Background: Several clinical studies have focused on the therapeutic effects of interferon gamma (IFN-γ) in patients with severe atopic dermatitis (AD), although the dosage of recombinant IFN-γ (rIFN-γ), therapeutic schedule, and the degree of clinical improvement were different among studies. Although the exact mechanism of action of IFN-γ therapy in AD is not clear, the beneficial effects of IFN-γ have been attributed mainly to an immunomodulating effect on the expression of certain immunologic markers. Objective: Our purpose was to study the therapeutic effect of two different dosages of rIFN-γ on AD and to investigate the change of lesional expression of infiltrating inflammatory cell markers associated with rIFN-γ therapeutic efficacy. Methods: Fifty-one patients with severe recalcitrant AD were treated with rIFN-γ. Twenty patients were treated with 0.5 × 10⁶ IU/m² of rIFN-γ (low-dose [LD] group); 21 patients received 1.5 × 10⁶ IU/m² of rIFN-γ (high-dose [HD] group); and 10 patients received placebo. The patients were injected subcutaneously 3 times a week for 12 weeks. Immunohistochemical study was performed in 20 patients of the HD group in the initial visit and after completion of rIFN-γ therapy with a panel of 14 monoclonal antibodies as markers of inflammatory cells and cytokines. Results: The disease severity of the 2 groups treated with rIFN-γ was reduced significantly at the end of treatment compared with that of the placebo group (P<.05). More rapid clinical improvement and more effective treatment outcome were seen in the HD group than in the LD group for the initial 6-week treatment period; however, the clinical improvement in both of the treated groups was stable and maintained after week 8 of treatment. Immunohistochemical findings showed statistically significant reduction in the lesional expression of CD25 and EG2 cells that infiltrated into skin after rIFN-γ therapy. Conclusion: This study demonstrated that rIFN-γ therapy for AD is safe and effective. In the early phase of therapy, a higher dosage of rIFN-γ is more effective; and for the maintenance of clinical improvement, a lower dosage of rIFN-γ is recommended when high cost and effectiveness of rIFN-γ are considered. The therapeutic efficacy of rIFN-γ in AD might be in part related to the decreased number of CD25⁺ and EG2⁺ inflammatory cells infiltrated into skin. (J Am Acad Dermatol 2000;42:1033–40.)