The long-term survival rate of shoulder arthroplasties is comparable to that of hip arthroplasties, at about 85% after 15 years. The diagnosis of rheumatoid arthritis is the most important predictive factor for the end-result, with a stronger influence than the condition of the rotator cuff or whether a hemiarthroplasty or a total arthroplasty is performed. Unlike the situation in rheumatoid arthritis, in osteoarthritis wear and tear on the glenoid is caused mainly by medial erosion with cranial migration. The functional results are slightly better with total shoulder replacements than with hemiarthroplasties. Arthroplasty in rheumatoid patients is adversely influenced by the poor quality of the rotator cuff, which is either primarily defective or develops secondary insufficiency as a result of the underlying illness. Secondary cranial migration, i.e. secondary rotator cuff failure, occurs in up to half these patients in the long term. It is frequently followed by glenoid loosening after total shoulder arthroplasty. Nevertheless, glenoid loosening is frequently not progressive over long periods and is well tolerated by the majority of patients. Thus, the rate of glenoid revision operations in rheumatoid patients seldom exceeds 5% in mid- and long-term studies. Owing to poor bone quality, glenoid erosion after hemiarthroplasty is more frequent than in patients with osteoarthritis. In contrast to glenoid loosening, glenoid erosion is almost always painful and leads to poor results requiring surgical revision. In shoulder joints that are centred in the coronal plane and in which the rotator cuff is still intact or only moderately worn, with no more than one defective tendon, total arthroplasty should be given preference. The results achieved with humeral head surface replacement are as good as those yielded by conventional hemiarthroplasty; combination with the implantation of a glenoid is technically demanding but leads to even better results. Reverse arthroplasties should be reserved for patients over 70 years of age with complete or almost complete rotator cuff destruction. In younger patients bipolar implants should be used in preference; the functional results attained with these are admittedly limited, but unlike the reverse implants they do not involve the risk of loosening of the metaglenoid.