Methods for analyzing occupational cohort data with application to lung cancer in U.S. uranium miners

We used two methods to examine how lung cancer death rates vary with cumulative exposures to radiation and tobacco among U.S. uranium miners. We assumed that exposures act proportionately on age-specific death rates among nonsmokers who are unexposed to uranium. The two methods produced similar results when death rate ratios were taken to be the product of radiation and tobacco effects. The estimates were discrepant when death rate ratios were taken to be the sum of radiation and tobacco effects. Both methods indicated better fit for the multiplicative model, as judged by the maximized loglikelihood values. Death rates estimated in this way for white males in the absence of mining and smoking increased only weakly with age. This weak relation suggests that our models of death rates fit poorly. It may be that cumulative exposures are inappropriate measures of the effects of radiation and tobacco on lung cancer death rates.