Shoulder arthroplasty in the management of native shoulder joint infections has a high complication rate and poor functional outcome – a systematic review

Abstract

Background

Shoulder arthroplasty is a treatment option of the sequelae of native shoulder joint infections. However, the functional outcomes and re-infection rates are unknown. The aim of this review was to analyse the outcome of shoulder arthroplasty in patients with native shoulder infections.

Methods

A review of the online databases MEDLINE and Embase was conducted according to PRISMA guidelines. The review was registered prospectively in the PROSPERO database. Studies reporting either primary or secondary infections of native shoulder joints treated with any form of arthroplasty were included and appraised using the methodological index for non-randomised studies (MINORS) tool.

Results

Fourteen studies were eligible for inclusion. Mean age ranged from 56 to 72 years and the mean follow-up from 20.5 months to 8.2 years. Primary shoulder infections were present in 50 patients and secondary infections in 86. 76 patients underwent a two stage: 46 patients a single-stage procedure whilst 14 refused second-stage surgery. Mean post-operative Constant score ranged from 38 to 56.2. The overall reported re-infection rate was 2.3% and complication rate was 26%.

Conclusion

Shoulder arthroplasty in the management of either primary or secondary native shoulder infections has a high complication rate and low functional outcome but low re-infection rates at short-term follow-up.

Introduction

Native shoulder infections account for 3%–15% of all septic arthritis cases^1,2 and often occur in older, medically complex patients or those with easily identifiable risk factors.^3,4 The most common pathogens are reported to be Cutibacterium acnes (C. acnes) and coagulase-negative staphylococcus (CoNS).^5,6 Although diagnosis relies on a combination of patient's history, clinical findings, blood infection markers, imaging, and joint aspiration, ⁷ the gold standard is repeated pathogen detection from multiple-tissue sampling ⁵ and the 2018 International Consensus Meeting (ICM) further defined a periprosthetic infection scoring system specific to shoulders. ⁸

Joint preservation by either arthroscopic or, in selected cases, open debridement should be the aim of initial treatment of shoulder infections. ⁹ Even after successful elimination of bacteria, cartilage and bone destruction is the consequence of prolonged inflammatory arthritis following infection mediated by pro-inflammatory cytokines. ¹⁰ Therefore, early and aggressive treatment should be used to optimise successful outcomes. ¹¹ A systematic review of joint preservation surgery in native shoulder infections revealed only 40% achieved infection eradication and functional improvement after a single arthroscopic procedure and 37% of patients had ongoing symptoms or complications even after eradication. ¹ Postinfectious sequelae of the native shoulder joint are frequently associated with poor clinical function, pain as well as diminished quality of life for affected patients.^12,13

If joint preservation is unsuccessful or when bone involvement occurs, then joint sacrificing surgery may become necessary. ⁹ Surgical options include resection shoulder arthroplasty,^14,15 glenohumeral arthrodesis ¹⁶ and shoulder arthroplasty. ¹⁷ Resection arthroplasty has been reported to give poor functional results.^14,15 The available evidence into use of shoulder arthroplasty in this cohort is limited. The aim of this systematic review was to analyse the outcome of shoulder arthroplasty in patients with native shoulder infections in terms of function, reinfection and complication rates.

Methods

A systematic review of the literature was conducted in accordance with the PRISMA ¹⁸ guidelines using the Pubmed and the Cochrane database. The review was registered on the PROSPERO database prospectively (CRD42022334856). The searches were performed independently by two authors on the 25^th of May 2022 and repeated on the 28^th of May 2022 to ensure accuracy. Search terms included ‘Shoulder infections’ and ‘Shoulder replacement’. Any discrepancies were resolved through discussion between these two authors.

The eligibility criteria were clinical studies published in the English language that reported on adult patients (over 16 years of age) managed surgically with either a long-term spacer, hemiarthroplasty (HA), total shoulder arthroplasty (TSA) or reverse shoulder arthroplasty (RSA) for a sequelae of a native shoulder joint infection. The shoulder infection could be defined as primary, either occurring after haematogenous spread or after an injection/aspiration, or secondary, after previous arthroscopy surgery including rotator cuff repair or shoulder fixation surgery. Studies reporting prosthetic joint infections (PJIs) were excluded and those studies where results of PJI and native shoulder infection could not be adequately separated were also excluded. The studies were required to report any of the following parameters: functional outcome, range of motion (ROM), reinfection or complication rates during follow-up for inclusion. Only primary research articles were considered for review. Abstracts, expert comments, review articles, biomechanical, and technique articles were excluded.

