OBJECTIVE To obtain further information about the risks of cancer associated with occupational exposure to ethylene oxide METHODS Follow up was extended by 13 years for a cohort of 2876 men and women with definite or potential exposure to ethylene oxide in the chemical industry or in hospital sterilising units. Subjects were traced through National Health Service and social security records, and their mortality was compared with that expected from rates in the national population by the person-years method. RESULTS Analysis was based on 565 deaths, of which 339 had occurred during the additional period of follow up. Mortality was close to or below expectation for all causes (565 deaths v 607.6 expected), all cancers (188 v 184.2), and for all specific categories of malignancy including stomach cancer (10 v 11.6), breast cancer (11 v 13.2), non-Hodgkin's lymphoma (7 v 4.8), and leukaemia (5 v 4.6). All five deaths from leukaemia occurred in the subset of subjects with greatest potential for exposure to ethylene oxide, but even in this group the excess of deaths was small (2.6 expected). CONCLUSIONS The balance of evidence from this and other epidemiological investigations indicates that any risk of human cancer from ethylene oxide is low, particularly at the levels of occupational exposure that have occurred in Britain over recent decades. This may reflect the capacity of human cells to repair DNA damage caused by the chemical, which is a potent genotoxin and animal carcinogen.