Ninety patients randomly allocated to receive auranofin, matching placebo, or sodium aurothiomalate have been followed up for three years. Inefficacy led to cessation of treatment in 14 patients receiving auranofin, 27 receiving placebo, and one receiving sodium aurothiomalate. Twenty seven of the patients receiving placebo were reallocated within the study and 16 continued therapy at three years. This group showed similar statistically significant improvement in clinical and laboratory parameters at one, two, and three years to those on an active drug from the outset. Patients who discontinued auranofin because of inefficacy were offered sodium aurothiomalate therapy--eight patients in this group completed three years of treatment on sodium aurothiomalate and showed significant improvement in some but not all parameters. A hand radiograph erosion score showed a deterioration in 80% of patients remaining on auranofin, 75% of those on sodium aurothiomalate, and 80% of the original placebo group who continued an active drug for three years. Although more patients discontinued auranofin over the study period because of inefficacy, no difference could be shown between the degree of improvement in the subgroup who remained on auranofin and those receiving sodium aurothiomalate. No disadvantage in outcome could be shown for patients originally assigned to placebo.