The clinical studies were appraised independently by two authors and quality assessment of non-randomised studies was completed using the methodological index for non-randomised studies (MINORS) tool ¹⁹ MINORS is a validated scoring tool for non-randomised studies. Each of the eight items in the MINORS criteria relevant to case series was given a score of 0, 1, or 2, with maximum score of 16 (Table 1).

Table 1.

Methodological items for non-randomized studies (MINORS) score.

	Kriechling et al. (25)a	Boelch et al. (20)	Goetti et al. (23)	Padegimas et al. (28)	Zhang et al. (33)	Morris et al. (27)	Schoch et al. (29)	Magnan et al. (26)	Stine et al. (30)	Coffey et al. (21)	Hattrup et al. (24)	Twiss et al. (32)	Themistocleus et al. (31)	Cuff et al. (22)
Score	22	11	12	13	12	15	16	11	12	12	12	13	7	22

^aThe global maximum score is 16 for non-comparative studies and 24 for comparative studies.

MINORS: methodological index for non-randomised studies.

Results

The search strategy is illustrated in Figure 1 and identified 1392 studies for consideration. After application of inclusion criteria, 14 studies were deemed eligible for inclusion (n = 136).^20–33 Kriechling et al., ²⁵ Schoch et al. ²⁹ and Morris et al. ²⁷ reported cases from a prospectively kept database but the remaining studies reported retrospective case series. Concise study characteristics are summarised in Table 2 and outcomes from these studies provided in Table 3.

Figure 1.

Flow diagram of review process.

Table 2.

Summary of studies reporting use of shoulder arthroplasty in native shoulder infections.

