Diagnosing a periprosthetic shoulder infection: A systematic review

Abstract

Introduction

The aim of this study was to systematically review the literature regarding accurate shoulder prosthetic joint infection (PJI) diagnosis.

Methods

Using PRISMA guidelines, we analyzed 25 studies reporting on 5535 patients and 646 infections.

Results

Cutibacterium acnes (C. acnes) cultures were positive in 60% of patients. Serum markers WBC, CRP, ESR, and IL-6 appear to lack diagnostic reliability. Synovial IL-6 and alpha-defensin may be more accurate in detecting infections.

Conclusion

Synovial IL-6 and alpha-defensin appear to have greater utility than serum markers. These may be incorporated into new criteria to accurately diagnose shoulder PJI.

Level of evidence

IV.

1. Introduction

Prosthetic joint infection (PJI) is one of the most feared complications following total shoulder arthroplasty with reported rates that range from 0.4 to 4% after primary arthroplasty and up to 15% after revision arthroplasty.^1,2 Almost 80,000 shoulder arthroplasties are now performed annually in the United States, with an expected increase of almost 300% in the near future.^1,3, 4, 5 Thus, PJIs can place a tremendous burden on the health care system as they are associated with substantial cost and morbidity.^6,7 Risk factors for periprosthetic shoulder infection include male gender, younger age, drug abuse, and certain medical comorbidities.⁴

PJI of the shoulder represents a difficult diagnostic challenge since it often presents without the typical signs and symptoms associated with a deep infection of the knee or hip as shoulder PJI is most often caused by low-virulence organisms, resulting in an indolent presentation.⁸ Furthermore, common screening makers such as C-reactive protein (CRP), and serum erythrocyte sedimentation rate (ESR), which are highly sensitive in identifying periprosthetic hip and knee infections, lack diagnostic reliability in the shoulder.⁸ Pre-operative aspiration and culture has similarly been shown to be unreliable with a high false-negative rate.^9,10 Delayed or incorrect diagnosis of infection in the shoulder can lead to delays in the proper treatment.¹¹ Early and accurate identification of an infection is critical for mitigating the long-term effects of PJI.

Given the difficulty in accurately diagnosing shoulder PJI with traditional approaches, attention has turned to evaluating additional serum and synovial cytokines and biomarkers. These include IL-6, leukocyte esterase, and alpha-defensin, amongst others. Given the current lack of consensus as to the most accurate diagnostic tests and criteria in detecting infection, the purpose of this study was to systematically review the literature regarding the diagnosis of shoulder PJI.

2. Methods

Utilizing the Medline, Embase, and Ovid databases, we searched for all studies between January 1st^, 1996 and October 1st^, 2020 that assessed diagnosis of periprosthetic shoulder infection. This was performed utilizing the Preferred Reporting Items for Systematic Review and Meta-Analyses guidelines.¹² We searched for studies published in English using the search string ((“revision” OR “failed”) AND “shoulder” AND (“arthroplasty” OR “replacement")), which yielded 632 studies. We then carefully screened the titles and abstracts utilizing the exclusion criteria: (1) included patients with shoulder infection without arthroplasty; (2) reported on patients with positive cultures not considered infection or that were “unexpected,” i. e, no strict definition of infection; (3) included patients with arthroplasty of joints other than the shoulder. After screening, a total of 25 studies which met inclusion criteria were included in our final analysis (Fig. 1). The search was performed by one of the authors (JJJ) and then again by two others (MWC/ATT) in an independent fashion to validate the included studies and avoid the omission of any relevant studies.

Fig. 1.

Summary of studies included and excluded in this study.

When evaluating included studies, we noted number of patients involved, number of confirmed infections, and mean patient age. For the purposes of this study, “true infection” was recorded according to the criteria used by each individual study given the lack of a gold-standard for the diagnosis of shoulder PJI. We recorded the objective clinical, laboratory, and histologic criteria used in each study (Appendix A).

Furthermore, we analyzed the effectiveness of all diagnostic markers utilized to detect PJI including white blood count (WBC), Interleukin-6, C-reactive protein (CRP), serum erythrocyte sedimentation rate (ESR), Leukocyte esterase, culture results, next generation sequencing, alpha-defensin, and other infection parameters. We accomplished our analysis by noting their cut-off values, mean, standard deviation, number of patients with values greater than the cut-off point, sensitivity, specificity, positive predictive value, and negative predictive value when provided in the studies. In addition, we also collected data on the causal organism. For organisms causing less than 2% of the infections, these were grouped into a miscellaneous category. All data was extrapolated to an electronic spreadsheet (Microsoft Excel, Microsoft Office, Redmond, Washington). This study was performed without external funding.

