CMStatistics 2018: Start Registration
View Submission - CMStatistics
Title: A Bayesian hierarchical approach to account for shared and unshared exposure uncertainty in radiation epidemiology Authors:  Sophie Ancelet - Institut de Radioprotection et de Surete Nucleaire (IRSN) (France) [presenting]
Sabine Hoffman - University of Munich (Germany)
Chantal Guihenneuc - Paris-Descartes University (France)
Abstract: Measurement error is ubiquitous and represents an important source of uncertainty in radiation epidemiology. When not or poorly accounted for, it can lead to biased risk estimates and to a distortion of the exposure-response relationship. One of the main reasons why measurement error is rarely accounted for is that classical methods lack the flexibility to account for complex patterns of exposure uncertainty. In occupational cohort studies, for instance, the type and magnitude of error can change over time depending on the methods of exposure assessment. Moreover, methods of group-level exposure estimation may give rise to errors which are shared between workers belonging to the same group or shared within workers. First, a simulation study is conducted to compare the effects of shared and unshared errors on risk estimation and shape of the exposure-response curve in proportional hazards models. Then, as a flexible framework to deal with complex error structures, several Bayesian hierarchical models are proposed to obtain corrected risk estimates on the association between exposure to radon and lung cancer mortality in the French cohort of uranium miners. The importance of making a careful characterization of shared and unshared errors to account for exposure uncertainty in risk estimates is highlighted.