Using splines to analyse the radon exposure-time-lung cancer risk in uranium miners cohort.
A. Rogel, A. Fouillet , D. Laurier, M. Tirmarche, International conference on statistics in health sciences, 23-25/06/2004, Nantes (France).
Numerous studies have shown a significant increase in death from lung cancer related to radon exposure among uranium miner’s 1. Uncertainty still remains concerning the exposure-risk relationship at low levels, in part due to imprecise knowledge of the effect of time-dependent modifying factors, such as time since exposure and age at exposure. This issue is very important for the knowledge of lung cancer risk due to indoor radon exposure. The French cohort of uranium miners is composed by a population exposed over a long duration, at level and concentration close to indoor exposure. A significant decrease in risk with time since exposure was observed 2. However, as most similar studies, the analysis was conducted using a piecewise constant model. More precise methods were proposed such as B-splines in which each year of exposure is weighted to assess the variation of exposure-risk 3. The aim of this paper is to study the sensibility of results to the choice of B-splines. The method is applied to the French cohort of uranium miners exposed to low level of radon exposure. An application to the Czech cohort of uranium miners is ongoing.