The aim of this study was to perform a randomized study to evaluate the role of plasma filtration in the treatment of severe thyroid-associated ophthalmopathy (TAO). 20 patients were enrolled, and all patients were treated with methylprednisolone IV pulses. 10 randomly chosen patients were also subjected to plasma filtration (twice weekly in Weeks 1, 2, 4, 7, and 10). The procedure proved to be safe. All immunoglobulin classes as well as autoantibodies directed against thyroglobulin, thyroid peroxidase, and TSH receptor exhibited statistically significantly decreases. Some markers of cell-mediated immunity such as soluble antigen CD30 and monocyte chemotactic protein 1 decreased, but serum levels of other markers such as CD40 ligand and soluble protein Fas/Apo-1 did not change significantly. The decrease of immunoglobulins was long lasting, whereas cytokine levels returned to basal values before the next apheresis. Although the clinical activity score (CAS) dropped in all patients, it occurred more rapidly in patients treated with plasma filtration. The CAS difference between the two groups was statistically significant (p = 0.027). The amplitude of visual evoked potentials improved after 3 months in the plasma filtration group. At the end of the study, there was no difference between patients treated with aphereses and the control group. Eye muscle width and proptosis measured by CT scan did not differ between the two groups. We conclude that apheresis can decrease disease activity more rapidly than standard high-dose IV glucocorticoid therapy. Whether this superior treatment effect could potentially avoid surgical procedures remains to be determined.