Analysis of Strategies for Reconstructing Exposures

Abstract Many retrospective epidemiologic studies require the reconstruction of exposures over long periods of time on the basis of incomplete or absent data. The modern requirements of acceptably accurate, quantitative exposure estimations for the construction of exposure-response relationships and extrapolation of estimated risks to untested regions suggest the existence of a conflict between the study requirements and the more or less qualitative reconstruction methodologies. While the importance of exposure reconstruction is accepted, there is no structured analysis of the logical basis of the reconstruction process. In an attempt to analyze exposure reconstruction strategies (ERS) mathematically, a formal set of properties for ERSs are developed. These properties are presented in the form of three necessary and six desirable/undesirable general attributes of any exposure reconstruction process. Even a cursory analysis of these formal attributes suggests that a careful analysis of the exposure reconst...