“…These results parallel those of a meta-analysis published in 2002 and other reviews published around the same time. 1,4,15 Because risk estimates in industrybased studies are usually based on comparison with an external population, uncontrolled confounding is virtually impossible to exclude with any level of confidence. 46 Community-based studies of case-control design, on the other hand, are likely to suffer from bias (eg, from selection of cases or controls, retrospective assessment, and absence of validation of exposure), from low-reported prevalence of exposures of interest or jobs related to such exposures, and from residual confounding (eg, by other occupational exposures).…”