3. Results

Our evaluation included 5535 patients, of which 646 were diagnosed with periprosthetic shoulder infections. The mean age of the infected patients ranged from 54 to 73 years. The mean number of males infected across studies was 65.7%, and the mean average time to infection diagnosis after surgery was 34 months (Table 1). C. acnes and coagulase-negative staphylococci were the culprits of infection in 60.0% and 10.8% of the patients, respectively (Table 2). Staphylococcus aureus accounted for 6.9% of the patients while less common organisms, such as Acinetobacter calcoaceticus, Alcaligenes species, Bacillus species, Corynebacterium species, Enterobacter cloacae, Enterococcus gallinarum, Peptostreptococcus, Proteus mirabilis, Pseudomonas species, and Streptococcus Viridans, accounted for approximately 18.3% of infections (Table 2).

Table 1.

Study flowchart of the 25 studies incorporated in this systematic review.

Author, Year	LOE	Number of patients in study	Number of infected patients	Number of males infected	Mean age in years (Range or SD)	Mean time of diagnosis from surgery in months (Range)
Primary Arthroplasties
Grosso et al., 2014	III	69	24	N/R	62 (35–86)	54 (1–204)
Dodson et al., 2010	IV	11	11	8	60 (47–78)	25 (12–48)
Frangiamore et al., 2015 (Synovial fluid …)	II	46	5	N/R	60 (±13.2)	N/R
Piper et al., 2010	IV	64	19	N/R	N/R	1
Nelson et al., 2015	II	45	5	N/R	70	N/R
Dilisio et al., 2014	IV	19	19	11	57 (±11.5)	36 (9–93)
Villacis et al., 2014	II	34	14	N/R	64 (36–81)	31 (2–134)
Grosso et al., 2014	III	45	30	20	N/R	N/R
Sperling et al., 2001	III	2279	23	14	54 (24–75)	42 (0–177)
Millett et al., 2011	IV	10	10	8	58 (24–81)	20 (1–96)
Butler-Wu et al., 2011	IV	87	42	N/R	N/R	N/R
Kelly II et al., 2009	V	28	8	N/R	62 (43–81)	22 (12–37)
Foruria et al., 2013	IV	107	11	10	60 (±13)	67 (1–300)
Frangiamore et al., 2015 (α-Defensin …)	III	30	4	N/R	62 (±12.4)	N/R
Topolski et al., 2006	III	75	74	50	60 (27–79)	35 (2–150)
Pottinger et al., 2012	II	193	108	N/R	N/R	N/R
Singh et al., 2012	II	1431	14	7	63 (±16)	96 (N/R)
Ecker et al., 2019	III	105	24	14	68 (±13)	N/R
Revision Arthroplasties
Frangiamore et al., 2015 (Early Versus Late …)	III	208	46	42	59 (±13)	N/R
Piper et al., 2009	III	136	33	22	60 (44–81)	14
Nelson et al., 2015	II	40	16	N/R	68	N/R
Namdari et al., 2019	II	44	13	N/R	68.7 (±9.7)	N/R
Sperling et al., 2001	III	194	8	N/R	N/R	N/R
Frangiamore et al., 2017	III	67	36	N/R	63.8	N/R
Hecker et al., 2020	III	106	24	NR	62	2 (0–6)
Falstie-Jensen et al., 2019	III	29	11	NR	63	NR
Meinshausen et al., 2019	II	33	14	NR	72.8	36

Table 2.

Compilation of identified infection-causing organisms reported by all studies.

Organism	Number of shoulders	Percent of shoulders
Cutibacterium acnes	377	60.0%
Coagulase negative staphylococcus	69	10.8%
Staphylococcus aureus	44	6.9%
Staphylococcus epidermidis	13	2.0%
Other organisms with <2%a	117	18.3%
Non-reported	19	3.0%
Total	639

^aAcinetobacter calcoaceticus, Alcaligenes species, Bacillus species, Corynebacterium species, Enterobacter cloacae, Enterococcus gallinarum, Peptostreptococcus, Proteus mirabilis, Pseudomonas species, and Streptococcus Viridans.

3.1. Diagnostic criteria: white blood cell count

Villacis et al. analyzed 14 patients undergoing revision surgery for infection and found that a WBC cutoff value of 11.0 × 10⁹ had a sensitivity of 7%, specificity of 95%, positive predictive value of 50%, and a negative predictive value of 59%.¹³ Other studies included in this analysis did not report a WBC cut-off sensitivity, specificity, PPV, or NPV. Millett et al. evaluated a cohort of 10 infected patients and reported a mean preoperative WBC of 6.8 × 10³ with a range spanning from 4.8 × 10³ to 10.4 × 10³.¹⁴ Similarly, Sperling et al. evaluated a cohort of 29 infected patients and found a preoperative mean WBC of 7.4 × 10³ with a range spanning from 3.8 × 10³ to 15.6 × 10³.¹⁵ The authors further specified that only 2 patients reported a preoperative WBC count greater than 10 × 10³.¹⁵

