PUBLICATION BIAS IN METAANALYSIS - A BAYESIAN DATA-AUGMENTATION APPROACH TO ACCOUNT FOR ISSUES EXEMPLIFIED IN THE PASSIVE SMOKING DEBATE

''Publication bias'' is a relatively new statistical phenomenon that o nly arises when one attempts through a meta-analysis to review all stu dies, significant or insignificant, in order to provide a total perspe ctive on a particular issue. This has recently received some notoriety as an issue in the evaluation of the relative risk of lung cancer ass ociated with passive smoking, following legal challenges to a 1992 Env ironmental Protection Agency analysis which concluded that such exposu re is associated with significant excess risk of lung cancer. We intro duce a Bayesian approach which estimates and adjusts for publication b ias. Estimation is based on a data-augmentation principle within a hie rarchical model, and the number and outcomes of unobserved studies are simulated using Gibbs sampling methods. This technique yields a quant itative adjustment for the passive smoking meta-analysis. We estimate that there may be both negative and positive but insignificant studies omitted, and that failing to allow for these would mean that the esti mated excess risk may be overstated by around 30%, both in U.S. studie s and in the global collection of studies.