**4. Discussion**

In this study, we evaluated the conversion of failed HA to RSA using a modular prosthetic system and assessed the clinical outcomes at different time intervals postoperatively. We reported significant postoperative improvements in the Constant score, ASES Shoulder Score, VAS for pain and satisfaction, and ROM.

During the revision of failed HA, the correct balancing of soft tissue tension is important. Insufficient tension can result in instability of the prosthesis, whereas excessive tension can result in acromial fracture, poor deltoid function, and/or neurological lesions [19]. Moreover, restoring adequate tension can be complicated by scarring and altered muscle function. However, most current modular prostheses allow the height of the humeral head to be adjusted enough to modify the compressive forces across the glenohumeral joint, and thereby achieve adequate soft tissue tension.

Postoperative arm lengthening is a typical occurrence after RSA. Lädermann et al. reported a mean postoperative lengthening of 23 ± 12 mm in primary and revision RSA using a different prosthesis than we did [19], while Teschner et al. measured a mean increase in implant height of 24 ± 2.6 mm in the conversion of failed HA to RSA using the same prosthesis as we did [14]. Both of these values are similar to the lengthening that can be achieved with the reverse modular prosthesis used in our series (23 mm) (Figure 3). Moreover, we did not observe any complications related to the under- or overtensioning of the deltoid tissue. We therefore believe that the modular prosthesis used in our series allowed for the optimal soft tissue balance required for retaining a firmly implanted humeral stem.

Although several modular convertible shoulder prostheses are available today for the successful conversion of HA to RSA [14], stem retention may not always be possible, especially when the humeral stem is poorly positioned or not well fixed [10,11]. In these cases, stem revision is needed [10,11]. Similarly, conversion RSA is not feasible in cases with infection or where the stem cannot be distalized enough, as seen in the present study. Finally, the use of stemless anatomical implants in primary procedures does not allow conversion RSA. We also found that not every patient in the present study could benefit

from conversion RSA. Of the 17 cases where conversion RSA was considered, four cases (23%) could not be converted due to overstuffing. A similar rate was found in a recent study where 22% of the patients could not be converted despite the presence of a modular prosthesis [10], indicating a limitation of today's convertible modular prosthetic systems. Other limitations include the need for highly skilled surgeons and a meticulous surgical technique for implanting a convertible modular prosthetic system during the primary procedure [16]. These limitations necessitate the continued use of modular non-convertible shoulder prostheses.

Revision surgery of failed HA remains technically challenging and may be associated with less predictable clinical outcomes and various complications [18]. Nevertheless, in the current study, we found a significant improvement in the mean Constant score (from 21.7 preoperatively to 57.9 at the last follow-up; *p* = 0.0001). Other studies on the revision of failed HA to RSA also found similar increases in the mean Constant scores: from 12.67 to 45.08 [6] and from 8.9 to 41.0 [18].

Besides the functional scores, intraoperative parameters such as the operating time are also important contributors to successful conversion RSA. In recent studies comparing conversions and revisions, conversion procedures resulted in shorter operating times [10,11,13,15]. We found similar results in our study: the mean operating time for conversion RSA was shorter than that for traditional revision RSA [118.4 min (range: 80.0 to 140.0) versus 150.0 min (range: 100.0 to 230.0), our unpublished results].

Despite successful clinical outcomes, complication rates remain high after conversion RSA, ranging from 22% to 43% [11]. We found complications in three (23%) patients who underwent conversion RSA. Revision with stem exchange was needed in two (15%) patients due to aseptic loosening. Although we observed an overall complication rate similar to that reported in other studies of RSA after failed shoulder arthroplasty [6,20,21], more clinical evidence is needed to draw definite conclusions in this regard. Previous studies also showed that modular prosthesis designs minimized the risk of periprosthetic humeral shaft fractures [18]. We confirmed this finding; no periprosthetic fractures or other implant-related complications were observed in our study.

Our study had a number of limitations; these included the absence of standardized radiographs for the accurate measuring of arm length and the small sample size owing to the nature of the indication, which limited the interpretation of our findings. This is why—although challenging—studies with larger patient cohorts are needed to investigate the full potential of modular systems in the conversion of failed HA to RSA. Finally, the results from the type of modular convertible shoulder prosthesis used in our study are not sufficient to make generalized conclusions on the outcomes of conversion RSA. For this, a clinical comparison with different implant types will be needed.
