OBJECTIVE: At the end of 2000, certain diseases including leukemia were reported among soldiers who participated in the Balkan and in the Gulf wars. Depleted uranium used during these conflicts was considered as a possible cause. Its radiotoxicity is close to that of natural uranium. This paper reviews the epidemiological knowledge of uranium, the means of exposure and the associated risk of cancer. METHODS: The only available epidemiological data concerns nuclear workers exposed to uranium. A review of the international literature is proposed by distinguishing between uranium miners and other workers of the nuclear industry. French studies are described in details. RESULTS: In ionizing radiation epidemiology, contamination by uranium is often cited as a risk factor, but the dose-effect relationship is rarely studied. Retrospective assessment of individual exposure is generally insufficient. Moreover, it is difficult to distinguish between uranium radiotoxicity, its chemical toxicity and the radiotoxicity of its progeny. A causal relation between lung cancer and radon exposure, a gas derived from the decay of uranium, has been demonstrated in epidemiological studies of miners. Among other nuclear workers exposed to uranium, there is a mortality deficit from all causes (healthy worker effect). No cancer site appears systematically in excess compared to the national population; very few studies describe a dose-response relationship. CONCLUSION: Only studies with a precise reconstruction of doses and sufficient numbers of workers will allow a better assessment of risks associated with uranium exposure at levels encountered in industry or during conflicts using depleted uranium weapons.