A recent commentary on epidemiological studies of nuclear workers notes that these studies can provide radiation risk estimates that complement those derived from the study of Japanese atomic bomb survivors. The author asserts that the results from some nuclear worker studies are difficult to interpret due to the fact that ERR/Gy estimates vary across subcohorts, and subcohort-specific estimates are not always equal to estimates obtained in the overall study population. We discuss settings in which it is reasonable to expect that an estimate of association in a subcohort should be similar to an estimate obtained in the full cohort and settings in which a subcohort analysis may differ from the estimate obtained in a full cohort analysis. Focusing on the INWORKS study, we describe some of the steps taken to understand variation in estimates of ERR/Gy between subgroups and upon restrictions, as well as interpretation of estimates of external dose-mortality associations in the total study population.