**4. Discussion**

The current study compares patients following an aTSR versus an rTSR aged 75 years and older. The baseline demographics of the two groups were similar, including PROMs. The overall outcomes in the 2-year follow-up period confirm our original hypothesis, as better ROM, clinical outcome scores including patient satisfaction, and less complications were demonstrated for the aTSR group than the rTSR group, as shown in Table 4.

These results differ to Kiet et al. who looked at outcomes of aTSR and rTSR results after 2 years [24]. They demonstrated no differences in the complication or revision rates, nor ASES and pain scores. The ROM in both groups was similar except for a slightly better external rotation in the aTSR group, which was also found in our study. In addition, in our study, the AE was also significantly better in the aTSR group (*p* < 0.001). It is difficult to compare our results with Kiet's study, as the average age of the patients is not documented, and further, they used a stemmed anatomic humeral prosthesis which was cemented and a metal humeral head [24].

Another study by Wright et al. compared aTSR with rTSR in patients aged 70 years and older with an intact rotator cuff [25]. They were able to identify similar postoperative ROM and outcome scores for both prostheses. Although they reported no significant differences in complication rates for the two prostheses types, they did identify that rotator cuff tears were the principal complication in the aTSR group leading to conversion to an rTSR, which occurred within an average of 28 months postsurgery. In the current study, there was one early cuff failure following an aTSR within 24 months follow-up, but revision of the prosthesis was not required.

An additional problem with an rTSR is that the medical complication rate has been reported to be significantly higher than with an aTSR [26]. The factor of age and complications was analysed by Koh et al. [27]. They assessed patients at 30 days postoperation and demonstrated that the older age group (>80 years) had significantly more complications and readmissions than younger patients for any type of shoulder arthroplasty [27].

Although all patients were 75 years of age and older, the rTSR patients were slightly older and a slightly lower ROM preoperatively, however, the PROMs and CS were very similar. Therefore, both groups started from a similar basepoint for all indices. However, the overall improvement in most indices and scores was greater in the aTSR group than the rTSR group. This resulted in the rTSR group having lower final postoperative clinical scores and a worse ROM than the aTSR group. It was not possible to analyse males' versus females' results due to the relatively few male patients.

Friedman and colleagues demonstrated that older patients with an rTSR had better outcome scores (ASES and SPADI) despite smaller improvement in abduction and forward flexion than younger patients [28]. Triplet et al. were able to demonstrate good improvements in motion (ER and AE), pain and function (ASES) in a smaller cohort of older patients (>80 years) for both types of shoulder replacements (aTSR *n* = 18 and rTSR *n* = 33). However, they did detect more postoperative complications and higher transfusion rates for rTSR patients [29]. This was different to Anakwenze and colleagues, who looked at the effect of age and outcomes in aTSR and rTSR [30]. They identified a higher odds ratio (OR) for patients older than 75 years for readmissions and an increased 1-year mortality in patients with an aTSR (OR 1.75 and 3.34) versus rTSR (OR 0.68 and 0.92). Furthermore, their hazard ratios (HR) for risk of revision were significantly lower for rTSR vs. aTSR patients in patients older than 75 years versus patients 75 years or younger, at HR 0.45 versus 1.24, respectively [30]. In our study there were no revisions. Another study from Wagner et al. demonstrated that no matter what sort of shoulder prosthesis is implanted, the risk of revision for mechanical failure, aseptic loosening or infection decreased with age 65 years and above [31]. According to their data, only instability remained an age-independent complication.

Stress fractures were the main complications seen in our rTSR group (21.6%), while none of the above-mentioned studies reported acromial or scapular stress fractures [25,26,32,33]. However, Zmistowski et al. reported an incidence of 10.5% for acromial stress fractures and reactions following rTSR, with an average follow-up of 407 days [34]. There was also no mention in the previous studies of acromioclavicular joint pain, as was seen in some cases in our study [25,26,32–34].

In our study, component loosening was not an issue in this short-term follow-up period. On radiographic evaluation of the humeral component side, there was no difference between prosthesis types, with only minor osteolysis seen for both the stemless aTSR and the long-stem rTSR. On the glenoid side, osteolysis for both types of component was again only minor, but there was a slight tendency towards more osteolysis with the rTSR than the aTSR, although it is difficult to compare in such different prostheses as notching is only seen with rTSR and radiolucent lines with aTSR.

Shields and Wiater [33] compared aTSRs which were revised to rTSRs because of rotator cuff failure or component loosening to matched patients with primarily an rTSR. They demonstrated that patients with a revision surgery for a failed aTSR not only had a lower satisfaction but also significantly poorer subjective outcome scores and more complications than the primary rTSR group. However, revision of a primary aTSR to an rTSR solely due to cuff failure has also been shown by Flury et al. to be a good salvaging procedure, as it improves ROM, clinical scores and patients' satisfaction when comparing pre- and post-operative scores [32]. The latter study results support a surgeon's decision to continue using an aTSR in older patients, especially as with a stemless prosthesis, revision of the components to an rTSR is relatively easy.

Despite prospective data collection, our study has limitations. The follow-up time is limited to 2 years, which is less than the average time of secondary rotator cuff failure in aTSR [25,35]. Despite this, during this short follow-up time, the rTSR demonstrated more complications, in particular stress fractures of the scapula, than aTSR. Patients receiving an rTSR were slightly older than the aTSR patients, skewing the data slightly, as there is a naturally increased chance with age of a later rotator cuff tear, which requires an rTSR rather than an aTSR [36]. Another limiting factor is the relatively small cohort size and the predominance of female patients, both of which are probably related to the increased age of the patients. Further research and longer follow-up studies need to be performed in order to address these issues.
