We have performed a clinical and radiological analysis of 105 shoulder arthroplasties in patients with rheumatoid arthritis. The clinical results showed improvements in the Constant-Murley and Association of Shoulder and Elbow Surgeons score of 21 and 35, respectively. Both were statistically significant (p < 0.001). This improvement was maintained over a period of 8.8 years. There was no statistically significant difference in the scores after hemiarthroplasty and those after total arthroplasty. The presence of an intact rotator cuff was associated with improved function in both groups. In spite of the use of an uncemented humeral stem, no implant was radiologically loose or at risk. There was lucency in a single zone in 14 implants. One glenoid component was at risk and 16 had lucency in a single zone. There was, however, a significant difference in the amount of lucency which was associated with pegged and keeled glenoid components (p = 0.005). In the group with hemiarthroplasty, two or more years after surgery there was superior migration of the humeral component by more than 5 mm in 18 shoulders (28%) and medial migration by more than 2 mm in eight (16%). Both superior and medial migration had an effect on the outcome. Revision was undertaken in four patients for persistent pain relating to medial migration. With revision taken as the endpoint for survival after eight years, 92% were found to be still in situ