Twenty-eight patients with refractory rheumatoid arthritis completed a randomized 24-week double-blind crossover trial comparing oral methotrexate (2.5 to 5 mg every 12 hours for three doses weekly) with placebo. The methotrexate group had significant reductions (P less than 0.01 as compared with the placebo group) in the number of tender or painful joints, the duration of morning stiffness, and disease activity according to physician and patient assessments at the 12-week crossover visit; reductions in the number of swollen joints (P less than 0.05) and 15-m walking time (P less than 0.03) also occurred. These variables, as well as the grip strength and erythrocyte sedimentation rate, showed significant (P less than 0.01) improvement at 24 weeks in the population crossed over to methotrexate. A significantly increased frequency (P less than 0.03) of the HLA-DR2 haplotype occurred in the eight patients with the most substantial response to methotrexate. Adverse reactions during methotrexate therapy included transaminase elevation (21 per cent), nausea (18 per cent), and diarrhea (12 per cent); one patient was withdrawn from the trial because of diarrhea. One patient died while receiving the placebo. Methotrexate did not affect measures of humoral or cellular immunity. We conclude that this trial provides evidence of the short-term efficacy of methotrexate in rheumatoid arthritis, but the mechanism of action is unknown. Longer trials will be required to determine the ultimate safety and effectiveness of this drug.