Accounting for uncertainty in systematic bias in exposure estimates used in relative risk regression.
Pacific Northwest Laboratory, U.S. Department of Energy, Battelle Memorial Institute 1995 Dec; :1-21
Techniques for addressing systematic biases in exposure estimates used for establishing exposure response relationships in epidemiologic studies were developed. The aim was to have strategies to deal with uncertainties associated with constructing dose response estimates for nuclear workers exposed to external ionizing radiation in studies examining the carcinogenic risk of exposure to ionizing radiation. Two approaches based on constructing 90 and 95% confidence intervals (CIs) for uncertainties in the dose estimates were used. The first approach used numerical methods based on the likelihood ratio statistic to compute the CIs. The second approach utilized computer simulations based on the likelihood ratio statistic to compute the CIs. Both approaches were applied to data obtained in a cohort study of leukemia in workers employed at the Hanford laboratory, Washington State, between 1944 and 1980, to a combined dataset of workers at the Hanford site, Oak Ridge National Laboratory, and the Rocky Flats weapons plant, and to simulated data sets created to examine the effects of variations in population size and exposure estimates on the risk estimates. When applied to the occupational data, both methods indicated that sampling errors accounted for most of the uncertainty in the risk estimates. Accounting for statistical biases, however, did not significantly alter the CIs for the estimates. In the simulated dataset, the CIs increased as the number of subjects with exposure data increased. The author concludes that accounting for statistical bias is very important when analyzing exposure response data obtained from a large population of workers.
NIOSH-Grant; Grants-other; Statistical-analysis; Radiation-exposure; Ionizing-radiation; Dosimetry; Dose-response; Laboratory-workers; Epidemiology; Statistical-analysis; Malignant-neoplasms; Risk-analysis; Mathematical-models
Health Risk Assessment Dept Battelle Memorial Inst PO Box 999 Msin P7 82 Richland, WA 99352
Final Grant Report
Other Occupational Concerns; Grants-other;
Pacific Northwest Laboratory, U.S. Department of Energy, Battelle Memorial Institute
Battelle Pacific Northwest Laboratories, Richland, Washington