3.2. Erythrocyte sedimentation rate

A study conducted by Grosso et al. of 24 infected patients showed sensitivity and specificity of 42% and 82%, respectively, for patients with an ESR above 15 mm/h (millimeters/hour).¹⁶ Another study conducted by Piper et al. evaluated 64 patients of which 19 had confirmed infection. The authors evaluated patients using two different sets of ESR cut-off values (30 mm/h and 26 mm/h). With a cut-off value of 30 mm/h, the reported sensitivity and specificity was 16% and 98% respectively with a positive predictive value and a negative predictive value of 75% and 73%, respectively. After applying the optimized ESR of 26 mm/h, sensitivity increased from 16% to 32% but specificity decreased (98%–93%). Moreover, this also resulted in a lower positive predictive value (75%–67%) and a higher negative predictive value (73%–76%).¹⁷ Other studies in our evaluation reported a wide range of ESR values which extended from 0 mm/h to 135 mm/h along with a wide range of cut-off values (10 mm/h to 30 mm/h). The sensitivity ranged from 21% to 61% and specificity ranged from 65% to 93% (Table 3).

Table 3.

Summary of studies reporting Erythrocyte Sedimentation Rate (ESR) serum and synovial cut-off, sensitivity, specificity, mean, and % greater than cutoff values for infected patients.

Author, Year	Type of Arthroplasty	Cut off (mm/hr)	Sensitivity %	Specificity %	Mean (Range or SD) in infected patients	Percent of patients with ESR greater than cutoff
Serum
Grosso et al., 2014	Primary	15	42	82	N/R	33%
Grosso et al., 2014	Primary	10	61	85	N/R	N/R
Piper et al., 2010	Primary	26	32	93	9 (1–71)	N/R
Villacis et al., 2014	Primary	30	21	65	(0–44)	29%
Frangiamore et al., 2015	Primary	14.2	N/R	N/R	N/R	N/R
Piper et al., 2009	Revision	30	N/R	N/R	N/R	24%
Topolski et al., 2006	Primary	22	N/R	N/R	12 (0–66)	8%
Kelly II et al., 2009	Primary	22	N/R	N/R	N/R	50%
Sperling et al., 2000	Primary	30	N/R	N/R	47 (10–135)	61%
Pottinger et al., 2012	Primary	15	N/R	N/R	N/R	18%
Synovial Fluid
Nelson et al., 2015	Revision	23 (mean)	N/R	N/R	N/R	N/R

3.3. Serum CRP

Villacis et al. noted the CRP range to be 0.3–3.5 mg/L among 14 infected patients. They utilized a cut-off value of 10 mg/L as part of their evaluation of 34 revision patients, which yielded poor sensitivity and a positive predictive value of 0%. However, this cut-off value led to a high specificity of 95%, a negative predictive value of 57%, and an accuracy of 56%.¹³

Another study conducted by Piper et al. also utilized 10 mg/L as a cut-off value for CRP and reported a sensitivity of 42% with positive predictive value of 53%. Furthermore, they noted a specificity of 84% and a negative predictive value of 78%. Then, after lowering the CRP cut-off value to 7 mg/L, they noted an increase in sensitivity (63%) and a decrease in specificity (73%).¹⁷ Other studies noted a wide range of CRP values which extended from 0.2 mg/L to 40 mg/L along with a wide range of cut-off values (1.0 mg/L to 10 mg/L). Furthermore, a wide range of sensitivity (0–75%) and specificity (44–95%) was also reported (Table 4).

Table 4.

Summary of studies reporting C-Reactive Protein (CRP) serum and synovial cut-off, sensitivity, specificity, mean, and % greater than cutoff values for infected patients.

Author, Year	Type of Arthroplasty	Cutoff (mg/L)	Sensitivity %	Specificity %	Mean (Range or SD) in infected patients	Percent of infected patients with CRP greater than cutoff
Serum
Grosso et al., 2014	Primary	10	46	93	N/R	33%
Grosso et al., 2014	Primary	N/R	33	85	N/R	N/R
Frangiamore et al., 2015	Primary	7	N/R	N/R	N/R	N/R
Piper et al., 2009	Revision	10	N/R	N/R	N/R	36%
Topolski et al., 2006	Primary	1	N/R	N/R	7 (0.5–32)	5%
Kelly II et al., 2009	Primary	10	N/R	N/R	N/R	63%
Piper et al., 2010	Primary	7	63	73	10 (median) (13–40)	N/R
Villacis et al., 2014	Primary	10	0	95	(0.3–3.5)	0
Pottinger et al., 2012	Primary	10	N/R	N/R	N/R	13%
Ecker et al., 2019	Primary	5	75	44	N/R	N/R
Synovial Fluid
Nelson et al., 2015	Revision	Infected	40.3 (±4.3)	Non-infected	5.1 (±0.7)	N/R