Study	Study design	Population	Patient demographics	Intervention (s)	One- or two-stage revision	Type of spacer	Antibiotic regimen	Mean follow-up	Culture results
Kriechling et al. 2021 (25) n = 14	Prospective 2005–2017	Primary sepsis (n = 1) Secondary sepsis (n = 13) – post ORIF (n = 4) – post cuff repair (n = 8) – post instability (n = 1) Matched comparison to n = 14 RSA for alternative indications 0 PJI cases	Mean age 64 ±8 Male 43%	100% Zimmer reverse shoulder system (Zimmer Warsaw, IN, USA)	100% one stage	Not used	3 peri-operative IV doses 8 patients received oral antibiotics until cultures negative	5 years (1–12)	MSSA 36% C. acnes 21% CoNS 7% Streptococcus 7% C. avidum 7% Serratia 7% Multiple 7%
Boelch et al. 2020 (20) n = 16	Retrospective 2007–2015	Primary sepsis (n = 8) Secondary sepsis (n = 8) – post ORIF (n = 6) – post cuff repair (n = 2) Excluded PJI (n = 23)	Mean age 72 (45–85) Male 31%	Debridement and custom spacer 100% 2nd stage with Aequalis reversed II prosthesis at mean 13 weeks (4–55)	100% two stage	Custom spacer around bent Steinmann pin Using Palacos gentamicin but adding vancomycin	At least 2 weeks tailored IV antibiotics Then oral course	Min 2 years (31–128 months)	CoNS 38% C. acnes 15% None 54%
Goetti et al. 2019 (23) n = 5	Retrospective 2012–2017	Primary sepsis (n = 1) Secondary sepsis (n = 4) – post ORIF (n = 2) – post cuff repair (n = 2) 0 PJI cases	Mean age 61 (47–70) 80% male	Debridement with >4 samples Custom spacer	100% 2nd stage at 6–12 weeks 2nd stage 40% HA and 60% RSA	Custom-made spacer with 1.2g tobramycin and 4g vancomycin	2 weeks IV antibiotics 3 months oral antibiotics	Mean 36.8 months (12–59)	C. acnes 40% Staphylococcus epidermidis 20% Streptococcus pyogenes 20% Streptococcus angiosus 20%
Padegimas et al. 2018 (28) n = 17	Retrospective 2010–2015	Primary sepsis (n = 5) – haematogenous (n = 1) – Post injection (n = 4) Secondary sepsis (n = 12) – post ORIF (n = 2) – post cuff repair (n = 8) – post ACJ excision (n = 1) – post labral repair (n = 1) 0 PJI cases	Mean age 65.4 ±12.2 Male 59%	Aspirate and multiple samples Four different implants utilised Zimmer Biomet (Warsaw, IN, USA) Depuy Synthes (Warsaw, IN, USA) DJO global (Vista, Ca, USA) Wright Medical (Memphis, TN, USA)	18% one stage 82% two stages 82% RSA, 12% HA, 6% TSA	Not specified	2 weeks of IV antibiotics unless positive cultures	4.1 years ±1.8 years	CoNS 24% C. acnes 18% Staphylococcus epidermidis 18% MRSA 12% MSSA 12% E. coli 12% Pseudomonas 6%
Zhang et al. 2015 (33) n = 7	Retrospective 2005–2012	Secondary sepsis (n = 7) – post cuff repair (n = 6) – post ORIF (n = 1) Excluded PJI (n = 11)	Mean age 69 (52–88) Male 83%	Extensive debridement and multiple cultures Custom spacer Further open biopsy prior to 2nd stage. If positive further debridement and exchange of spacer	100% two stage	Custom antibiotic-loaded spacer with vancomycin 1g and tobramycin 1.2g	6 weeks of IV antibiotics	22.3 months (12–36)	Staphylococcus epidermidis 57% C. acnes 29% MSSA 14% P. avidum 14%
Morris et al. 2015 (27) n = 8	Prospective 2004–2011	Primary sepsis (n = 1) Secondary sepsis (n = 7) – post ORIF (n = 1) – post cuff repair (n = 6) 0 PJI cases	Mean age 64.1 ±11.2 Male 75%	If acute or positive aspirate/biopsy then 2 stage. If remote infection and negative aspirate/biopsy 1 stage. Extensive debridement and culture 100% Aequalis reverse shoulder arthroplasty system (Tornier, Edina, Minnesota)	Two stage 62.5% One stage 37.5%	Custom antibiotic-loaded cement spacer with either vancomycin or tobramycin	24 h of IV antibiotics and further 4 days of oral clindamycin	Mean 4.4 years ±1.7	MRSA 12.5% Negative 87.5%
