Retrospective estimates of internal doses received by workers in the nuclear industry following intake of radionuclides, based on bioassay data, are a benchmark method in epidemiological studies. Nonetheless, full information relative to thousands of people included in an epidemiological cohort is rarely available, thus implying difficulties to estimate exposure precisely. To evaluate the cumulative exposure to uranium in a cohort of the AREVA NC Pierrelatte plant workers, we compared the epidemiological Job Exposure Matrix (JEM) method with the dosimetric method based on biological monitoring of exposure for 30 workers randomly selected within the cohort. A moderate to strong correlation was observed between the estimators resulting from the two approaches, thereby validating the JEM as a tool that can be used to characterise cumulative exposure to uranium in the cohort. In addition, this study showed that the JEM is a valuable complement to the interpretation of bioassy, (1) in providing information on exposure periods as well as on physical and chemical form of the radionuclides and (2) in compensating for the lack of exposure data regarding the very earliest periods. Combining the two methods may improve the precision in reconstructing cumulative exposure for epidemiological studies.