Clinical Outcome of Two-Stage Revision after Periprosthetic Shoulder Infection

Background: Periprosthetic shoulder infections are devastating complications after shoulder arthroplasty. A potential treatment concept is a two-stage prosthesis exchange. Data are sparse in terms of clinical outcome, including infection-free survival and patient satisfaction after this procedure. In the present study, we investigated recurrence of infection, revision-free survivorship and clinical outcome following two-stage revision due to periprosthetic shoulder infection. Furthermore, reasons for poor outcome were analyzed. Methods: Sixteen patients undergoing two-stage revision after shoulder joint infection were retrospectively identified. Recurrence of infection was analyzed by Kaplan–Meier survival curve. Clinical outcome was quantified with subjective shoulder value (SSV), “quick” Disabilities of the Arm, Shoulder and Hand (qDASH) and Rowe score. Range of motion (ROM) was measured pre- and postoperatively. Postoperative scores and ROM were compared in a subgroup analysis according to different reimplanted prosthesis types. Results: The reinfection-free implant survival was 81% after one year and at final follow-up (FU; mean of 33.2 months). The overall revision-free survival amounted to 56% after one year and at final FU. Patients who received reverse shoulder arthroplasty (RSA) as part of reimplantation had less disability and long-term complications. This group demonstrated better subjective stability and function compared to patients revised to megaprostheses or large-head hemiarthroplasties. Conclusions: Two-stage revision following periprosthetic joint infection of the shoulder allows appropriate infection control in the majority of patients. However, the overall complications and revision rates due to mechanical failure or reinfection are high. Reimplantation of RSA seem superior to alternative prosthesis models in terms of function and patient satisfaction. Therefore, bone-saving surgery and reconstruction of the glenoid may increase the likelihood of reimplantation of RSA and potentially improve outcome in the case of infection-related two-stage revision of the shoulder.


Introduction
Primary shoulder arthroplasty (SA) is an expanding field in treating proximal humeral fractures and degenerative shoulder disorders, including osteoarthritis and cuff tear arthropathy (CTA) [1]. In 2012, 34 glenohumeral joint replacements per 10 5 inhabitants were performed in Germany, which means a tripling of the annual incidence in the past decade [2]. With an increasing number of shoulder replacement surgeries, orthopedic surgeons are increasingly confronted with complications like implant failure [3]. Apart from mechanical disorders, periprosthetic joint infections (PJIs) are serious complications associated with prolonged hospitalization and long-term functional constraints [4]. Several influencing risk factors, such as bone and soft tissue status, comorbidities, socioeconomic status, age and gender, that may predispose for PJI are discussed [5]. However, male gender and young age at primary implantation were identified as the most important influencing factors [6,7]. It is reported that the incidence of PJI following primary SA ranges from 0 to 4% [8,9]. Distinct guidelines for the clinical management are missing and appropriate treatment concepts are heterogeneous [10][11][12][13]. Treatment ranges from exclusive antibiotic treatment to surgical revision with irrigation and debridement or staged prosthesis exchange up to definitive resection arthroplasty [14][15][16], whereas most studies recommend removal of the implant for chronic infections [8,17,18]. Infection control can also be achieved by resection arthroplasty [13,[19][20][21] or permanent spacer retention [22]. However, both procedures have a poor functional outcome with a high rate of dissatisfied patients [13,19,22]. Therefore, the most established procedures are the one-and two-stage prosthesis exchange utilizing similar approaches as for hip and knee PJI [22][23][24][25].
A clear superiority of one procedure over the other in terms of infection control and functional outcome has not been identified to date [17]. However, several authors consider a two-stage approach the gold standard, as it is not limited by previous identification of the causative organism or soft tissue conditions and is associated with a potentially better infection control [18]. Usually for a two-stage approach, the infected prosthesis is first removed completely and a temporary bone cement spacer is implanted [17]. Reimplantation of a new prosthesis is performed several weeks later, provided that the infection appears to have been completely eliminated. However, two-staged exchange is associated with the morbidity of a second surgery and recurrence of infection varies; it may still be high according to Strickland et al. [26], who report a reinfection rate of up to 50%. Furthermore, there is a lack of studies investigating functional outcome and patient satisfaction after twostage exchange due to PJI. Therefore, in the present study, infection control, revision-free survival and functional outcome after this procedure were studied and reasons for poor outcome and reduced patient satisfaction were analyzed.

