Uncontrolled study has demonstrated the usefulness of somatostatin in the treatment of mild Graves' ophthalmopathy (GO). We performed a prospective study to evaluate the usefulness of somatostatin as compared to corticosteroid in the treatment of moderately severe GO. All patients were rendered euthyroid and observed for 3 months to exclude spontaneous improvement without active treatment. They were randomized to receive either somatostatin (SS, octreotide 200 micrograms q8h subcutaneously, n = 8) or corticosteroid (CS, prednisone 1 mg/kg/day in decreasing doses, n = 10). Assessments of soft tissue inflammation, exophthalmos, palpebral aperture, intraocular pressure, diplopia, cornea, and visual acuity were made every 4 weeks for 3 months. MRI of the orbit was performed before and after treatment. Both SS and CS therapy decreased the palpebral aperture and activity score after 3 months (p < 0.05), but those treated with CS had a lower activity score after treatment when compared to SS [2.5 (1-7) v.s. 3.5 (0-4), median (range), p < 0.05]. Only CS, but not SS, was able to reduce intraocular pressure and muscle size as documented by MRI, but no significant reduction in proptosis was observed in either group. Also, patients' self-assessments of the eye changes after treatment were similar between the two groups. Both groups showed significant elevation of urinary glycosaminoglycan (GAG) excretion before therapy (SS 24.6 +/- 10.8; CS 27.8 +/- 11.4 mg/24 h), which was reduced after treatment (SS 12.5 +/- 7.3; CS 10.8 +/- 6.3 mg/24 h, p < 0.05). However, no significant correlation could be observed between the degree of GAG reduction and the clinical outcome of the patients. In conclusion, the long acting SS octreotide was effective in reducing soft tissue inflammation and providing symptomatic relief in GO but not as effective as corticosteroid in reducing muscle size. In view of the minimal side-effects and similar efficacy as compared to corticosteroid in patients with minimal extraocular muscle enlargement, it is suggested that a trial of SS may be considered in selected patients with GO.