As the number of primary shoulder arthroplasties increases, the prevalence of periprosthetic shoulder infections (PSIs) will also increase.1,2 Although hip and knee periprosthetic infections have a higher prevalence compared with PSIs,3–5 the diagnosis of PSI is often more challenging because the majority of PSIs are caused by nonsuppurative bacteria such as Cutibacterium (formerly Propionibacterium) acnes (C acnes) and coagulase-negative staphylococci species (CoNS).5–8 Until recently, no commercially available pre- or intraoperative test could reliably predict the presence of C acnes.5 No validated tests for shoulder periprosthetic infection are approved currently by the Food and Drug Administration.5 The purpose of this review is to discuss current and new diagnostic tests for preoperative and intra-operative detection of PSI, particularly C acnes and other less virulent organisms.
Periprosthetic shoulder infections remain a diagnostic challenge, with no single test or diagnostic algorithm proven to effectively diagnose this condition. The diagnostic criteria used to diagnose a periprosthetic hip or knee infection created by the Musculoskeletal Infection Society (MSIS)9,10 have been shown to be unreliable for diagnosis of PSI.6,11 This challenge exists because a majority of PSIs are caused by nonsuppurative bacteria, such as C acnes and CoNS, that do not reliably cause elevation of inflammatory markers commonly used in the diagnosis of hip and knee prosthetic joint infection.5–8
When considering the diagnosis of a PSI, a combination of clinical findings, serum and synovial studies, and intraoperative pathologic studies have traditionally been used. However, studies by Millet et al,6 Pottinger et al,7 and Dodson et al11 have shown that erythema, purulence, and drainage are rarely observed in the infected prosthetic shoulder. Furthermore, systemic markers of inflammation found in the serum, such as erythrocyte sedimentation rate (ESR), C-reactive protein (CRP), and white blood cell (WBC) count, are often not elevated and, unlike in hip and knee prosthetic joint infections (PJIs), serum interleukin-6 is not useful.6,7,11,12
In contrast to hip and knee PJIs, preoperative aspiration rarely is diagnostic in the shoulder for a multitude of reasons. First, there is often not enough fluid, as demonstrated by a study by Millet et al6 where no fluid was obtained for a majority of aspiration attempts, even under fluoroscopic guidance. Second, the culture time for C acnes is much longer than for other bacteria, often taking more than 8 days for cultures to turn positive, making extended duration cultures of 14 to 21 days necessary.6,7,11 Third, the culture may not be diagnostic and has a low sensitivity for infection, ranging from 30% to 66%.6,11 Finally, intraoperative frozen sections often do not correlate with the incidence of PSI, as demonstrated by Topolski et al13 with only 6 (8%) of 73 patients with positive intraoperative tissue culture and frozen sections with acute inflammation at the time of revision. This lack of diagnostic evidence based on markers of inflammation and infection is likely due to the slow growing, nonsuppurative bacteria that are the hallmark of PSI.
Diagnosing Periprosthetic Shoulder Infection
Because the traditional diagnostics for more virulent and suppurative bacteria are less effective for the nonsuppurative organisms commonly found in PSI, different techniques are required to identify the underlying pathogen. Some of these techniques include serum alpha defensin, serum D-dimer, culturing intraoperative tissue in blood culture media, polymerase chain reaction (PCR) and restriction fragment length polymorphism (RFLP), and next generation sequencing (NGS). These newer technologies hold promise in the diagnosis of PSI.
Defensins are antimicrobial peptides expressed in epithelial cells and leukocytes that show broad-spectrum antimicrobial activities against gram-positive and -negative bacteria. As part of the innate immune system, their role is to neutralize invading pathogens without prior sensitization.14–16 Alpha defensin is 29 to 35 amino acids long, is highly concentrated in neutrophils, and is secreted into the synovial fluid in response to infection.
There are 2 methods of testing for alpha defensin. The laboratory-based immunoassay was developed from studies of genomes and proteomes and gives a qualitative result specific for synovial fluid (CD Diagnostics, Wynnewood, Pennsylvania). The quicker, on-table lateral flow test kit (Zimmer Biomet, Warsaw, Indiana) was developed to test intraoperative synovial fluid for PJI (Figure 1). The systematic review and meta-analysis of 10 studies (2 of which included the shoulder joint) by Suen et al17 showed that the immunoassay test demonstrated higher sensitivity and specificity (95% and 96%, respectively) than the lateral flow test (77% and 91%, respectively).