Inclusion Criteria and Patients
In this single-center study, we retrospectively analyzed the consecutive patients undergoing revision shoulder arthroplasty performed in our department from November 2010 to November 2019. We identified 122 patients. Of these, 106 were aseptic revisions (one-stage exchange, prosthesis conversions, component changes) and 16 (6 male, 10 female) underwent a two-stage exchange for PJI. Mean age at time of reimplantation was 65 years (range 41-76). Further demographic details are presented in Table 1. The exclusion criterion were revision of SA for other cause than infection, and repeat infections after already performing two-stage exchange following PJI were not considered new cases.
PJI comprised low-grade (symptoms > 3 months, chronical pain, joint fistula, increased infection parameters) and acute infections (symptoms for hours or a few days, systemic sepsis). The cases between the two definitions with symptom duration between 7 days and 3 months were referred to internally as "sub-acute". However, these cases were statistically combined with the acute cases. When diagnosing patients, a standardized algorithm was followed, irrespective of the presentation (acute infections vs. elective in low-grade infections). This algorithm always included a clinical examination, a laboratory analysis and, if the suspicion of infection was confirmed, joint aspiration. For selected cases with high suspicion, but a "dry tap", or clotted specimens, an open biopsy was performed. In detail, diagnosis was based on clinical symptoms, such as local warmth, stiffness, intermittent joint swelling or fistula in combination with elevated infection parameters, especially serum C-reactive protein (CRP) (>0.5 mg/dL) and elevated leucocytes in blood count. Further criteria were positive aspiration with elevated leucocytes (total leucocyte count >3000 and percentage of neutrophils >65%) and positive aspiration culture. Additionally, the majority of patients' positive microbiological findings (culture in open biopsy, intraoperative cultures) or macroscopic intraoperative findings like pus could confirm the suspected PJI.
The mean follow-up (FU) amounted to 32 months (range 12-85). According to Xu et al. [27], a 12-month minimum follow-up period is sufficient to investigate treatment failure in PJI, therefore, the minimum follow-up in the present study was 12 months. Range of motion (ROM) was measured prior to two-stage revision and at last FU. Postoperative outcome was measured with subjective shoulder value (SSV) which, according to Gilbart et al. [28], shows a high correlation to the constant score (CS). The "quick" Disabilities of the Arm, Shoulder and Hand (qDASH) and Rowe score were also measured [29,30]. The postoperative scores and ROM were compared in a subgroup analysis (reverse shoulder arthroplasty (RSA) vs. revision prostheses (hemiarthroplasty (HA)/reverse proximal humeral replacement (RPHR)) and all reverse prostheses vs. HA with CTA head).

Treatment of PJI by Two-Stage Exchange
The first step of two-stage revision included complete removal of implant components and implantation of an antibiotic-impregnated bone cement spacer (polymethylmethacrylate-PMMA) ( Figure 1). After surgery, antibiotic treatment was initiated using broad-spectrum antibiotics. The intravenous antibiotic treatment was carried out for at least two weeks with subsequent oral application for a further 4 weeks. During the interval until prosthesis reimplantation, the shoulder was immobilized in a brace. Reimplantation was indicated in the presence of normal CRP and the absence of clinical signs of infection. After reimplantation (Figure 2), the shoulder was again immobilized for 4 weeks. On day one, rehabilitation was started with passive movement exercises under physiotherapeutic supervision. After seven to fourteen days, active assisted exercises were allowed according to intraoperative joint stability. If the same microorganism was detected during reimplantation as during explantation, or a new microorganism was detected in more than one of 5 samples during reimplantation, an antibiotic treatment was administered for 6 weeks.

