The Portuguese Thorotrast study cohort consists of a group of patients who received Thorotrast for diagnostic reasons between 1929 and 1956, and a group of similar patients who were given nonradioactive contrast agents. The cohort members were identified from medical records that contained information on reasons for the radiological investigation, type of procedure employed, and name and dose of the contrast medium used. This cohort was assembled in 1961, but its follow-up was interrupted in 1976. We have now reactivated this cohort and extended its follow-up through the end of 1996. Similar methods were used to follow up and ascertain cause of death for the Thorotrast-exposed and unexposed subjects. A total of 1931 patients who received Thorotrast systemically and 2258 unexposed subjects were initially identified from medical records. We were able to successfully follow up 58.6% (1131) of the Thorotrast patients and 45.7% (1032) of the unexposed patients. By the end of 1996, 92.2% of the Thorotrast patients and 75.2% of the unexposed patients were dead. Mortality from all causes was increased in the Thorotrast patients compared to those who were not exposed. This excess in mortality increased with time since exposure, peaking 30 years after administration of Thorotrast. The rise in overall mortality was essentially due to neoplasms [relative risk (RR) adjusted for sex, age and period = 6.04, 95% CI = 4.42-8.26], nonmalignant liver disorders (RR = 5.67, 95% CI 3.13-10.3) and nonmalignant hematological conditions (RR = 14.2, 95% CI = 2.54-79.3). The increase in mortality from neoplasms was explained mainly by increases in the risk of liver cancer (RR = 70.8, 95% CI = 19.9-251.3) and, to a much lesser extent, leukemia (RR = 15.2, 95% CI = 1. 28-181.7).