Schoch et al. 2014 (29) n = 23	Prospective 1977–2009	Primary sepsis (n = 14) – haematogenous (n = 10) – Post injection (n = 4) Secondary sepsis (n = 9) – post ORIF (n = 2) – post cuff repair (n = 2) – acromial surgery(n = 2) – other arthroscopic surgery (n = 3) 0 PJI cases	Mean age 56 (38–75)	Multiple soft tissue and bone samples including intra-operative frozen specimens 1 stage procedure 11 TSA and 12 HA	100% one stage	Not used	1 to 7 days of IV antibiotics	8.2 years (2–21.6)	C. acnes 9% CoNS 9% Pseudomonas 4% Negative 78%
Magnan et al. 2013 (26) n = 6	Retrospective	Primary sepsis (n = 3) Secondary sepsis (n = 3) – post ORIF (n = 3) Excluded PJI (n = 1)	Mean age 68.8 (52–82) Male 50%	Debridement and cultures Spacer 2nd stage if indicated at mean 7 months ±1 with mono-articular prosthesis	Two stage 60% 40% refused second stage and LT spacer	Interspace gentamicin loaded shoulder spacer (Tecres S.p.a. (Verona), Italy) with collar of vancomycin loaded cement	IV antibiotics for mean 3 weeks (2–5)	40 months (12–66)	MSSA 33% Staphylococcus epidermidis 33% Negative 33%
Stine et al. 2010 (30) n = 13	Retrospective 2003–2007	Primary sepsis (n = 4) – haematogenous (n = 4) Secondary sepsis (n = 9) – post ORIF (n = 1) – post cuff repair (n = 6) – instability surgery (n = 1) – other arthroscopic surgery (n = 1) 17 PJI cases	Mean age 61 (41–86) Male 62%	Aggressive debridement with multiple samples Custom spacer	Unable to separate from PJI data Overall 60% refused 2nd stage	Custom-made spacer using vancomycin 1g and tobramycin 1.2g	6 weeks of antibiotics	3.4 years (0.25 to 35)	Unable to separate from PJI data
Coffey et al. 2010 (21) n = 5	Retrospective 2006–2008	Primary sepsis (n = 4) Secondary sepsis (n = 1) – post ORIF (n = 1) Excluded PJI (n = 11)	Mean age 58.9 (45–89)	Extensive debridement, multiple cultures, bone biopsy with frozen section Pre-moulded spacer Revision to either TSA or RSA	Two stage for majority 25% refused 2nd stage including PJI patients	Interspace gentamicin loaded shoulder spacer (Tecres S.p.a. (Verona), Italy) with collar of vancomycin loaded cement	IV antibiotics (2–6 weeks)	20.5 months (12–30)	Unable to separate from PJI data
Hattrup et al. 2010 (24) n = 5	Retrospective 1997–2005	Primary sepsis (n = 2) Secondary sepsis (n = 3) – post cuff repair (n = 2) – post ORIF (n = 1) Excluded PJI (n = 20)	Mean age 66.9 (36–87) Male 60%	Extensive debridement and multiple cultures Custom spacer Revision to HA, TSA or RSA	Two stage 100% at mean 6 months	Custom antibiotic-loaded spacer with 4.8g gentamicin, 2g vancomycin and 2g cefazolin	6 weeks IV antibiotics	4.2 years (2 to 9.1)	Unable to separate from PJI data
Twiss et al. 2010 (32) n = 5	Retrospective 2006–2008	Primary sepsis (n = 4) Secondary sepsis (n = 1) – post ORIF (n = 1) Excluded PJI (n = 25)	Mean age 58.9 (45–89) Male 60%	Extensive debridement and multiple cultures Spacer Revision to HA, TSA or RSA	Unable to separate from PJI data Overall 30% refused 2nd stage	Antibiotics spacer either custom or Interspace gentamicin loaded shoulder spacer (Tecres S.p.a. (Verona), Italy)	IV antibiotics until biochemical markers normal	21.2 months (12–40)	Unable to separate from PJI data
Themistocleus et al. 2009 (31) n = 7	Retrospective	Primary sepsis (n = 3) Secondary sepsis (n = 4) – post ORIF (n = 2) – post cuff repair (n = 2) 5 PJI cases excluded	Mean age 64 (36–79) Male 86%	Radical debridement and samples Spacer	100% two stage	Custom-made spacer using vancomycin 1g and tobramycin 1.2g	6 weeks IV antibiotics	22 months (15–26)	MSSA 91% Neisseria sicca 9%
Cuff et al. 2007 (22) n = 5	Retrospective 1998–2005	Secondary sepsis (n = 5) – Post cuff repair (n = 5) 17 PJI cases excluded	Mean age 67.6 (52–82) 80% male	Extensive debridement, multiple cultures including frozen section 100% Implanted reverse shoulder prosthesis (Encore Medical Corporation, Austin, Texas)	Single stage (n = 3) Two stages (n = 2)	Not specified	2 weeks if culture negative 6 week IV if culture positive or sinus	43 months (25–66)	MRSA 40% MSSA 20% CoNS 20% Serratia 20%

