**4. Discussion**

Despite the abundance of available literature focusing on two-stage exchange arthroplasty in patients with shoulder PJI, there is a widespread heterogeneity among most studies. The most important fact is the different definition criteria of the shoulder PJI diagnosis, as well as of its treatment success, leading to a lack of comparability. With the expected increase of performed shoulder arthroplasties over the next years, a consensus diagnostic definition and a consensus definition of treatment success for shoulder PJI are getting increasingly relevant and important not only to create a more comparable scientific reporting, but also to diagnose, counsel, and treat patients in a standardized matter.

There is considerable variation in reimplantation rates among studies in literature, ranging from 37% to 97% [7,15–21]. Only a few studies have dealt with these patients and tried to report on their clinical outcomes and causes for their attrition [7,18,22]. The current study aimed at evaluating all patients who underwent an attempted two-stage exchange arthroplasty for shoulder PJI, irrespective of the subsequent clinical course, and demonstrated that almost one-third of all patients who underwent the first stage

of the procedure, did not complete a subsequent reimplantation. Despite having an acceptable infection eradication rate in patients with a permanent spacer, the clinical outcome scores are poor and progressive bone loss can occur with the extended retention of the spacer [18,22]. The current accepted goal of a two-stage exchange arthroplasty protocol is still the reimplantation of a new prothesis to ensure best functional outcome for patients and every effort should be made to improve shoulder function by providing the opportunity for reimplantation.

A variety of reasons can lead to failure to reimplant, including mortality, medical comorbidities, uncontrollable infection leading to amputation of the limb or lifetime antibiotic suppression, and unwillingness of patients to undergo a second surgery, as well as patient's satisfaction with the current status. Similar to the hip and knee literature [23,24], premature mortality and high general morbidity were the most common causes for not being able to proceed with the intended reimplantation in this study. CCI and older age seemed to be risk factors for failure to reimplant, as the patients in the group without reimplantation had significantly higher CCI and were significantly older compared to patients with reimplantation. Patients with a higher CCI mostly have poorer health as well as a compromised immune status, which can be due to scar tissue and vascular damage. This local and systemic immune failure can massively decrease the minimal infecting dose of bacteria, predisposing to problems with infection eradication [25].

Furthermore, several studies in hip and knee literature were able to show an association of the microorganism type and likelihood of inability to achieve a reimplantation. To our knowledge, there are no existing studies investigating this association in patients with shoulder PJI. Barton et al. found that patients with a polymicrobial infection had a nearly 8-fold greater likelihood of inability to undergo reimplantation compared to patients with monomicrobial or culture-negative infections [23]. Although it did not achieve statistical significance, we were able to identify more polymicrobial infections in patients who did not undergo reimplantation, compared to patients with a successful reimplantation.

The treatment success rate in terms of infection eradication was as high as 86% among patients with reimplantation. This is comparable to the almost 90% infection control rate reported in a recent systematic review and meta-analysis of 30 studies reporting on two-stage exchange arthroplasty in patients with shoulder PJI [6]. However, when taking the patients without reimplantation into consideration, the overall infection eradication rate dropped to 78%, which is lower than most of the studies in the literature dealing with two-stage exchange arthroplasty. The majority of these studies do not encompass patients who do not complete the second stage of an attempted two-stage exchange arthroplasty and exclude them from their treatment success analysis, thereby leading to a possibly overestimated success of this surgical procedure. In addition, the considerably high inflammatory response in many patients of the study cohort may be a further factor affecting our infection eradication rate, as patients with persistent infection had a significantly higher C-reactive protein on admission compared to patients with successful eradication of infection. High virulent microorganisms mostly induce an acute response and much inflammation with the release of cytokines and elevation of CRP, leading to a potentially worse postoperative clinical course [26,27].

Although mortality and morbidity associated with two-stage exchange arthroplasty for hip and knee periprosthetic joint infection has been one of the main research topics [8,23,24], there is a lack of knowledge in shoulder PJI literature. Cancienne et al. recently showed a mortality rate of 2.2% within the first postoperative year in patients undergoing removal of an infected shoulder prosthesis [7]. This is significantly less than the mortality rate reported in the current study, which was 10% within ninety days after resection arthroplasty and 25% at the latest follow-up, which is similar to what has been reported previously in hip and knee literature [23,24,28]. The mortality in five patients, who deceased in the first 90 days after resection arthroplasty, was related to the infection, whereas the other patients died due to other health issues. This indicated that shoulder PJI and its treatment is associated with a high risk of mortality, especially in patients with

older age and higher CCI, as shown in our study. In these patients, it may be reasonable to have a detailed discussion with the patient about their likelihood of treatment success, as well as postoperative mortality and consider alternative treatment options, such as long-term antibiotic suppression. Thus, an optimal treatment algorithm should be based on patient-specific general health status, risk factors, and patient expectations.

Only looking from the perspective of infection eradication would lead to overlook a grea<sup>t</sup> subset of patients who are too fragile to endure further surgery for reimplantation, decease prematurely, or refuse further surgery because of low functional demand after implant removal. We therefore sugges<sup>t</sup> that the success of two-stage exchange arthroplasty should be accounted from the point of the first stage, rather than following reimplantation, to consider the failures occurring between the stages and better represent the clinical course of two-stage exchange arthroplasty in patients with shoulder PJI [8].

This study has some limitations. Although the patients' data were longitudinally collected in our database, the retrospective nature of the study may lead to bias. Despite being the largest cohort in literature dealing with this topic, the study may be underpowered, preventing the significant differences between analyzed groups. Furthermore, we did not include any assessment of functional outcomes, which may be seen as a potential weakness. However, the most compelling outcome measures of the current study were the infection eradication, reimplantation, and mortality rates. In addition, there is already an abundance of available literature focusing on the functional outcome of patients after two-stage exchange arthroplasty or antibiotic cement spacer retention. The complexity of our study cohort, due to multiple previous revision surgeries, can furthermore alter our results, making our results to be generalized and difficult to compre with other studies. Finally, infection-related mortality was difficult to confirm in patients who died outside of the hospital and the precise cause of death in these cases could not be determined, which can alter our treatment success rate.