A recent study conducted by Unter Ecker et al. utilized a cut-off value of 5 mg/L for a study in 105 patients, which demonstrated a sensitivity of 75%, specificity of 44%, negative predictive value of 84%, and positive predictive value of 32%.¹⁸ The authors suggest a potential benefit to decreasing the cut-off value for CRP given the indolent nature of many shoulder PJI's as less virulent organism such as C. acnes have been associated with a lower CRP.¹⁸

3.4. Synovial or serum IL-6

In a cohort of 24 infected patients, Grosso et al. investigated the utility of IL-6 in diagnosing shoulder PJI. Utilizing 5 pg/mL as their cut-off value, they reported 2 patients with values greater than cut-off with maximum reported levels of 8 pg/mL. Furthermore, the reported sensitivity and specificity of IL-6 was 12% and 93%, respectively. They concluded that serum IL-6 is not an effective diagnostic tool due to poor sensitivity.¹⁶ Villacis et al. also evaluated role of serum IL-6 and found only 2 patients of 14 infected patients had IL-6 levels greater than the cut-off value of 10 pg/mL. They reported a sensitivity and specificity of 14% and 95%, respectively, similarly concluding that serum Il-6 is not a useful test to screen for shoulder PJI.¹³

Frangiamore et al. analyzed synovial IL-6 in patients undergoing primary¹⁹ and revision²⁰ shoulder arthroplasty, and their results suggested that synovial IL-6 may be a more useful marker than serum IL-6. In their cohort of primary arthroplasties, they determined the ideal cut-off value for synovial IL-6 was 359.3 pg/mL, with sensitivity, specificity, and positive and negative likelihood ratios of 87%, 90%, 8.45, and 0.15, respectively (Table 5). In their cohort of revision arthroplasties, they utilized a multiplex immunoassay to examine levels of 9 cytokines (IL-6, granulocyte macrophage colony stimulating factor [GM-CSF], IL-1β, IL-12, IL-2, IL-8, interferon-γ [IFN-γ], IL-10, and tumor necrosis factor-α [TNF-α]). Notably, they showed that a combined cytokine model consisting of IL-6, tumor necrosis factor-α, and IL-2 showed better diagnostic test characteristics than any cytokine alone, with a sensitivity of 0.80, specificity of 0.93, positive and negative predictive values of 0.87 and 0.89, and positive and negative likelihood ratios of 12.0 and 0.21.²⁰

Table 5.

Summary of studies reporting Interleukin-6 (IL-6) serum and synovial cut-off, sensitivity, specificity, and frozen histology median values for infected patients.

Author, Year	Type of Arthroplasty	Cutoff (pg/mL)	Sensitivity %	Specificity %	Frozen histology median IL-6 (pg/ml) in C. acnes
Serum
Grosso et al., 2014	Primary	5	12	93	N/R
Villacis et al., 2014	Primary	10	14	95	N/R
Synovial Fluid
Frangiamore et al., 2015	Primary	359.3	87	90	8,531
Frangiamore et al., 2017	Revision	453.6	82	87	N/R

3.5. Synovial leukocyte esterase strip testing

Leukocyte esterase (LE) is a relatively new marker for diagnosing periprosthetic shoulder infection. Nelson et al. investigated its utility in five patients diagnosed with infection after primary shoulder arthroplasty and sixteen patients diagnosed after revision shoulder arthroplasty. Their reported specificity of leukocyte esterase in infected patients who were diagnosed after primary shoulder arthroplasty and revision arthroplasty was 85% and 75%, respectively. Sensitivy in patients diagnosed after primary shoulder arthroplasty was not reported. The reported positive predictive value and negative predictive value of LE in the revision group was 57% and 75%, respectively. Overall, their study indicated that LE should not be used in diagnosing periprosthetic shoulder infection because of poor sensitivity, which was 23.8%.²¹

A follow-up study conducted by Unter Ecker et al. sought to further characterize the utility of the leukocyte esterase test in diagnosing infection. In their study, the authors utilized a cohort of 42 patients, 12 of which were found to have a PJI based on their diagnostic criteria. The leukocyte esterase test was positive in 6 of the infected patients and negative in the remaining 6. The authors calculated a sensitivity of 50% and a specificity of 87%. The negative predictive value was found to be 81% and the positive predictive value to be 60%. The study's poor sensitivity was attributed to poor synovial fluid yield and sample admixture with blood, leading to many “indeterminate” samples (Table 6).¹⁸

Table 6.

