Compartment model approaches for estimating the parameters of a chronic disease process under changing risk factor...

Compartment model approaches have been proposed for the analysis of the age incidence of specific types of cancer. These models represented the age increases in incidence as the result of a compound hazard function where individual level risks were described by the Weibull hazard function and where the population level hazard rate is a continuous mixture of the Weibull hazards. These formulations assumed that the mixing function, which described differences in risk due to different exposure histories, was constant after the age at which the model was first applied. In this paper we show how the mixing distribution can be allowed to change with time reflecting changing exposures. The model is fitted to U.S. lung cancer mortality data where for recent male cohorts there appear to be changing patterns of exposure possibly related to recent declines in male smoking. The implications for future lung cancer mortality trends in the United States are discussed.