OBJECTIVE Prednisolone reduced the progression of joint destruction over 2 yr in early, active rheumatoid arthritis. The response to discontinuation of prednisolone under double-blind conditions is now reported. METHODS A randomized, double-blind, placebo-controlled trial of prednisolone 7.5 mg daily in addition to routine medication over 2 yr in 128 patients with early rheumatoid arthritis, using radiological progression (changes in the Larsen score) and the development of erosions as primary outcome measures. Study medication was blindly discontinued and follow-up maintained for a further year. Other assessments included disability, joint inflammation, pain and the acute-phase response. RESULTS Similar results were obtained when all available radiographs were included for each year of assessment (maximum 114) and when only patients with radiographs at all time points were included (75 patients). In these 75, the mean progression in the prednisolone group was 0.21 Larsen units in year 1, 0.04 units in year 2 and 1.01 units in year 3 (P = 0.587, 0.913 and 0.039 for change within each year, respectively). The equivalent placebo group means were 2.34, 1.00 and 1.63 Larsen units (P = 0.001, 0.111 and 0.012; difference between groups: 2.13, 0.96 and 0.67 units, P = 0.082, 0.02 and 0.622). The percentage of hands which had erosions at each time point was: prednisolone group: 27.8, 29.2, 34.7 and 39.2; placebo group: 28.2, 48.7, 59.0 and 66.5. There was little evidence for a flare in clinical symptoms after discontinuation of prednisolone. CONCLUSIONS Joint destruction resumed after discontinuation of prednisolone. This corroborates the previously reported therapeutic effect and challenges current concepts of disease pathogenesis.