A prospective study was carried out to determine the factors which influence response to antithyroid drug treatment in Graves' disease and to assess their predictive value. Eleven variables were included in the assessment and were subjected to discriminant analysis, log rank test and "survival" analysis. The patients were observed for a considerable period (mean duration 51 months). Carbimazole (mean total dose 8 g) was given in combination with thyroxine for an average of eight months to 72 patients. Thirty-five patients relapsed and 37 remain in remission. Thyrotrophin binding inhibiting immunoglobulins (TBII) were detectable in 74 per cent of patients at diagnosis and thyroid stimulating antibodies detectable in 70 per cent. At the end of treatment thyrotrophin binding inhibiting immunoglobulins and thyroid stimulating antibodies were present in 36 and 27 per cent of patients respectively. Levels of thyrotrophin binding inhibiting immunoglobulins were significantly higher both before and after treatment in the group who relapsed, but were not of prognostic significance in an individual patient unless the value was extremely high (TBII index greater than 70). The presence of thyroid stimulating antibodies was of no value in predicting outcome. HLA typing confirmed the known association of Graves' disease with HLA B8 and HLA DR3 but neither of these antigens conferred and increased likelihood of relapse. The likelihood of relapse is shown to be directly related to the severity of the disease at the time of diagnosis, as measured by the serum total T3, and to the size of the thyroid gland; it is not affected by age, family history of thyroid disease or ophthalmopathy. The data indicate that antithyroid drug treatment can be expected to induce long-term remission in patients with mild disease (T3 less than 5 nmol/l) and small thyroids; carbimazole at this dose level is inappropriate for patients with severe disease (T3 greater than 9 nmol/) and large goitres.