Summary of studies reporting Leukocyte Esterase synovial cut-off, sensitivity, specificity, and frozen histology median values for infected patients.

Author, Year	Type of Arthroplasty	Synovial level cuttoff	Sensitivity %	Specificity %	Positive predictive value %	Negative predictive value %
Synovial Fluid
Nelson et al., 2015	Primary	N/R	N/R	85	N/R	90
Ecker et al., 2019	Pimary	N/R	50	87	60	81
Nelson et al., 2015	Revision	N/R	25	75	57	75

3.6. Synovial alpha-defensin

The utility of alpha-defensin in diagnosing periprosthetic shoulder infection was evaluated by a prospective study conducted by Frangiamore et al. They used 0.48 s/co (signal to cut-off ratio) as the cut-off value for synovial alpha-defensin. Noninfected patients had much lower levels of alpha-defensin (0.21) than the infected patients (3.2). Furthermore, sensitivity and specificity were reported as 63% and 95%, respectively. The positive and negative likelihood ratios were 12.1 and 0.38, respectively. They concluded that synovial alpha-defensin showed good diagnostic accuracy in shoulder infection because a positive likelihood ratio of 12.1 suggests that positive alpha-defensin value should raise the pretest odds of having an infection by a factor of 12.1.²²

Unter Ecker et al. also investigated the utility of using a-defensin as a diagnostic lab value. The alpha defensin test outperformed CRP, WBC, and the leukocyte esterase test in terms of accurately diagnosing PJI. With a cut-off value of 0.9, and sensitivity and specificity were found to be 75% and 96%, respectively. Negative predictive value and positive predictive values were found to be 93% and 86% (Table 7).¹⁸

Table 7.

Summary of studies reporting alpha-Defensin synovial cut-off, sensitivity, and specificity values for infected patients. Median infected vs. non-infected patient values were included as well.

Author, Year	Type of Arthroplasty	Cut off value (s/co)	Sensitivity %	Specificity %	Median in infected patients	Median non infected patients
Synovial Fluid
Frangiamore et al., 2015	Primary	0.48	63	95	3.2	0.2
Ecker et al., 2019	Primary	0.9	75	96	N/R	N/R

3.7. Culture

Cutibacterium acnes is a slow growing microorganism. Frangiamore et al. evaluated the appropriate amount of time required to optimize sensitivity and specificity for shoulder PJI diagnosis. They retrospectively evaluated a cohort of 208 patients, 46 patients of which had confirmed infections. They found that mean time required for culture was 13.1 days (±3 SD) with a range of 8–26 days.²³ Dodson et al. showed similar culture data, with a reported mean time to isolate C. acnes of 8.8 days (range: 8–10 days).²⁴ Furthermore, Pottinger et al. investigated a cohort of 108 infected patients of which 44 cases were caused by C. acnes. Their data indicated that 86% and 100% of cultures were positive in 14 and 28 days respectively (Table 8).²⁵ Lastly, Hecker et al. studied a cohort of 106 patients and demonstrated that aspirations for fluid and culture analysis were of limited diagnostic accuracy with a sensitivity of just 33% for a positive bacterial yield.²⁶

Table 8.

Summary of studies reporting relevant culture data for confirmed infected patients.

Author, Year	Type of Arthroplasty	Mean/median time for culture growth (days)	% of culture positive in 14 days	% of culture positive in 28 days	Number of patients with peri-op positive frozen section	Percent of patients with peri-op negative frozen section
Serum
Frangiamore et al., 2015	Revision	13.1	N/R	N/R	9	34
Dodson et al., 2010	Primary	8.8	N/R	N/R	N/R	N/R
Kelly II et al., 2009	Primary	7	N/R	N/R	N/R	N/R
Dilisio et al., 2014	Primary	10.1	N/R	N/R	N/R	N/R
Millett et al., 2011	Primary	7	N/R	N/R	N/R	N/R
Pottinger et al., 2012	Primary	N/R	86	100	N/R	N/R

3.8. Next generation sequencing

Next generation sequencing (NGS) is a relatively new method that involves using PCR to sequence an entire bacterial genome from a synovial fluid sample. Namdari et al. analyzed 44 revision arthroplasty patients to determine any correlation between traditional synovial fluid culture and NGS data. The authors determined that there was “fair concordance” between conventional culture data and next generation sequencing results.²⁷ However, the authors notably point out that bacterial thresholds are currently not defined, which makes NGS-based diagnosis challenging at this time.²⁷ Additionally, NGS identified bacteria such as E. coli and other atypical organisms which were never grown on culture, raising the concern that NGS may be over-sensitive and identifying non clinically relevant remnant DNA from dead bacteria.²⁷ Ultimately, NGS appears to have promising features but requires future studies for clarification on its role in diagnosing PJI.