Statistical Analysis
All data were collected using a Microsoft Office Excel sheet (Microsoft Corp, Seattle, WA, USA) and all analyses were performed using SPSS for Windows Version 26 (IBM Corp, Redmont, VA, USA). We analyzed data distribution using the Kolmogorov-Smirnov test and present mean/medians with corresponding ranges or 25-75% interquartile ranges depending on data normality. Groupwise comparisons were done using the chi-squared test for cross tables and the Mann-Whitney U test or Student's t-test for (non-)parametric testing. Implant survivorship was calculated using the Kaplan-Meier method and survival curves for the primary endpoints (reinfection, overall revision) were generated. A p-value of <0.05 was considered significant.     Postoperative anterior-posterior X-ray imaging of a right shoulder after prosthesis reimplantation of a reverse shoulder arthroplasty (uncemented standard stem) and glenoidal augmentation with human allograft due to glenoid erosion.

General
We identified 12 low-grade PJIs and four acute infections. CRP was elevated in 12 patients with a mean of 3.0 mg/dL (range < 0.5-15.3; IQR 0.7-4.3). In four cases, joint fistulas indicated chronical PJI. The median retention period of the PMMA spacers before prosthesis reimplantation was 18.7 weeks (IQR 10-15.8). Further baseline data are summarized in Table 1.

Recurrence of Infection and Revision-Free Implant Survivorship
The mean period from index prosthesis implantation until diagnosis of acute or chronic PJI followed by prosthesis explantation and spacer interposition amounted to 14.2 months (IQR 3.2-28.4; range 1.5-34). The infection recurred in 19% (3/16) of patients. One patient who had undergone reverse proximal humeral replacement ( Figure 4) had acute early-onset PJI after 1.5 months, while two patients (1 RPHR, 1 RSA) showed late-onset and low-grade reinfection after seven and twelve months, respectively. These patients underwent repeat revision surgery, with two patients undergoing two-stage exchange again and the other undergoing definitive resection arthroplasty. The latter patient was subsequently free of pain in the mid-term with significantly reduced function, but has now been free of infection for over 4 years.

Recurrence of Infection and Revision-Free Implant Survivorship
The mean period from index prosthesis implantation until diagnosis of acute or chronic PJI followed by prosthesis explantation and spacer interposition amounted to 14.2 months (IQR 3.2-28.4; range 1.5-34). The infection recurred in 19% (3/16) of patients. One patient who had undergone reverse proximal humeral replacement ( Figure 4) had acute early-onset PJI after 1.5 months, while two patients (1 RPHR, 1 RSA) showed late-onset and low-grade reinfection after seven and twelve months, respectively. These patients underwent repeat revision surgery, with two patients undergoing two-stage exchange again and the other undergoing definitive resection arthroplasty. The latter patient was subsequently free of pain in the mid-term with significantly reduced function, but has now been free of infection for over 4 years.  The reinfection-free implant survivorship was 81% (95% CI (62-100%)) after one year and at final FU of 33.2 months (IQR 13.25-59.75, range 1.5-87.0) ( Figure 5). The overall revision-free survival (revision for any cause) amounted to 56% (95% CI 32-90%) after one year and at final follow-up. Thirty-one percent (5/16) of patients underwent revision surgery following non-infective complications (Table 3). Two patients with dislocations underwent inlay exchange and did not suffer from further instability, while, on the other hand, two patients with RPHR had recurrent instability despite inlay revision and lengthening of the modular reverse proximal humeral replacement. One was converted to anatomic modular megaprosthesis and one remained chronically unstable. Another patient with chronic instability after reimplantation of large-head HA refused repeat revision surgery. With the numbers available, there were no differences in implant survival with respect to the type of prosthesis used during reimplantation surgery. However, there was a 75% (3/4) revision rate for infection and non-infection reasons in RPHR while only 11% (1/9) of patients who underwent reconstruction with a non-megaprosthetic RSA suffered reinfection and 33% (3/9) underwent overall revision surgery.

