To examine the time-dependent effects of exposure histories on disease we use sliding time windows as an exploratory alternative to the analysis of variables like time since last exposure and duration of exposure. The method fits a series of risk models which contain total cumulative exposure and an additional covariate for exposures received during fixed time intervals. Characteristics of the fitted models provide insight into the influence of exposure increments at different times on disease risk. A simulation study is performed to check the validity of the approach. We apply the method to data from a recent German case-control study on smoking and lung cancer risk with about 4300 lung cancer cases and a similiar number of controls. The sliding time window approach indicates that the amount of cigarettes smoked from two to 11 years before disease incidence is most predicitive of lung cancer incidence. Among different smoking profiles that result in the same lifelong cumulative number of cigarettes smoked, those with a concentration of smoked cigarettes within 20 years before interview bear substantially larger risk than others.
Copyright 2000 John Wiley & Sons, Ltd.