Tritium (3H) is a radioactive isotope of hydrogen. A number of factors combine to create a good deal of interest in the risks arising from exposure to tritium of both workers and members of the public. Tritium is ubiquitous in environmental and biological systems and is very mobile due to its occurrence as tritiated water. In this paper we systematically review epidemiological data relating to tritium exposure with a view to assessing the risk of such exposure using those studies that are potentially informative. The usefulness of the available studies of cancer and other adverse health effects in workforces and members of the general public is often impaired by a lack of tritium-specific dose data, low doses and small numbers of cases. A number of workforce studies have been identified in which tritium-specific individual doses have been estimated, although none of them, as presently reported, enable reliable inferences to be made on risks associated with exposure to tritium. In general, the available epidemiological studies on the offspring of radiation workers or on pregnancy outcome in areas subject to releases of tritium do not contain enough detail to estimate risks from tritium exposure. Although the studies presently reported are uninformative on risks from tritium, a number of the occupationally exposed cohorts would be potentially informative, particularly if data were suitably combined.