BackgroundThe optimal design of an elbow prosthesis for badly damaged elbows is unkown. We evaluated 23 GSB III semi-constrained (sloppy-hinged) total elbow arthroplasties in 16 consecutive patients with rheumatoid arthritis.Patients and methods After a mean follow-up period of 5 (2-9) years, we assessed quality of the cementing technique, signs of aseptic loosening, patient satisfaction, range of movement, and determined the Hospital for Special Surgery (HSS) elbow score. 3 patients had died before follow-up; thus, 20 replacements in 16 patients were available for clinical and radiographic study. All patients had end-stage rheumatoid arthritis (RA) of the elbow joint.Results In 2 patients, humeral components were revised due to malorientation. 1 arthroplasty was revised due to aseptic loosening of the humeral component. There were 4 cases of intraoperative fracture which healed uneventfully. The total rate of complications was thus one-third. In 17 of 40 components, the cementing technique was rated as marginal or inadequate. We found no association between cementing technique and loosening. The arc of extension/flexion increased by 19º (0-80), and the range of pronation/supination increased by 31º (0-130). There were no cases of infection or ulnar nerve dysfunction. At the latest follow-up, the HSS elbow score was 84 (40-100) points. 11 of 20 elbows were rated as excellent, 4 elbows were rated as good, 2 elbows were rated as fair, and 3 elbows were rated as poor. 14 of 16 patients were satisfied with the result and the 2 patients who were not satisfied had persistent pain.Interpretation Despite the inherent problems of cementing in small-calibre medullary cavities, the clini-