Beyond cohort versus case-control: the primacy of exposure assessment over study design in occupational cancer epidemiology.

In occupational cancer epidemiology, prospective cohort studies are conventionally ranked above case-control studies. However, exposure assessment quality-central to occupational risk estimation-receives less methodological attention than study design. This review argues that divergent cohort and case-control risk estimates reflect exposure assessment limitations rather than study design. Three findings support this. First, variance decomposition demonstrates that group-level exposure assignment attenuates effect estimates by a reliability ratio approaching zero as within-worker variability becomes dominant. Second, Berkson error in job-exposure matrices, combined with model non-linearity, biases estimates toward the null. Third, the healthy worker effect, immortal time bias, and left truncation compound misclassification and shift cohort estimates further toward the null. Two case studies illustrate this pattern. For night shift work and breast cancer, the case-control odds ratio (OR) of 1.34 contrasted with the cohort relative risk (RR) of 0.98; at 30 years' cumulative exposure, these estimates were 1.88 and 1.13. For welding fumes and lung cancer, the pooled cohort RR of 1.29 contrasted with a case-control OR of 1.87; Sørensen's cohort, using individual breathing-zone measurements, recovered a hazard rate ratio of 2.34 at the highest cumulative exposure. Thus, an occupational cancer study's inferential value depends on its exposure assessment method rather than its design label. Group-level assignment attenuates effect estimates and may compromise etiological inference regardless of nominal design. Individual exposure reconstruction-through occupational interviews, industrial hygiene assessment, or personal monitoring-supports precise dose-response estimation. Quality assessment tools and meta-analyses should stratify evidence by exposure assessment method rather than design label.