Aim. To identify and quantify the reasons for failure of primary shoulder arthroplasty.

Materials. Seventy-nine Revision Shoulder Arthroplasties (RSA) in 75 patients were performed between January 1990 and July 2003. Twelve patients underwent more than one revision. Five patients were lost to follow-up. The indications for PSA (38 total shoulder arthroplasties, 37 hemiarthroplasties) were trauma (22), osteoarthritis (13), rheumatoid arthritis (11), avascular necrosis (9), cuff tear arthropathy (9), traumatic arthritis (5), capsulorraphy arthropathy (4),and instability (2). Glenoid and humeral bone stock, and rotator cuff integrity were analysed and the reasons for failure of the index procedure were determined.

Results. The median interval between PSA and RSA was 46 months (1 month to 22.8 years; mean 28 months). Thirty-six (58.1%) PSAs (of which 26 of 37 hemiarthroplasties (70%)) failed in the first three years. Failure of the PSA was caused by rotator cuff deficiency (24), glenoid bone erosion (19), glenoid component loosening (25), humeral bone erosion (3), infection of the implant (3) and periprosthetic fracture (1). Revision of a loose humeral component was performed in 6 cases, but 31 humeral components were revised in order to deal with glenoid or rotator cuff conditions.

Conclusion. When faced with a limited surgical armamentarium or tactic, there is a predictable rate of failure of shoulder replacement involving the interaction of the failed or failing rotator cuff and the evolution of glenoid deficiency.