[Epidemiological model development using BSE as an example--observations from a statistical viewpoint]

Since first BSE cases in cattle born in Germany were recognized, questions have been raised concerning the future development of the disease and the epidemiological dynamics of BSE, and, consequently, modelling approaches that might answer these questions. The database for such modelling efforts is formed by BSE incidence numbers or incidence rates, broken down by age at onset of clinical disease, and by time of onset or time of birth, respectively, from available information gathered for suspect and confirmed BSE cases. To describe such data, statistical age-period-cohort-models and two epidemiologically/biologically oriented modelling approaches are discussed: the so-called three-factor-model used by the Central Veterinary Laboratory of the British Ministry for Agriculture, Fisheries and Food (MAFF) (now Department of the Environment and Rural Affairs (DEFRA)), and a back-calculation-model developed by a working group at the University of Oxford. Resulting model calculations are supposed to serve several purposes, including a prediction of future BSE incidence numbers, and, especially based on the "Oxford"-model, a back-calculation of the epidemic of BSE infections from the epidemic of clinically diseased BSE cases. Analysis of these approaches reveals some problems even to identify unique age, period, or birth cohort effects. An additional estimation of epidemiological components of BSE, for example the frequency distribution of incubation times, has to rely on further assumptions that cannot be validated by the model fit as such. Therefore, modelling results should be interpreted with caution. However, the limitations demonstrated by this discussion emphasize the need for specific studies to investigate certain aspects of the BSE epidemic, for example the distribution of times from infection to disease onset, and for the centralised collection of valid and detailed population data for cattle.