Search

Cohort-based dose-response analyses can be biased if based on a comparison group that is not comparable to the exposed persons with respect to uncontrolled factors related to disease incidence or mortality. When data exist over a range of doses including the very low dose region, internal regression standardized analyses based on the regression intercept derived from the exposed subcohort alone can provide risk estimates that are not subject to such comparison-group bias. In the Life Span Study cohort of atomic-bomb survivors, persons with dose estimates of zero comprise a broader geographic distribution than that of persons with non-zero dose estimates. Because there is geographic variation in mortality rates, the zero-dose persons might bias background rate estimates thereby affecting inference about radiation risk. This is illustrated using mortality due to all causes. Restricting the comparison group to certain geographically defined subcohorts resulted in as much as a 6% increase or 8% decrease in the risk estimate. This bias can be corrected using an SMR-type estimate in the regression model, allowing retention of the comparison group in the analysis if it is needed for stability or precision in estimating age, time, and sex effects. Consideration of heterogeneity in comparison groups is particularly important in dose-response studies focused on low doses at which the response may be comparable in magnitude to such heterogeneity.

We used the EM algorithm in the context of a joint Poisson regression analysis of cancer and non-cancer mortality in the Radiation Effects Research Foundation (RERF) Life Span Study (LSS) to assess whether the observed increased risk of non-cancer death due to radiation exposure (Shimizu et al., RERF Technical Report 02-91, 1991) can be attributed solely to misclassification of cancer as non-cancer on death certificates. We show that greater levels of dose-independent misclassification than are indicated by a series of autopsies conducted on a subset of LSS members would be required to explain the non-cancer dose response, but that a relatively small amount of dose-dependence in the misclassification of cancer would explain the result. The adjustment for misclassification also results in higher risk estimates for cancer mortality. We review applications of similar statistical methods in other contexts and discuss extensions of the methods to more than two causes of death.