Objective This epidemiologic study was undertaken after a cluster of five cases of rare forms of cancer (bone sarcoma, non-Hodgkin's lymphoma) had been observed among biomedical research workers at the Pasteur Institute in Paris to ascertain whether their disease was connected with exposure during this research. Methods A mortality study included 3765 people who worked at the Pasteur Institute between 1971 and 1986 and were followed until the end of 1987. Within this cohort a nested case-referent study included 23 cases of cancer [non-Hodgkin's lymphoma (6), multiple myeloma (I), leukemia (3), pancreatic cancer (7), bone cancer (3), brain tumor (3)], and four referents per case, matched for gender and year of birth. Results Total mortality from cancer was less than expected, the standardized mortality ratio (SMR) being 72 for the men and 82 for the women. Among the women the proportion of pancreatic cancer cases was larger than expected [SMR 490, 95% confidence interval (95%CI) 158-1144], as was the number of brain cancer cases (SMR 239, 95% CI 48-696). Among the men, mortality from bone cancer was greater than expected (SMR 553,95% CI 62-2006). In the nested case-referent study, more cases than referents had worked in the areas of molecular biology [odds ratio (OR) 7.1, 95% CI 1.5-331 and microbial genetics (OR 6.7, 95% CI 1.3-35). These cases especially included non-Hodgkin's lymphoma and bone cancer. Associated with this finding was the fact that more cases had used certain chemicals, including ethidium bromide, acrylamide, methylnitronitrosoguanidine and ethylmethanesulfonate, and radioactive compounds (essentially 32phosphorus). C O~~C~U S~O~~S As the products used are potent genotoxicants, the present findings suggest that work in biomedical research might well involve an increased risk of certain types of cancer; this conclusion should be balanced by the fact that two of the five index cases were included in the mortality study and four in the nested casereferent study.