A negative Synovasure PJI lateral flow test with only the control band (“C”) is present (A). The positive test shows the control band (“C”) and the alpha-defensin band (“a-D”) (B). The control band must be visible to have a valid test result. PJI, prosthetic joint infection. [Reprinted from Journal of Arthroplasty, 31(12), Kasparek MF, Kasparek M, Boettner F, et al, Intraoperative diagnosis of periprosthetic joint infection using a novel alpha-defensin lateral flow assay, 2871–2874, 2016, with permission from Elsevier.]
The alpha defensin test shows promise, as demonstrated by many studies. Although the lateral flow test is not yet approved by the Food and Drug Administration for the diagnosis of PJI in the United States, it is approved for use in Europe. Deirmengian et al18–21 are credited with the invention of this test and have performed much of the validation research. In one of their earliest studies, they evaluated the synovial fluid of 95 patients who met the MSIS criteria for PJI of the hip or knee. From an original list of 43 potential biomarkers, 16 proteins were consistently elevated in the presence of PJI and 5 biomarkers demonstrated 100% sensitivity and specificity (alpha defensin, neutrophil elastase, bactericidal/permeability-increasing protein, neutrophil gelatinase-associated lipocalin, and lactoferrin). Notably, patients taking antibiotics and patients with systemic inflammatory disease, normally excluded from similar reports, were included in this study.18
Berger et al22 performed a prospective study analyzing the diagnostic accuracy and reliability of the lateral flow device for detecting alpha defensin proteins in synovial fluid vs the MSIS criteria for diagnosing hip or knee PJI. Importantly, patients with metallosis and/or recent antibiotic use were not excluded. The results of this study showed that the lateral flow test had very high sensitivity and specificity (97.1% and 96.6%, respectively; 1 false-negative and 3 false-positives) and a high positive and negative predictive value (91.7 and 98.8, respectively). In addition to being highly sensitive and specific, the test responds well to the less virulent organisms, including C acnes and Staphylococcus epidermidis, making it helpful and relevant for the diagnosis of shoulder PJI.19,22
Two other studies examining the performance of the alpha defensin test on patients who had received antibiotics prior to joint aspiration were performed.21,23 Deirmengian et al21 prospectively evaluated the diagnostic characteristics of synovial fluid alpha defensin in 149 synovial aspirates and assessed the benefits of adding synovial CRP to the diagnostic evaluation. Included in this study population were those diagnosed with PJI, metallosis, and aseptic failure. Of the PJI group, 27% had been given antibiotics prior to diagnostic evaluation. The results demonstrated a high sensitivity and specificity for the alpha defensin test (97.3% and 95.5%, respectively). When combined with the alpha defensin test, the synovial fluid CRP increased the specificity to 100%.21
Shahi et al23 retrospectively reviewed 2 groups of patients with PJI, as diagnosed by the MSIS criteria. One group had been given antibiotics within 2 weeks before the diagnostic workup (aspiration and serologic measurements) and the other group had not been given antibiotics. The results demonstrated no difference in the median alpha defensin levels between those given antibiotics compared with those not given antibiotics. In contrast, median CRP, synovial fluid WBC, and percent of polymorpho-nuclear cells (%PMNs) were significantly lower in the group treated with antibiotics. Furthermore, sensitivity of the alpha defensin test was 100% in the group treated with antibiotics, whereas sensitivities in the same group for ESR, CRP, %PMNs, and fluid culture ranged from 69% to 79%.
As a testament to its potential impact in the field of shoulder arthroplasty, the work of Frangiamore et al24 from 2015 demonstrated the utility of alpha defensin as a predictor of PSI. Samples from patients undergoing revision shoulder arthroplasty (n=30; 33 samples) and those undergoing primary arthroscopic rotator cuff repair (used as negative controls; n=16) were analyzed using the laboratory-based alpha defensin immunosorbent assay. Given the lack of a gold standard for diagnosing PSI, the revision patients were subclassified into a spectrum of infection categories based on clinical, laboratory, and histological criteria (definite infection, n=4; probable infection, n=7; probably contamination, n=1; no evidence of infection, n=21). The results of this study demonstrated that synovial fluid alpha defensin levels in the patients undergoing revision shoulder arthroplasty subclassified as no evidence of infection were not different from the negative control group. However, there was a significant difference between the alpha defensin levels in the infection and no infection groups (P=.006). The alpha defensin test demonstrated a 63% sensitivity, 95% specificity, and an area under the curve of 0.78 for diagnosis of PSI. Based on their results, an alpha defensin value greater than 5 mg/L increased the likelihood of infection by a factor of 12.1, whereas a value less than 0.2 mg/L reduced the likelihood of infection by a factor of 0.38. Finally, the alpha defensin test demonstrated the strongest association with frozen section histology and growth of C acnes.24
Although the alpha defensin test shows great promise in the diagnosis of PSI, there are limitations. There are few studies on alpha defensin for the prosthetic shoulder, and the one highlighted in the article by Frangiamore et al24 demonstrated a lower sensitivity for diagnosing infection in the shoulder vs studies performed on the hip and knee.18,20,21
Furthermore, there is some concern that the rapid, lateral flow test is less sensitive and specific than the laboratory-based test17; the laboratory-based test results are available in 24 hours instead of instantly. The cutoff value to be positive for infection for alpha defensin for diagnosis of a PJI remains undetermined; currently, the quoted range is 5.30 to 7.72 mg/L.21 Finally, an elevated alpha defensin level alerts the surgeon whether there is an infection present. It will not be able to provide information about the causative organism.
