Twenty-six patients with severe rheumatoid arthritis who had completed a randomized crossover trial of methotrexate elected to continue to receive the drug in a long-term prospective study. At 36 months, 16 patients remained in the study. Over this period of time, significant improvement was noted in the number of painful and swollen joints, physician and patient global assessments, erythrocyte sedimentation rate, and prednisone dose. Adverse reactions occurred in 16 patients (62%), including nausea, alopecia, headache, stomatitis, herpes zoster, and diarrhea. Mild leukopenia (3 patients), thrombocytopenia (3 patients), and elevated transaminase levels (8 patients) resolved with temporary drug discontinuation. No patient withdrew due to drug toxicity. Liver biopsy specimens in 17 patients after 24 months of treatment showed no evidence of fibrosis or cirrhosis. A significant increase in the percentage of T3 and T4 blood cells and increases in lymphocyte proliferation to concanavalin A and purified protein derivative of tuberculin were found after 2 years of therapy. Our findings indicate that methotrexate has remained effective over 36 months of therapy, with acceptable toxicity levels and no evidence of systemic immunosuppression.