Cancer risks were studied in 834 thyroid cancer patients given 131I (4,551 MBq, average) and in 1,121 patients treated by other means in Sweden between 1950 and 1975. Record-linkage with the Swedish Cancer Register identified 99 new cancers more than 2 years after 131I therapy [standardised incidence ratio (SIR) = 1.43; 95% confidence interval (CI) 1.17-1.75] vs 122 (SIR = 1.19; 95% CI 0.88-1.42) in patients not receiving 131I. In females treated with 131I overall SIR was 1.45 (95% CI 1.14-1.83) and significantly elevated were noted for tumours of the salivary glands, genital organs, kidney and adrenal gland. No elevated risk of a subsequent breast cancer or leukaemia was noted. SIR did not change over time, arguing against a strong radiation effect of 131I. Organs that were estimated to have received more than 1.0 Gy had together a significantly increased risk of a subsequent cancer following 131I treatment (SIR = 2.59; n = 18). A significant trend was seen for increasing activities of 131I with highest risk for patients exposed to greater than or equal to 3,664 MBq (SIR = 1.80; 95% CI 1.20-2.58). No specific cancer or group of cancers could be convincingly linked to high-dose 131I exposures since SIR did not increase after 10 years of observation. However, upper confidence intervals could not exclude levels of risk that would be predicted based on data from the study of atomic bomb survivors. We conclude that the current practice of extrapolating the effects of high-dose exposures to lower-dose situations is unlikely to seriously underestimate radiation hazards for low LET radiation.