Functional Outcome
Overall postoperative patient satisfaction was low with a subjective shoulder value (SSV) at last FU of 56/100 (IQR 28-78), qDASH with a mean of 40/100 (range 11-83) and a mean Rowe score of 52/100 (range 20-85) ( Table 4). However, patients who underwent RSA showed significantly better postoperative SSV compared to RPHR and prostheses with large-head HA (72/100 vs. 29/100, p = 0.001) ( Table 5) Table 4. Postoperative scores after second step of prosthesis reimplantation and range of motion (ROM) prior to prosthesis explantation (pre) and after second step of prosthesis reimplantation (post) of all patients. IQR: interquartile range, ER: external rotation, IR: internal rotation, ABD: abduction, FLEX: forward flexion, SSV: subjective shoulder value, qDASH: "quick" Disabilities of the Arm, Shoulder and Hand. Patients who underwent SA for proximal humerus fracture had a trend for a lower Rowe score after two-stage exchange  vs. 73 (51-85), p = 0.056) while the qDASH and SSV were not different (p = 0.43 and p = 0.18, respectively).

Discussion
The objectives of this study were the analysis of the recurrence of infection and infection-free survival after two-stage exchange of shoulder prosthesis following PJI. The most important finding was that infection control was achieved in the majority of patients, at over 80%. However, overall revision-free prosthesis survival was 56% and therefore major complications leading to revision surgery were common after two-stage treatment. If the anatomical conditions, particularly on the glenoid side, allowed reimplantation of an RSA, there was a trend towards increased patient satisfaction and improved postoperative function than after treatment with alternative prostheses.

Infection Control after Two-Stage Revision
The primary objective when performing revision surgery for infection is infection control [31][32][33], with most comprehensive data available for the two-stage approach [8,17,18,26]. In a recent review from 2019, Kunutsor et al. [17] included 27 studies with 351 patients. The recurrence rate of infection was 16.2% after a median FU of 3.9 years. Therefore, our results are in line with the literature. In contrast, eight studies of a one-stage prosthesis exchange with a FU of 3 years were reviewed and it is reported that 8% of the included patients (12 of 147) presented recurrent infection. According to these data, the less common one-stage treatment seems equivalent or even beneficial in terms of infection control. However, especially in the case of preoperatively unidentified microorganisms, a two-stage approach is considered as the method of choice by several authors to eradicate biofilm-forming bacteria, as local targeted antibiotic treatment is crucial [31,34]. Appropriate surgical restoration with extensive soft tissue and bone debridement appears to be crucial for adequate infection control [18]. In this context, an advantage of two-stage revision is the opportunity of a further tissue debridement without the need to perform reconstruction right away [5]. Considering that PJI after SA is still a rare but growing complication, future comparative studies should evaluate the ideal treatment approach.
The fact that MRSE was detected in almost 50% of cases (7/16; Table 3) at the time of the second stage of reimplantation is remarkable. For our part, however, only those cases in which MRSE had already been detected at the time of the first step (explantation) were evaluated as genuine infections, as the risk of contamination is present. Two of these patients in our study underwent revision for reinfection, however. Future studies should focus on optimizing antibiotic treatment for these bacteria as they might persist despite radical debridement.

Complication Rate after One-and Two-Stage Revision
The revision rate of 31% in our cohort for non-infectious complications is consistent with reports from the literature: Strickland et al. [26] reported up to 74% overall postoperative complications and a 29% revision rate following reimplantation in 17 patients (19 shoulders) [18,26]. Frequent complications were joint dislocation, chronic instability, acromion fracture, superficial wound healing disorders or hematoma formation. It has been reported that the complication rate may be lower with a single-stage procedure [17]. It may be discussed whether secondary damage caused by the spacer in the two-stage approach is a cause of this. Nevertheless, only one patient of the present study showed anterior spacer dislocation with resulting glenoid erosion ( Figure 6). Moreover, the increasing contracture of the surrounding soft tissue caused by the spacer remaining for a longer time may lead to higher risk of postoperative instability.
However, successful stabilization was achieved in all patients with postoperative dislocation following reimplantation of RSA. Our data show that reimplantation of alternative prostheses (large-head HA, proximal humerus replacement) may lead to more chronic instability in the long term. However, glenoid bone stock appears to be a limiting factor for conventional RSA. Therefore, the role of glenoidal augmentation is important in this approach. Nonetheless, future studies are needed on the use of bone grafting for glenoid bone defects in the setting of infection [35].

