OBJECTIVE To study whether the reported superior effect of methotrexate (MTX) compared to azathioprine (AZA) in retarding radiologic progression after one year in rheumatoid arthritis was sustained at 2 and 4 years. METHODS All 64 patients enrolled in the original randomized double blind study were invited for an open extension of followup to 4 years including 4-monthly clinical and laboratory assessments and radiographs of hands, wrists, and feet at 2 and 4 years. RESULTS After 4 years, 18 patients (58%) from the MTX group and 7 patients (21%) from the AZA group continued the initial study drug. During followup more patients (n = 21) switched from AZA to MTX than vice versa (n = 5). In an intention-to-treat analysis improvement of clinical and laboratory variables at 4 years was more pronounced in the MTX group. Mean radiologic scores increased in both treatment groups during followup. According to an intention-to-treat analysis increase in erosion score at one and 2 years in the MTX group was significantly lower than in the AZA group: after one year MTX group 1.8 versus AZA group 5.3 (p = 0.002); after 2 years MTX 3.5 versus AZA 6.5 (p = 0.05). After 4 years there was a trend toward less progression in the MTX group: MTX 6.8 versus AZA 10.8 (p = 0.09). For the total score, progression in the MTX group was less after one and 2 years. After 4 years marked radiologic progression was observed more often in the AZA group. CONCLUSIONS Drug continuation after 4 years of followup was better for MTX than for AZA. In an intention-to-treat analysis the beneficial effect of MTX on radiologic progression compared with AZA was sustained after 2 years of followup. Thereafter differences between treatment groups leveled off, probably mainly due to the greater number of switches from AZA to MTX than vice versa.