Abstract

In cohort studies, the regression estimates of excess relative risk use information on background rates corresponding to zero exposures. These are usually estimated internally from cohorts. This needs some stratification by factors like age, calendar period or birth cohort. In addition, such background rates depend on the regression model. The aim of the presentation is to assess which factors are substantial for the stratification. Results are given form Czech, French, and German cohorts of uranium miners. In the model that takes into account cumulated radon exposure and calendar period specific excess relative risks including modifying effects of age at exposure and time since exposure, the effect of calendar year stratification period was found non-significant (p = 0.23). This is in line with approaches used in evaluation of 11 studies of underground miners by the National Cancer Institute in 1994 that used only age stratification. In addition, the estimates of excess relative risk per unit exposure for strata excluding calendar year were higher in the present joint study by 23% from estimates when age, calendar year, and cohort are used in stratification. The likely reason is that calendar period stratification remove the calendar period variability in exposure.