Multistage modelling of lung cancer mortality in asbestos textile workers.

The Armitage-Doll multistage model of cancer is applied to data from a cohort study of lung cancer in 1261 white male workers from one asbestos textile manufacturing plant. Three approaches are used: induction time analysis; analysis of the relationship of the excess incidence rate to age at first exposure and time since first exposure; and direct fitting of the Armitage-Doll model. Poisson regression was used for all analyses. The induction time analysis was conducted using Rothman's 'window of exposure' method. This suggested that the increase in rate ratio was primarily due to exposures occurring 15-24 years previously, whereas there was little effect from exposures occurring 0-14 or 25+ years previously. The excess incidence rate increased both with age at first exposure and time since first exposure, suggesting that asbestos acted at a stage intermediate between the first and penultimate stages. Direct fitting of the Armitage-Doll model suggested that the best fit was obtained by assuming that asbestos acts at stage 3, 4 or 5 of a six-stage process. Most analyses of the type presented here are unable to determine at which stage a carcinogen acts, due to the small numbers of cancer deaths occurring in typical occupational cohorts. Furthermore, there is reason to doubt the validity of the Armitage-Doll model. However, such analyses can at least suggest whether a carcinogen appears to act at an early, intermediate or late stage, and the general statistical methods applied here will retain their usefulness as further models are developed and larger data sets become available.