Tumour necrosis factor alpha (TNF alpha) is a critical inflammatory mediator in rheumatoid arthritis, and may therefore be a useful target for specific immunotherapy. In support of this hypothesis, we previously observed beneficial responses in patients with active rheumatoid arthritis after open-label administration of a chimeric monoclonal antibody to TNF alpha (cA2). We now report the results of a four-centre, randomised double-blind trial of a single infusion of 1 or 10 mg/kg cA2 compared with placebo in 73 patients with active rheumatoid arthritis. The primary endpoint of the study was the achievement at week 4 of a Paulus 20% response, an amalgam of six clinical, observational, and laboratory variables. Intention-to-treat analysis of data from individual patients showed only 2 of 24 placebo recipients responding at this time, compared with 11 of 25 patients treated with low-dose cA2 (p = 0.0083) and 19 of 24 patients treated with high-dose cA2 (p < 0.0001). Over half of the high-dose cA2 patients responded by the more stringent 50% Paulus criteria at this time (p = 0.0005). The magnitude of these responses was impressive, with maximum mean improvements in individual disease-activity assessments, such as tender or swollen-joint counts and in serum C-reactive protein, exceeding 60% for patients on high-dose treatment. There were two severe adverse events. 1 patient on 1 mg/kg cA2 developed pneumonia ("possibly" treatment-related) and 1 on 10 mg/kg had a fracture ("probably not" treatment-related). The results provide the first good evidence that specific cytokine blockade can be effective in human inflammatory disease and define a new direction for the treatment of rheumatoid arthritis.