3.9. WBC/BM SPECT CT

The gold standard for radionuclide imaging of lower extremity PJI is combined labeled leukocyte (WBC) and technetium 99 m sulfur colloid bone marrow imaging (WBC/BM). Several studies have demonstrated sensitivities and specificities above 90% for WBC/BM with single-photon emission computed tomography (SPECT) in lower extremity PJI.^28,29 In contrast, WBC/BM SPECT CT has not been shown to be useful for the diagnosis of shoulder PJI. For instance, Falstie-Jensen et al. reported the WBC/BM scan results of 11 infected patients, and found 2 true positives along with 9 false negatives for a sensitivity of 18%.³⁰ The authors highlighted that none of the patients in their study culture-positive for C. acnes had a positive WBC/BM SPECT CT study, further illustrating the difficulty in the diagnosis of infections caused by low-virulence organisms.³⁰

3.10. Terminal complement pathway

The complement system serves to activate immune defense cells, destroy foreign or dead cells and enhances phagocytosis via opsonization of antigens on pathogens. Meinshausen et al., in a study of 33 revision arthroplasties, investigated if components of the terminal part of the complement system, specifically C3, C5 and C9, would allow for differentiation between septic and aseptic loosening in shoulder arthroplasty. A statistically significant increase in all three complement factors was seen in 14 septic shoulders versus 19 aseptic shoulders; furthermore, C9 was found to distinguish between aseptic and septic tissue with a specificity of 100% and a sensitivity of 88.89%.³¹

4. Discussion

Early and accurate diagnosis of shoulder PJI remains elusive. Patients often present only with symptoms of pain and stiffness making differentiation from other etiologies such as aseptic loosening or rotator cuff tears difficult. This indolent presentation is in large part due to the low virulence of the most common infecting organisms such as C. Acnes and coagulase-negative Staphylococcus. Commonly utilized blood tests such as CBC, ESR and CRP have shown poor sensitivity in detecting shoulder PJI. This is in contrast to hip and knee PJI, where both ESR and CRP are first-line screening tests given their well-researched high sensitivities and specificities.³² Similarly, shoulder joint aspiration and culture has been shown to be far less sensitive than in the hip and knee which is in part due to the fact that C. acnes is typically shielded from the immune system via a biofilm.³³

These challenges have driven the investigation of various diagnostic tests to potentially improve diagnosis of periprosthetic shoulder infection after arthroplasty. Data from our pooled results indicate that markers known to be reliable in diagnosing periprosthetic infection in the hip and knee, such as serum ESR and CRP, have limited utility with PJI of the shoulder. For shoulder PJI, across the studies that were analyzed, both of these markers demonstrated no reliable cutoff value that would yield high specificity and sensitivity. As such, these markers may not be of high clinical relevance alone when attempting to diagnose a shoulder PJI because of the possibility of normal values in the presence of infection. However, they may be useful in the context of other exam findings and diagnostic testing.

New tests for the diagnosis of PJI have recently been investigated and early results suggest that they may be more efficacious for evaluation of infection in the shoulder. Amongst these new tests are peptide markers such as α-defensin and IL-6. Defensins are antimicrobial peptides that represent a part of the innate immune system, whose role is to directly neutralize invading pathogens.³⁴ Furthermore, their production may be regulated by proinflammatory cytokines such as IL-6, which are known to be elevated in the setting of PJI.34, 35, 36 Recent studies have demonstrated that synovial levels of α-defensin³⁵ and IL-6³⁷ have sensitivities and specificities >95% in the diagnosis of hip and knee PJI.

Our pooled results reaffirmed similar findings regarding the utility of assaying antimicrobial peptides in the setting of shoulder PJI, wherein both α-defensin and IL-6 had better sensitivities and specificities than traditional markers of infection. Studies demonstrated that synovial α-defensin had a sensitivity and specificity of 63% and 95% respectively.²² Synovial IL-6 demonstrated far better diagnostic ability than serum IL-6, with synovial IL-6 values having a sensitivity and specificity of 82–87% and 87–90%, respectively.^19,20 Conversely, serum IL-6, in multiple studies demonstrated sensitivity ranging from 12 to 14%.¹⁶ These results suggest that it may be more useful to focus on these synovial fluid markers rather than serum IL-6 because of their improved diagnostic utility.

The findings of this review highlight the imperative to investigate new biomarkers for the diagnosis of shoulder PJI. Several recent studies have investigated the use of synovial peptides as markers of PJI in the hip and knee. Although more difficult to obtain than serum samples, synovial samples capture cytokine profiles directly in the affected joint. Deirmengian et al. identified several synovial biomarkers in the setting of hip and knee PJI, and follow up studies have demonstrated these markers to have areas under the curve (AUCs) greater than 0.9, with specificities of 95%–97% and sensitivities of 89%–100%.^35,37 This systematic review notes the superiority of synovial IL-6 to serum IL-6 and also suggests that a combined cytokine model should be investigated in future studies given its potential promise. In addition, further studies with a larger cohort of patients are necessary to expand upon the potential use of the complement system in identifying and diagnosing shoulder PJI.

