Purpose: The purpose of this study was to analyse total shoulder arthroplasty failures and the outcome after simple ablation of the prosthesis or revision with a Grammont inverted prosthesis.

Material and methods: This was a longitudianal prospective study of patients with a failed shoulder prosthesis who underwent either simple ablation of the prosthesis or revision athroplasty with an inverted Grammond prosthesis. Clinical (Constant score) and radiographic analysis was performed before revision and at last follow-up using identical criteria. The shoulder prosthesis was removed in case of failure due to infection. For all other causes of failure, an inverted Grammond prosthesis was implanted.

Results: The series included nine patients (eight women and one man) reoperated between January 1st 1995 and December 31st 1999. Mean follow-up was 47 months (12–108). Delay between the first procedure and revision surgery was 26 months (6–72). The cause of the failure of the first implant was: infection (four patients), dislocation (one patient), three-tendon rotator cuff tears (four cases). The overall Constant score before revision surgery was 18.5 (6–30). Mean Constant score at last follow-up was 40.1 (35–54). Mean gain in pain score was 9.4 points (0–15) and mean gain in hand position was 2.75 points (0–10).

Discussion: Complications after shoulder arthroplasty are not uncommon (14% according to Wirth, 1994) and treatment is difficult (Sperling 1999). Instability, rotator cuff tears, glenoid loosening, and infection are the most frequent causes of failure (Wirth 1994). The patients in this series had a very poor Constant score involving all the subscores, although deterioration of the pain score predominated. The gain, both with ablation and revision total shoulder arthroplasty, was greater than 25 points on the average. This gain was proportional to the initial score before revision and patients who had a revision total shoulder arthroplasty had a better gain (p < 0.001). Simple implant ablation did however improve the mean Constant score among patients with infection whose initial score was lower than the others (p < 0.001). the final outcome was moderate. The only patients who recovered muscle force were those who had a total revision prosthesis (p < 0.05).

Conclusion: Revision of a shoulder prosthesis gives disappointing results in terms of absolute outcome score, even though the gain over the initial functional situation is encouraging. Simple ablation of an implant is still indicated in certain patients, in particular those who have an initial Constant score under 20 points.