ORIF: open reduction and internal fixation; IV: intravenous; PJI: prosthetic joint infection; HA: hemiarthroplasty; TSA: total shoulder arthroplasty; ACJ: acromioclavicular joint; WCC: white cell count; CRP: C-reactive protein; MRSA: methicillin-resistant staphylococcus aureus; MSSA: methicillin-sensitive staphylococcus aureus; CoNS: coagulase negative staphylococcus; C. acnes: Cutinbacterium Acnes; RSA: reverse shoulder arthroplasty.

Table 3.

Summary of outcomes in studies reporting use of shoulder arthroplasty in native shoulder infections.

Study	Mean follow-up	Outcome	Re-infection rate	Complications	Functional outcome and ROM
Kriechling et al. 2021 (25) N = 14	5 years (1–12)	Re-infection ROM CMS SSV Complications	0 cases of re-infection	Complication rate 36% and revision rate 21% - 2 peri-prosthetic fractures requiring surgery - 2 neuropathies - 1 instability requiring revision - 1 revision for excision of scar tissue	Comparison to matched cohort with RSA for alternative indication - CMS 24 ± 10 to 38 ± 17 - CMS 33 ± 16 to 75 ± 8 - CMS p < 0.001 - SSV 24 ± 14 to 43 ± 26 - SSV 29 ± 20 to 85 ± 10 - SSV p < 0.001 - Flexion 63° ± 36 to 82° ± 31 - Flexion 82° ± 49 to 131° ± 16 - Flexion p = 0.014 - ER 28° ± 28 to 18° ± 25 - ER 19° ± 21 to 35° ± 22 - ER p = 0.07
Boelch et al. 2020 (20) N = 16	Min 2 years (31–128 months)	Re-infection ROM VAS pain CMS QuickDASH Complications	0 cases of re-infection	Complication rate 12.5% - 1 acromial fracture at 3 months - 1 humeral peri-prosthetic 4 years	- Constant 48 (7–85) - QuickDASH 40 (11.4–93.2) - VAS 1.6 (0–7) - Flexion 113° (20–180) - ER 12° (0–40)
Goetti et al. 2019 (23) N = 5	Mean 36.8 months (12–59)	Re-infection ROM DASH SSV Complications	0 cases of re-infection	Complication rate 0%	- DASH 18.4 (7.5–40) - SSV 70% (40–95%) - Flexion 102° (70–130) - ER 25° (10–45)
Padegimas et al. 2018 (28) N = 17	4.1 years ± 1.8 years	Re-infection VAS pain SANE ASES SST Complications	0 cases of re-infection	Complication rate 23.5% - 2 dislocations requiring revision - 1 acromial fracture - 1 cuff tear in HA requiring revision	- VAS 4.6 ± 2.3 - SST 6.9 ± 2.6 - ASES 57.6 ± 15.5 - SANE 59.3 ± 23.7 -
Zhang et al. 2015 (33) N = 7	22.3 months (12–36)	Reinfection ROM ASES Pain	0 cases of re-infection at final follow-up 1 case of positive culture at open biopsy requiring 2nd debridement (14%)	Not reported	- ASES 77 (52–95) - Flexion 119° (50–140) - ER 36° (0–80)
Morris et al. 2015 (27) N = 8	Mean 4.4 years ± 1.7	Reinfection rate ROM CMS ASES SANE Complications	0 cases of re-infection	Complication rate 37.5% - 2 acromial fractures - 1 intra-operative humeral fracture	- CMS 14.4 ± 6 to 56.2 ± 19.4 - ASES 38 ± 14.8 to 78.4 ± 15.2 - SANE 24.3 ± 27.2 to 59.9 ± 34.5 - Flex 7.5° ± 13.9 to 143.3° ± 21.3 - ER 3.8° ± 7.5 to 20° ± 14.6
Schoch et al. 2014 (29) N = 23	8.2 years (2–21.6)	Reinfection ROM Pain Subjective satisfaction Complications	2 cases of re-infection (8.7%) both requiring revision surgery	Complication rate 35% - 5 implants loosening - 2 haematomas - 1 instability	- VAS 4.5 to 2.1 (p < 0.001) - 70% report being better or much better - Abduction 62° to 110° - ER 14° to 47°
Magnan et al. 2013 (26) N = 6	40 months (12–66)	Re-infection rate ROM ASES CMS SECEC elbow score Complications	0 cases of re-infection	Not reported	ASES - Primary sepsis 12.3 to 21.3 - Secondary sepsis 11 to 22 CMS - Primary sepsis 41 to 79 - Secondary sepsis 38 to 78 SECEC elbow score - Primary sepsis 34.5 to 67.5 - Secondary sepsis 33.7 to 76. Flexion 102.5° (60–150)
Stine et al. 2010 (30) N = 13	3.4 years (0.25 to 35)	Reinfection rate ROM DASH STT Complications	0 cases of re-infection	Unable to separate PJI and native infection data	Unable to separate PJI and native infection data
Coffey et al. 2010 (21) N = 5	20.5 months (12–30)	ROM UCLA shoulder rating scale SST ASES CMS Recurrence	0 cases of re-infection	Unable to separate PJI and native infection data	Unable to separate PJI and native infection data
Hattrup et al. 2010 (24) N = 5	4.2 years (2 to 9.1)	Reinfection ROM ASES SST VAS pain	Unable to separate PJI and native infection data	Unable to separate PJI and native infection data	Unable to separate PJI and native infection data
Twiss et al. 2010 (32) N = 5	21.2 months (12–40)	Reinfection ASES UCLA VAS	0 cases of re-infection	Not reported	Unable to separate PJI and native infection data
Themistocleus et al. 2009 (31) N = 7	22 months (15–26)	Reinfection rate Pain Satisfaction Loosening	0 cases of re-infection	Not reported	- No or occasional pain only - 100% satisfied
Cuff et al. 2007 (22) N = 5	43 months (25–66)	Re-infection ASES Satisfaction score	0 cases of re-infection	Complication rate 20% - Haematoma requiring washout	- ASES 41 to 72 - Excellent 40%, good 40% and satisfied 20%

ROM: range of motion; VAS: visual analogue score; ER: external rotation; SSV: simple shoulder value; ASES: American Shoulder and Elbow Surgeons Shoulder score; UCLA: University of California Los Angeles; DASH: the Disabilities of the Arm, Shoulder an Hand questionnaire; SST: the Simple Shoulder Test; SANE: single assessment numeric evaluation; CMS: Constant Murley Score; HA: hemiarthroplasty; RSA: reverse shoulder arthroplasty; PJI: prosthetic joint infection.

The mean ages of patients in the included studies ranged from 56 to 72 years and the mean follow-up ranged from 20.5 months to 8.2 years. Gender was reported in 13 studies. Overall, 60% of patients were male and in only two studies were females more prevalent. The studies included both primary (n = 50) and secondary shoulder infections (n = 86). The primary group included infections secondary to haematogenous spread (n = 15), following shoulder injection/aspiration (n = 8) and unspecified causes (n = 27). The secondary group included infections following previous rotator cuff repair (n = 49), previous fracture fixation (n = 27) or other arthroscopic surgery (n = 10). The commonest micro-organisms identified at initial debridement ranged from, methicillin-sensitive staphylococcus aureus (MSSA) 12–91%, CoNS 7–38%, C. acnes 9–40% and staphylococcus epidermidis 18–57%. Seventy-six patients (56%) underwent a two-stage surgical procedure with implantation of a spacer initially before implantation of prosthesis during a second stage. Forty-six patients (34%) underwent a single-stage procedure with implantation of the prosthesis at first surgery and the remaining 14 patients (10%) opted not to proceed with a planned second-stage choosing to live with spacer in situ.

