Many cancer-risk factor associations are known to be modified by aspects of time, including age at first exposure or age at risk, time since first exposure or since exposure ceased, duration of exposure, and calendar time. In addition, many exposures vary in intensity over time, raising the possibility that the pattern of exposure (ex.: whether intermittent or continuous, whether increasing or decreasing, etc.) might also modify cancer risks. All these aspects bear on the choice of a suitable index of exposure and risk model for epidemiologic analysis. Furthermore in some cases, time itself may be the object of inquiry, as in descriptions of the latent period and whether it is affected by the intensity of exposure. Despite considerable evidence of the importance of such time-related factors in cancer epidemiology, relatively little is known about appropriate methods of analysis and most published reports either ignore them or use rather simplistic models. To address these issues, a Symposium on Time-Related Factors in Cancer Epidemiology was recenty held and this proposal represents an outgrowth of ideas discussed there. As methodological contributions, we propose (1) to develop methods of constructing exposure indices for long-term exposure histories based on models of carcinogenesis and to study their performance using simulation; (2) to develop alternatives to the proportional hazards model, such as those in which exposure and time effects combine additively; (3) to develop simple methods for describing the evolution of risk over time and to explore their use in examining the fit of parametric models; and (4) to study the implications of time-related factors for the design of case-control studies. We will also apply the methods we have developed to a number of epidemiologic studies of cancer available to us, evaluating such factors as occupational exposures, hormonal and reproductive events, medical treatments, and lifestyle (smoking, sunlight, etc.).
Showing the most recent 10 out of 20 publications
