The development of painful glenoid arthrosis is the most common reason for reoperation after replacement of the humeral head. We performed twenty-two revision total shoulder arthroplasties, between 1983 and 1992, for the treatment of painful glenoid arthrosis in shoulders that had a prosthetic replacement of the humeral head. Eighteen shoulders (seventeen patients) were included in the study as their preoperative and operative records were complete and they had been followed for at least two years (mean, 5.5 years; range, 2.3 to 10.0 years). The indications for the hemiarthroplasty were trauma (ten shoulders), osteoarthrosis (four), rheumatoid arthritis (two), and osteonecrosis secondary to the use of steroids (two). The mean interval between the hemiarthroplasty and the total shoulder replacement was 4.4 years (range, 0.8 to 12.7 years). The mean score for pain in the shoulder decreased from 4.3 points before the revision to 2.2 points after it (p = 0.0001). The mean active abduction increased from 94 degrees before the revision to 124 degrees after it (p = 0.01), and the mean external rotation increased from 32 to 58 degrees (p = 0.007). Two shoulders needed another operation after the revision because of a late infection in one and particulate synovitis associated with instability in the other. With the numbers available for study, we did not detect a significant difference in pain relief and range of motion with respect to gender, diagnosis, subluxation, or the presence of periprosthetic radiolucency.Our findings indicate that most patients with painful glenoid arthrosis after a hemiarthroplasty have marked pain relief and improvement in motion after revision to a total shoulder replacement. However, seven of the eighteen shoulders that had this procedure had an unsatisfactory result due to a limited range of motion or the need for a subsequent operation. Therefore, long-term studies are necessary to evaluate the durability of total shoulder replacement in this group of patients.