Only three studies provided details on the delay between primary infection and joint replacement clearly. Kreichling et al. performed joint replacement between 0.1 and 10.6 years, Schoch et al. performed after a mean of 8.2 years (range 0.25–48years) and Themistocleus et al. after 1 year (range 3 m – 1.5years). All but one study ruled out infection with tissue sample and monitored bloods. Eleven studies reported that bloods markers were checked to monitor treatment response to infection prior to joint replacement surgery while three also re-took intra-operative samples prior to any arthroplasty. Two studies did not clarify what investigations were undertaken prior to the surgery.

Spacer and anti-microbial management

Ninety patients underwent a planned two-stage procedure with use of a shoulder spacer. In three case series, the spacer used was pre-manufactured and contained gentamicin, Interspace gentamicin loaded spacer (Tecres S.p.a. Verona, Italy). This spacer was additionally secured within the humeral shaft with cement loaded with 1 g of vancomycin. Five case series used a custom spacer made at the time of surgery combining the cement with vancomycin 1 g and tobramycin 1.2g.^{23,27,30,31,33} A sixth case series exchanging tobramycin for 1.2 g of gentamicin. ²⁰ The final series reporting on the use of a custom spacer loaded the cement with three antibiotics; gentamicin 4.8 g, vancomycin 2 g and cefazolin 2g. ²⁴

Patients undergoing one- and two-stage surgery all received post-operative antibiotics. Three studies provided only peri-operative antibiotics for under 1 week, three studies routinely prescribed 2 weeks of intravenous antibiotics unless cultures were positive and the final eight treated with prolonged intravenous for between 3 and 6 weeks. Interestingly, the two largest papers reporting on one-stage surgery reported giving only three peri-operative intravenous antibiotic doses ²⁵ and under 1 week ²⁹ respectively.

Re-infection rate

Thirteen (n = 131)^{20–23,25–33} of the included 14 studies reported the rate of reinfection. The data for reinfection from the final study from Hattrup et al. ²⁴ could not be separated from the combined PJI data and therefore was excluded from this part of the analysis. Only three cases of reinfection were reported out of the 131 included cases (2.3%) during the follow-up (mean range 20 months to 8.2 years). Of the 46 patients undergoing single-stage surgery, two were subsequently found to be reinfection (4.6%) and of the 76 undergoing two-stage procedure, one patient had infection at later biopsy (1.3%).

Zhang et al. ³³ reported a 0% reinfection rate, but the authors routinely performed open biopsy prior to the second-stage procedure. One case was identified as having positive cultures at this stage for Staphylococcus epidermidis necessitating further debridement, exchange of spacer and further 6 weeks of intravenous antibiotics and thus is identified as a re-infection in this review. Schoch et al. ²⁹ had the longest follow-up period in the included studies at 8.2 years and reported the highest rate of reinfection, two cases out of 23 (8.7%) using a single-stage technique. The first case developed humeral and glenoid loosening 6.5 years post-operatively and at revision cultures demonstrated Staphylococcus aureus; however, given initial infection was 40 years prior it is unknown if this was the same causative pathogen. The second patient developed an abscess at 1.6 years post-operatively and samples at resection arthroplasty demonstrated a CoNS which did not match the initial MSSA isolating suggesting a de novo infection.

Complications

Seven studies (n = 88)^{20,22,23,25,27–29} reported complication data that could be separated from PJI data. The overall complication rate was 26% with 23 out of the 88 patients suffering a complication after arthroplasty following native shoulder infections. The commonest complications reported were implant loosening 5.7%, acromial fractures 4.5%, humeral peri-prosthetic fractures 4.5% and joint instability 4.5%.

Morris et al. ²⁷ reported the highest complication rate at 37.5% but included only eight patients in their series; two acromial fractures managed non-operatively and one humeral fracture intra-operatively during placement of the uncemented prosthesis. Kriechling et al. ²⁵ reported the second highest rate of complications with 6 of their 14 patients (36%) suffering a complication within their 5-year follow-up. Two patients suffered a subsequent peri-prosthetic fracture requiring revision, one patient had shoulder instability requiring revision surgery, two patients had ongoing neuropathy and the final patient had excessive scar tissue requiring excision. Schoch et al. ²⁹ reported a 35% complication rate in their 23 patients but also reported the longest follow-up period at 8.2 years: 5 implant loosening, 2 haematoma and 1 instability.