The coagulation cascade (both the intrinsic and extrinsic pathways) results in thrombin cleaving fibrinogen resulting in soluble fibrin monomers. These monomers associate to form fibrin polymers that are eventually cross-linked by activated factor XIII to produce an insoluble, cross-linked fibrin clot. With parallel activation of the fibrinolytic pathway, plasmin cleaves these cross-linked fibrin polymers into fibrin degradation products, producing a D-dimer (Figure 2).25
Fibrinogen is clotted by thrombin, and the fibrin monomers that are produced polymerize spontaneously into protofibrils. The tensile strength of the fibrin network is enhanced by factor XIIIa, which cross-links adjacent monomers. Plasminogen activation is enhanced with fibrin formation, and the resultant plasmin digests the individual fibers. Plasmin cleavage between the D and E domains yields (DD)E, the noncovalent complex of D-dimer (DD) and fragment E. Further proteolysis liberates fragment E from DD. [Reprinted from Journal of the American College of Cardiology, 70(19), Weitz JI, Fredenburgh JC, Eikelboom JW. A test in context: D-dimer, 2411–2420, 2017, with permission from Elsevier.]
The D-dimer level was previously used as a screening test for venous thromboembolism but has been abandoned given the low specificity of the test result.26 D-dimer has been used in emergency and intensive care medicine as a predictor of mortality because there have been studies demonstrating an association between high D-dimer levels and multiple organ dysfunction and mortality.27 In the veterinary, rheumatologic, and orthopedic literature, D-dimer has been used in the diagnosis of joint infections.28–31
In the animal world, Ribera et al28 investigated the synovial fluid of septic foals with septic joints, healthy foals with septic joints, septic foals without septic joints, and healthy foals with healthy joints. There was a strong association with the activation of the synovial fibrinolytic pathway in foals with septic joints, as indicated by elevated synovial D-dimer levels, when compared with foals that were septic from a source other than a joint and from healthy controls.28 Although this study examined synovial fluid D-dimer and not serum D-dimer, it does highlight the activation of the fibrinolytic pathway as the result of joint inflammation and infection.
More relevant to the diagnosis of PJI is the prospective study by Shahi et al31 of 5 groups from a total of 245 patients: primary arthroplasty (n=23), revision for aseptic loosening (n=86), revision for PJI (n=57), reimplantation of a prosthetic joint after treatment for a PJI (n=29), and an infection at a site other than a joint (n=50). Patients with systemic inflammatory diseases, patients taking antibiotics, or patients with metallosis were not excluded. All participants had preoperative ESR, CRP, and D-dimer levels drawn. The median D-dimer level was significantly higher (P<.001) for patients with PJIs. The median D-dimer for patients undergoing revision for PJIs was 3.7 and 2.5 times greater than the median D-dimer for patients undergoing revision for aseptic loosening and patients with infections at other sites, respectively. Furthermore, the serum D-dimer outperformed both the serum ESR and CRP in terms of predicting PJI. The sensitivity and specificity of D-dimer were 89% and 93%, respectively, whereas the sensitivity and specificity for ESR and CRP were lower at 73% and 78%, respectively, and 79% and 80%, respectively. Of note, 2 of 5 patients who had positive D-dimer levels and negative levels for both ESR and CRP at the time of re-implantation were found to have positive intraoperative cultures with more indolent bacteria (S epidermidis and C acnes) and went on to failure.31
However, the studies on D-dimer for PJI have not included prosthetic shoulders. As shown by Shahi et al,31 the elevated serum D-dimer was helpful in identifying C acnes and more indolent bacteria, which is promising. Similar to the limitations of the alpha defensin test, the optimal threshold levels must be refined. Encouragingly, the test performed well in patients with systemic inflammatory diseases and those taking immunosuppressive agents.31 Larger and multi-institute studies still need to be performed to determine the influence of concomitant inflammatory arthritis, distant infection, or infection at a site other than the prosthetic joint of interest. As with the alpha defensin test, the D-dimer will not be able to provide information on the causative organism.
