Mechanistic modelling in large case-control studies of lung cancer risk from smoking.

The two-step clonal expansion (TSCE) model is applied to large case-control studies, frequency matched for age, which allow estimation of the RR of lung tumour risk caused by smoking. For estimating background hazard rates, mortality data from the study areas are used to supplement the case-control data. Two approaches are used to analyse the data, based on the unconditional and the conditional likelihoods. They are demonstrated to give nearly identical results. Some model diagnostics are performed and demonstrate a good model fit. Our results indicate that smoking acts on the promotion and transformation parameters, but not on the initiation parameter of the TSCE model. The fitted relative risk of current smokers peaks between ages 50 and 60 years. The relative risk of male ex-smokers decreases strongly with time since end of exposure, but does not reach the risk of non-smokers, and does not decrease as much as for female ex-smokers. Copyright 2002 John Wiley & Sons, Ltd.