Functional outcome and range of motion

Eight of the included studies (n = 78)^{20,22,23,25–28,33} reported patient reported outcome measures that could be isolated from PJI data. Concise details of the functional outcomes are available in Table 3. The Constant Murley Score was reported in four studies and the mean post-operative score ranged from 38 to 56.2. The American Shoulder and Elbow Surgeon Shoulder score was similarly reported in four studies and the mean post-operative score ranged from 57.6 to78.4. Kriechling et al. ²⁵ provided a matched control group undergoing RSA for alternative indications and reported that there was a significantly inferior outcome in RSA performed in native shoulder infections with both CMS and SSV lower 38 ± 17 versus 75 ± 8 (p < 0.001) and 43 ± 26 versus 85 ± 10 (p < 0.001), respectively.

Seven studies (n = 79)^{20,23,25–27,29,33} reported ROM data that could be separated from concomitant PJI data with mean flexion achieved ranging from 82 to 143 and external rotation ranging from 18 to 36. Kriechling et al. ²⁵ reported that patients undergoing RSA for native shoulder infections had a significantly inferior ROM compared to matched individuals undergoing RSA for alternative indications; flexion 82° ± 31° versus 131° ± 16° (p = 0.014), abduction 80° ± 40° versus 148° ± 29° (p < 0.001) and external rotation 18° ± 25° versus 35° ± 22° (p = 0.07).

Discussion

This review has demonstrated a low reinfection rate (2.3%) but a high overall complication rate of 20%, when treating either primary or secondary native shoulder joint infections with shoulder arthroplasty, although the short follow-up period across the studies may partly explain this low value. Numerous previous studies have demonstrated the safety and efficacy of hip and knee arthroplasty for the treatment of degenerative joints with a history of infection with reinfection rates ranging from 0% to 9.5%,^{5–7,9,34–36} and it appears shoulder arthroplasty in this setting is also a viable option. The majority of previous literature relates to prosthetic shoulder joint infections and report a higher reinfection rate than the current review; for example, Buchalter et al. demonstrated a 21% reinfection rate at mean of 63 months. ³⁷ This increase in re-infection rates in the PJI setting is supported by two papers from this review who reported both on patients with native shoulder infections and shoulder PJIs; Boelch et al. ²⁰ report a 9% reinfection rate in PJIs compared to 0% for native shoulder infection and Hattrup et al. ²⁴ reported 15% reinfection rate in PJIs and 0% for native shoulder cases. These higher values in PJIs have been reproduced in hip and knee systematic reviews; Masters et al. ³⁸ demonstrated reinfection rates of 0–41% for two-stage and 0–11% in single-stage knee revisions (46) and Kunutsor et al. ³⁹ reported reinfection rates of 8.2% (6.0–10.8) after single-stage and 7.9% (6.2–9.7) after two-stage hip surgery.

Studies in this review utilised both a one-stage (n = 46) and two-stage (n = 76) surgical approach. Although the reinfection rate was higher after one-stage (4.6%) versus two-stage surgery (1.3%), there were no comparative studies, the numbers were low and there was a high level of data heterogeneity precluding statistical testing of these results. Advantages of a two-stage procedure include the use of spacers to elucidate local antibiotics, ⁴⁰ maintenance of a functional soft tissue envelope for later reconstruction ⁴¹ and offering a long-term management option for frail patients. ⁴² However, one-stage procedures reduce the morbidity associated with multiple procedures, avoid repetitive damage to the soft tissue envelope around the shoulder and should therefore be considered if proven to give comparable outcomes to the two-stage alternative. Direct comparative studies in the management of native shoulder infections are not available however systematic reviews comparing one- and two-stage procedures for shoulder PJIs are. Mercurio et al. ⁴³ and Ruditsky et al. ⁴⁴ both report higher reinfection rates after two-stage (14–15%) when compared to one-stage surgery (4–5%). Potential explanations for this higher reinfection rate after two-stage procedures could be its use in more complex cases, in cases where pathogens are harder to eradicate and de novo infections. Whether these results relating to shoulder PJIs can be extrapolated to a native shoulder infection situation is currently unknown. In addition, Ruditsky et al. concluded that no consensus is yet available and that both treatment strategies are effective form of managing shoulder PJIs. The authors suggested that decision-making between one- and two-stage revisions is multifactorial with pathogen type, timing of infection, clinical findings, patient's opinion, surgeon preference and intraoperative soft-tissue/osseous appearance needing to be considered ⁴⁴ . In addition, a three-stage revision protocol has recently been suggested for shoulder PJIs with promising results of 0% reinfection in 28 patients, ⁴⁵ and this approach for native shoulder infections was reported in one study during this review. Zhang et al. ³³ identified one patient out of seven at this additional step with ongoing infection that was successfully treated with further debridement and antibiotics before reimplantation ³³ suggesting this three-stage approach may have a role.

