The clinical outcome of 1028 Hong Kong Chinese patients with Graves' disease treated with radioiodine therapy and followed for a mean of 9.85 +/- 4.84 years (range 2-20) was analysed. Retreatment was required by 413 patients (40.2 per cent), with 134 patients (13.0 per cent) requiring more than two 131I doses. One hundred and eighty-nine patients received carbimazole after 131I until euthyroidism was achieved. The cumulative incidence of hypothyroidism at one, five, 10 and 15 years was 9.6 per cent, 31.4 per cent, 53.8 per cent and 65.8 per cent, respectively. The average incidence of hypothyroidism after the first two years was 3.3 per cent per annum. Stepwise logistic regression analysis of pretreatment variables suggested that a combination of adjunctive carbimazole therapy, absence of ophthalmopathy and longer effective half-lives of 131I were of value in predicting which patients were less likely to develop permanent hypothyroidism. However, the probability of accurately predicting permanent hypothyroidism based on the present model was only 60 per cent. We believe that no single pretreatment variable, or combination of variables, predicts long-term hypothyroidism with sufficient confidence to justify the use of a 'formula' approach for prescribing 131I therapy for Graves' disease.