patients and to background cancer incidence. The data for nine cancer types were described adequately with a set of biologically meaningful parameters for each cancer. These results suggest that the combined short-long-term approach is a potentially promising method for predicting radiogenic cancer risks and interpreting the underlying biological mechanisms.