A major limitation in most of the studies analyzed is the lack of consensus criteria for the diagnosis of infection (i.e., lack of a gold-standard). The majority of studies included in this systematic review utilized guidelines proposed by the Musculoskeletal Infection Society (MSIS) to diagnose PJI in the hip and knee.^38,39 However, these guidelines are often difficult to interpret and require a complex algorithm of serum, synovial fluid, and histologic parameters to diagnose infection. The most recent International Consensus Meeting (ICM) on Musculoskeletal Infection in 2018 proposed a scoring system based on history, exam findings, and diagnostic testing to both aid in diagnosing a periprosthetic infection as well as to create a standardized definition for future research studies on shoulder PJI. Prior to this, there had been no established consensus criteria in PJI diagnosis, which led to much of the primary literature using different methods to create a diagnosis of infection. The authors proposed a series of both major and minor criteria, similar to that of the MSIS criteria that has been consistently used for PJI diagnosis of the hip and knee. Meeting any one of the major criteria was determined to be diagnostic of a definite shoulder PJI and included “a sinus tract communicating with the prosthesis, gross intraarticular pus, or two positive cultures with phenotypically-identical virulent organisms.“⁴⁰ The minor criteria included many different diagnostic tests, as well as radiographic and exam findings and placed a weighted value on each in order to categorize failed shoulder arthroplasties into 3 different groups: probable PJI, possible PJI, and unlikely PJI.⁴⁰ Our Appendix A illustrates just how heterogenous in nature diagnostic criteria is. The utilization of this above consensus criteria in future studies could help clarify and standardize the statistical efficacy of the biomarkers investigated in this review in terms of shoulder PJI diagnosis.

When considering revision arthroplasty, a correct preoperative diagnosis is important when deciding on a single-stage or two-stage exchange arthroplasty in patients with chronic PJI. However, the indolent nature of infecting organisms, such a C. acnes, presents an additional diagnostic challenge. It is not uncommon for a one-stage revision to be performed because perioperative markers of infection were negative at the time of revision, only to be followed by an unexpected positive result of intraoperative culture after surgery. Ultimately, the use of synovial biomarkers reviewed in this study, such as α-defensin and IL-6, in helping identify shoulder PJI may lead to improved decision making, such as determining the appropriate indications for one or two-stage revision. However, a limitation of this review is the limited data and sample sizes available regarding studies of these new biomarkers of shoulder PJI. Large multicenter studies are necessary to better characterize infected patients and more accurately define the diagnostic and predictive parameters of these new biomarkers.

5. Conclusion

This study highlights the lack of a standard definition for an infected shoulder arthroplasty as well as the difficulty in accurate and timely diagnosis of such an infection. The poor sensitivity of common diagnostic makers such as ESR and CRP, along with the indolent nature of Cutibacterium acnes, which is the most common infecting organism, makes establishing the diagnosis extremely challenging. Biomarkers such as synovial IL-6 and synovial alpha-defensin have shown promising results in identifying periprosthetic shoulder infections, whereas serum markers have low sensitivity. Complement pathway marker analysis and combined cytokine models have demonstrated promise in improving diagnostic accuracy. These specific markers along with clinical signs, symptoms, and imaging can be helpful in accurately detecting infections in their early stages. Further evaluation of additional synovial markers in isolation or combination and their potential utility is necessary to improve the ability to accurately diagnose shoulder PJI.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Disclaimer

None.

CRediT authorship contribution statement

Julio J. Jauregui: Conceptualization, Methodology, Validation, Formal analysis, Investigation, Resources, Writing – original draft, Writing – review & editing, Visualization. Andrew Tran: Formal analysis, Investigation, Resources, Writing – original draft, Writing – review & editing. Samir Kaveeshwar: Formal analysis, Writing – original draft, Writing – review & editing, Visualization. Vidushan Nadarajah: Methodology, Formal analysis, Writing – original draft, Writing – review & editing. Moiuz W. Chaudhri: Methodology, Formal analysis, Writing – original draft, Writing – review & editing. R. Frank Henn: Writing – original draft, Writing – review & editing, Supervision. Mohit N. Gilotra: Writing – original draft, Writing – review & editing, Supervision. S. Ashfaq Hasan: Writing – original draft, Writing – review & editing, Methodology, Data curation, Formal analysis, Project administration, Supervision, Validation.

