Criteria for the use of Sartwell's incubation period model to study chronic diseases with uncertain etiology.

This study explores the conditions under which Sartwell's incubation period model may be appropriate for identifying a primary time period of etiologic risk for chronic diseases with uncertain etiology. The investigation begins with a description of the evolution of the application of Sartwell's model from infectious to chronic diseases. The model's underlying assumptions and some concerns about its use in the chronic disease context are specified. These concerns are addressed by data simulations and analyses of empirical data from the Connecticut Tumor Registry and the Radiation Effects Research Foundation. The results indicate that the distribution of age at diagnosis (i.e. onset) for chronic diseases is not necessarily lognormal. However, the representativeness of age distribution of the case series can affect the distribution's form; hence, it is important to determine the extent of "missing" cases, particularly those lost through truncation. Moreover, a lognormal age distribution may occur with both prenatal and age-related postnatal exposures. These findings suggest that only under certain conditions will Sartwell's model be useful in the study of chronic diseases of uncertain etiology, and indicate some caveats for interpretation of the results.

Duke Scholars

Author Gregory P. Samsa Biostatistics & Bioinformatics, Division of Biostatistics