A study of the incubation period, or age at onset, of the transmissible spongiform encephalopathies/prion diseases.

In order to model epidemics of infectious diseases, particularly to estimate probable numbers of cases with onset at any particular time, it is necessaiy to incorporate a term for the incubation period frequency distribution. Sartwell's hypothesis states that the incubation period frequency distribution for infectious disease is generally a log-normal distribution, based on his examination of disease with short incubation periods. However, it may not apply to diseases with long incubation periods. During the course of an epidemic of a disease with a long incubation period, left and right censoring makes direct observation of the frequency distribution highly unreliable; in addition, time of infection is often unknown. Therefore, for a previously undescribed disease, methods other than direct observation must be employed. One method is to extrapolate from information available for other diseases. In evaluation of Sartwell's hypothesis as applied to diseases with long incubation periods, examination of transfijsion-associated AIDS data was inconclusive. Examination of data for experimental transmissible spongiform encephalopathy (TSE)/prion disease in several species suggests that it may not apply. For natural TSE/prion disease, age at onset is used generally as a 'proxy' for incubation period since infection time is rarely known; the validity of this may vary with the disease type and species being examined. Using this measure, again Sartwell's hypothesis was not confirmed. For both incubation period and age at onset, evidence presented suggests that observed frequency distribution coefficient of skewness is associated with modal age at onset (and thus indirectly with prior age at infection, where appropriate), an earlier modal age at onset resulting in a larger observed coefficient of skewness. The relationship of this association with Sartwell's findings is discussed; they are not incompatible. In addition, an association between observed coefficient of skewness and sample size is demonstrated and the implications discussed.