Although diet therapy for arthritis has received considerable publicity, there is little objective information about its efficacy. We undertook a 10-week, controlled, double-blind, randomized trial of patients with active rheumatoid arthritis (RA). Twenty-six patients completed the study; 11 were on an experimental diet (a specific popular diet free of additives, preservatives, fruit, red meat, herbs, and dairy products) and 15 were on a "placebo" diet. Of 183 variables analyzed, there were no clinically important differences among rheumatologic, laboratory, immunologic, radiologic, or nutritional findings between patients on experimental and placebo diets. Six RA patients on the placebo and 5 on the experimental diet improved by objective criteria. Improvement averaged 29% for patients on placebo and 32% for patients on experimental diets. Two patients on the experimental diet improved notably, elected to remain on the experimental diet following the study period, have continued to improve, and noted exacerbations of disease upon consuming nonexperimental diet foods. Our study failed to provide evidence of objective overall clinical benefit of this diet as followed by a group of patients with longstanding, progressive, active RA. However, our data are not inconsistent with the possibility that individualized dietary manipulations might be beneficial for selected patients with rheumatic disease.