Clinical Outcome and Postoperative Function after Surgical Treatment of PJI
While pain reduction seems to be achievable in more than half of patients by resection arthroplasty, the functional outcome is poor due to severe limitations of movement [19,20]. Patient satisfaction is low, as has been reported by Braman et al. [19] and Romano et al. [13], who reported a postoperative CS of 30 and 32, respectively. Postoperative abduction and external rotation were described as up to 28° and 8°, respectively [36]. However, these results appear to be poor and somewhat controversial regarding the results reported by Sperling et al. [21], who reported an abduction of 70° and an external rotation of 31°. Considering the serious functional disadvantages, the indication for resection arthroplasty should be reserved for patients with high surgical risk.
The functional results following reconstruction with RSA are consistent with those in the literature. Nevertheless, patient satisfaction and functional outcome in our patients were significantly higher after RSA compared to revision prostheses (RPHR + large-head HA). One reason might be the reduced functionality of large-head HA and modular megaprosthesis. Therefore, the outcome seems to be closely associated with the anatomical conditions and extent of bone and soft tissue loss caused by previous interventions. Correspondingly, patients show a trend towards poorer function and reduced satisfaction after prosthetic treatment due to a humerus fracture.

Clinical Outcome and Postoperative Function after Surgical Treatment of PJI
While pain reduction seems to be achievable in more than half of patients by resection arthroplasty, the functional outcome is poor due to severe limitations of movement [19,20]. Patient satisfaction is low, as has been reported by Braman et al. [19] and Romano et al. [13], who reported a postoperative CS of 30 and 32, respectively. Postoperative abduction and external rotation were described as up to 28 • and 8 • , respectively [36]. However, these results appear to be poor and somewhat controversial regarding the results reported by Sperling et al. [21], who reported an abduction of 70 • and an external rotation of 31 • . Considering the serious functional disadvantages, the indication for resection arthroplasty should be reserved for patients with high surgical risk.
The functional results following reconstruction with RSA are consistent with those in the literature. Nevertheless, patient satisfaction and functional outcome in our patients were significantly higher after RSA compared to revision prostheses (RPHR + large-head HA). One reason might be the reduced functionality of large-head HA and modular megaprosthesis. Therefore, the outcome seems to be closely associated with the anatomical conditions and extent of bone and soft tissue loss caused by previous interventions. Correspondingly, patients show a trend towards poorer function and reduced satisfaction after prosthetic treatment due to a humerus fracture.

Strength and Limitations
We are providing data in a little-explored field, considering that current reviews include only about 350 patients. Despite this large amount of data, we were ultimately only able to accrue 16 cases. A subgroup analysis of different reimplanted prostheses was carried out and has displayed clear differences in function and survival, although the numbers available are too small for a meaningful statistical analysis. The shortcomings of this study are the retrospective design and the small number of patients in an inhomogeneous population with multiple influencing factors. Therefore, significance and risk analyses must be viewed with reservation due to the reduced statistical power.

Conclusions
Two-stage revisions following PJI of the shoulder allow infection control in the majority of patients. However, the overall levels of complications and revision rates due to mechanical failure are high. Reverse shoulder prostheses seem to be superior to alternative models regarding postoperative function and patient satisfaction. Therefore, bone-saving surgery and reconstruction of the glenoid may increase the likelihood of reimplantation of RSA and potentially improve outcome in the case of infection-related two-stage revision of the shoulder. Positive cultures at reimplantation appear to be an issue. Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author. The data are not publicly available due to hospital internal data.

Conflicts of Interest:
The authors declare no conflict of interest.