The overall complication rate reported in this review was high at 26% with the commonest being implant loosening (5.7%), joint instability (4.5%) and associated fractures (4.5%). A high complication rate may be as a result of the poor quality of the soft tissue envelope secondary to both previous tissue damage from infection and multiple surgeries with resultant difficulties balancing soft tissue tension. ²⁸ However, the reported complication rate associated with primary TSA and RSA for arthritic shoulders is relatively high at 13–15% and in the setting of revision RSA the complication rate has been reported to increase to 40–50%.^23,46,47 A review from Bois et al. ⁴⁸ created subgroups for revision RSA dependent on what previous surgery patients had undergone and the authors demonstrated the highest complication rates when revising a failed RSA (56.2%), followed by HA (27.7%), TSA (23.6%), soft-tissue repair group (20.6%) and then ORIF group (19.0%) ⁴⁸ (57). In the current review, all patients had undergone previous surgery and therefore it is not surprising to see a complication rate that is higher than that associated with primary arthroplasty surgery.

Functional outcomes were reported in only eight studies and included reporting a variety of outcome measures. The commonest reported were the CMS (mean post-operative score 38 to 56.2) and the ASES score (mean post-operative score ranged from 57.6 to 78.4). These results appear suboptimal when compared to shoulder arthroplasty performed in for primary arthritis.^47,49 This is supported by Kriechling et al. ²⁵ who provided a matched cohort of RSA that were performed for alternative indications in their study and reported statistically worse outcomes in the native shoulder infection group in terms of Constant Score (p < 0.001), SSV (p < 0.001), flexion (p = 0.014) and abduction (p < 0.001). Previous literature related to reverse arthroplasty has demonstrated that functional outcomes are significantly lower in both post-traumatic cases and in revision surgery.^42,50 In addition, Buchalter et al. ³⁷ also demonstrated similarly low levels of function after revision surgery for shoulder PJI with mean ASES score of 69 (range 32–98). ³⁷ The inferior outcomes reported in this review after previous native joint infections may be secondary to patients having several previous procedures with severely compromised soft tissues affecting both function and stability.^17,25 Although RSA potentially is less reliant on the rotator cuff for function and therefore maybe resistant to prior debridement of soft tissues and maybe expected to provide a superior functional outcome to HA or TSA in this revision setting, the available data in this review is not sufficient to directly compare functional outcomes after HA, TSA and RSA in the treatment of native joint infections. Interestingly, four studies^21,26,30,32 reported cases where patients declined to proceed with the second stage as the introduction of a cement spacer had resulted in an acceptable outcome for the patient. Overall, this occurred in 16% of cases that had been planned for a two-stage revision and appears a reasonable option for patients who choose to decline further surgery although functional outcomes for this subgroup are not available.

Authors suggest the following recommendations as illustrated in Figure 2. (a) Clinical examination, inflammatory markers, MRI to assess for extent of bone involvement and obtain previous microbiology results. (b) Pre-operative multidisciplinary discussion involving surgeons, radiologist and microbiologist. (c) Aspiration of joint to guide choice of antibiotic used in subsequent spacer. (d) Extensive open debridement, removal of any implants and non-viable tissue, multiple cultures including frozen section where available. (e) Implantation of cement spacer loaded with antibiotics targeted at previous microbiology sensitivities where possible. (f) Empirical IV antibiotics until prolonged cultures available then conversion to complete 6 week of antibiotics according to sensitivities in discussion with microbiology team. (g) Monitor clinically and biochemical makers off antibiotics and proceed to second stage after 3 months if no concerns and patient dissatisfied with function with spacer. (h) If ongoing concerns over infection then further open biopsy/cultures with subsequent debridement and exchange of spacer if positive cultures. (i) If no concerns then proceed to second-stage arthroplasty with further sampling and use of prolonged antibiotics only if latest cultures are positive.

Figure 2.

Suggested algorithm for the management of native shoulder infection with arthroplasty.

Limitations of this systematic review are acknowledged. Table 1 illustrates the MINOR grading of the included studies which ranged from 7 to 22. Recurrent weaknesses identified included the majority of studies were retrospective, contained low numbers and had either short- or medium-term follow-up only. Long-term follow-up is required to gain a fair reflection on the successful eradication of infection. Data generally included both native shoulder infections and PJI with difficulty experienced in some papers to separate this data for some outcome measures.

Conclusion

Shoulder arthroplasty in the management of primary or secondary native shoulder joint infection is a viable option but has a high complication rate of 26% and relatively poor functional outcome. While the re-infection rates are low at short-term follow-up, long-term studies are required to ensure these re-infection rates remain low and functional outcomes are maintained in this challenging patient population.