Appendix A. Diagnostic Criteria Used by Each Study to Define Infection

Study	Diagnostic Criteria
Sperling et al., 2001	Clinical course, purulence at the time of surgery and sinus communicating with the joint
Topolski et al., 2006	At least one positive intraoperative culture after revision shoulder arthroplasty
Kelly II et al., 2009	Infection if any 1 of the following: •Any associated skin erythema, wound drainage, or obvious purulence or tissue synovitis at the time of surgery •Clinical aspiration yielding a positive Gram stain or culture, •Positive intraoperative Gram stain or frozen section showing more than five polymorphonuclear leukocytes per high-powered field
Piper et al., 2009	Infection if any 1 of the following: •visible purulence •acute inflammation on histology •a sinus tract communicating with the prosthesis
Dodson et al., 2010	2 positive intraoperative specimens
Piper et al., 2010	Infection if any 1 of the following: •visible purulence, •acute inflammation via histology, •sinus tract, •positive periprosthetic culture, •positive sonicate culture
Butler-Wu et al., 2011	Infection if any 1 of the following: •2 or more cultures positive of same organism •1 culture positive with more than 5 neutrophils
Millett et al., 2011	Culture-positive specimens obtained deep to the deltoid in patients with symptoms of pain, prosthetic dysfunction, or joint sepsis
Pottinger et al., 2012	Clinical evidence of infection (stiffness, pain, or loosening) AND positive culture
Singh et al., 2012	Infection if any 1 of the following: •positive joint fluid culture from needle aspiration, arthroscopic procedure, fluid obtained at surgery or fluid draining from a wound communicating with the humerus •clinically suspected septic arthritis plus either culture-negative purulent or serosanguineous joint fluid or necrotic joint tissue (or culture not performed) or positive blood culture •frank pus/purulent material at surgery •positive synovial or bone tissue culture.
Foruria et al., 2013	One positive preoperative culture (aspirate) AND 1 positive intraoperative culture with the same organism
Dilisio et al., 2014	Clinical evidence of infection (stiffness, pain, or loosening) AND positive tissue culture
Grosso et al., 2014	At least 1 preoperative or intraoperative finding of infection AND more than 1 positive intraoperative culture OR 1 positive preoperative (aspirate) culture and 1 one positive intraoperative culture
Grosso et al., 2014	At least 1 preoperative or intraoperative finding of infection AND more than 1 positive intraoperative culture OR 1 positive preoperative (aspirate) culture and 1 one positive intraoperative culture
Villacis et al., 2014	At least one positive intraoperative culture of peri-implant tissue
Frangiamore et al., 2015*, 2017	Infection if any 1 of the following: •At least 1 positive preoperative or intraoperative finding of infection AND more than 1 positive culture (preoperative or intraoperative) OR One positive preoperative culture (aspirate) AND 1 positive intraoperative culture with the same organism •At least 1 positive preoperative or intraoperative finding of infection AND one positive culture (preoperative or intraoperative) or No preoperative or intraoperative findings of infection AND more than one positive culture preoperative or intraoperative).
Nelson et al., 2015	Presence of sinus tract, pathogen isolated by culture from 2 or more different samples AND at least 4 of the following 6: 1.Elevated CRP or ESR, 2.Elevated synovial leukocyte count, 3.Elevated synovial neutrophil percentage, 4.Purulence in the affected joint, 5.Isolation of microorganism in one culture 6.Greater than 5 neutrophils per high-powered field
Meinshausen et al., 2018	Microbiological testing evidencing positive tissue culture
Falstie-Jensen et al., 2019	Presence of sinus tract OR successful pathogen isolated by culture from at least 3 separate samples OR at least two of the following: 1.Elevated CRP or ESR or WBC, 2.A single positive culture 3.Positive radionucleide imaging 4.Unexplained nightly pain or excessive joint stiffness
Ecker et al., 2019	Infection if any 1 of the following: •Direct fistula communicating with shoulder joint •Two periprosthetic positive cultures with phenotypically identical organisms
Namdari et al., 2019	Divided into 4 groups: 1.Definite Infection: ≥1 positive preoperative or intraoperative finding of infection and >1 positive NGS result with the same organism identified 2.Probable Infection: ≥1 positive preoperative or intraoperative finding of infection and 1 positive NGS result, or 0 preoperative or intraoperative findings of infection and >1 positive NGS result with the same organism identified 3.Probable Contaminant: 0 preoperative or intraoperative findings of infection and 1 positive NGS result 4.No evidence of infection: 0 preoperative or intraoperative findings of infection and 0 positive NGS results
Hecker et al., 2020	Infection was defined as the presence of at least 2 intraoperative tissue cultures growing the same pathogen

*The diagnostic criteria used by Frangiamore et al. is applicable to 4 studies included in this systematic review, three published in 2015 and one published in 2017.