The research proposed in this application is expected to lead to analyses of data from epidemiologic studies of nuclear workers that account for errors in external radiation dose estimates that are used in these studies. The proposed analyses will correct for bias in risk estimates resulting from both random and systematic errors in dose estimates, provide confidence limits that reflect additional uncertainty resulting from such errors, and, in the case of combined analyses of data from several studies, account for variation in the quality of dose estimates in different studies. A major objective of these worker studies is to provide a direct evaluation, based on data at low doses and dose rates, of the health risk estimates that serve as the basis for radiation protection standards. The proposed research will increase the appropriateness and validity of this evaluation. These objectives will be accomplished as follows. First, an approach for quantifying biases and uncertainties in external dose estimates will be described in a journal article that focuses on the kind of information that is needed for adjustment of epidemiologic dose-response analyses. A major contribution will be providing a feasible method of addressing sources of error that are correlated across workers and across different dosimeter readings for the same worker. This article will make methodology currently described in an extensive technical report accessible to a wider audience. Second, an approach for accounting for uncertainty in systematic bias in doses will be developed and applied to data on Hanford workers. Under this task, the correction factors developed in the report noted in the first task will be applied, and computer simulations will be conducted to allow the estimated uncertainty in these factors to be reflected in confidence limits on risk estimates. Third, a model will be developed for describing laboratory measurement errors in cumulative doses of workers. Finally, an approach for accounting for all random uncertainties in estimates of cumulative external dose used in nuclear worker studies will be developed and applied to data on Hanford workers. The proposed approach for handling laboratory uncertainties is generally similar to that recently used to account for random error in analyses of Japanese atomic bomb survivor data. For other types of uncertainties, a Berkson error model is considered more appropriate.