Blood Culture Bottles for Periprosthetic Tissue Culture
Another interesting new development in the diagnosis of PSI is placing the intraoperative tissue into standard blood culture tubes (eg, BD BACTEC; BD, Franklin Lakes, New Jersey) for culture. These tubes are widely available and are partially or fully automated with fluorescent sensor detection. This not only reduces the cost and labor hours compared with direct agar plating, it also has been shown to have improved sensitivity.32–34
Hughes et al32 compared 4 culture methods on samples taken at the time of revision surgery. These 4 methods were automated BACTEC blood culture, cooked meat enrichment broth, fastidious anaerobic broth, and direct agar plating. Although all tests were highly specific, ranging from 97% to 100%, there was a large difference in sensitivity. The automated BACTEC had the highest sensitivity at 87%, followed by cooked meat enrichment broth at 83%, fastidious anaerobic broth at 57%, and direct agar plating at 39%.
In a similar prospective, consecutive cohort, Peel et al33 compared cultures of periprosthetic tissue specimens, including prosthetic shoulders, on standard agar, thioglycolate broth, and blood culture bottles for 369 patients. Test sensitivity improved by 47% to 92% with the inoculation of tissues into blood culture bottles from the 62% of conventional agar and broth cultures. The blood culture bottles had a combined sensitivity (aerobic and anaerobic) of 92% and combined specificity of 99.7%, whereas the agar had a combined sensitivity of 48% and combined specificity of 99.7% and the thioglycolate had a sensitivity of 75% and a specificity of 99.5%. In addition, the time to microorganism detection was shorter with the blood culture bottles (21 hours) than with the standard media (41 hours for the aerobic agar, 62 hours for the anaerobic agar, and 65 hours for the thioglycolate broth).33
In a later study, Peel et al34 investigated the laboratory workflow of tissue samples analyzed with blood culture bottles vs the standard agar plating. Their findings suggest that there is a significant cost and labor savings when using the blood culture bottles instead of the standard agar plating (savings of up to $10,000 annually based on a model of the Mayo Clinic work-flow).34 A fundamental limitation of this method, as with standard culture methods, is that it still requires time for the organism to grow. Moreover, if the patient was pretreated with antibiotics, the sensitivity of the test will be affected.
Polymerase chain reaction is a commonly used method to amplify a specific region of DNA. The process uses a thermostable DNA polymerase, Taq polymerase. Polymerase chain reaction works by repeated cycles of heating and cooling to amplify regions of interest on particular DNA segments. With the primers flanking the region of interest, a particular DNA segment (if present) will be amplified exponentially (Figure 3). This segment of interest can then be detected using gel electrophoresis.35 Restriction fragment length polymorphism is the process by which restriction enzymes cut DNA at particular nucleic acid sequences. These fragments are then separated by size with gel electrophoresis and represent specific DNA sequences.36
The DNA sample is heated, and the double helix is denatured into single-stranded DNA. The DNA is cooled in the presence of primers specific to the DNA sequence of the DNA to be amplified. The primers anneal to these segments. The samples are reheated in the presence of Taq polymerase and the regions adjacent to the primers are amplified. With repeated cycles of heating and cooling, a particular segment of DNA will be amplified. (From the National Human Genome Research Institute, www.genome.gov.)
Holmes et al35 took advantage of this science to detect C acnes in surgical biopsy specimens from shoulders and were awarded the 2017 Neer Award for their work. The study group used PCR and RFLP analysis to detect the known and highly conserved 564 base pair amplicon of C acnes 16S rRNA gene. It combines the sensitivity of PCR and the specificity of RFLP. Furthermore, the time to diagnosis compared with standard anaerobic cultures is much faster (24 hours vs 6 to 7 days). The primers used in this study were able to distinguish between different subspecies of the Cutibacterium species; the RFLP cleavage site is a unique site in the amplicon that was specific to C acnes relative to other subspecies.
This method does have several limitations in the diagnosis of PSI. Currently, there are limited studies in shoulder PJIs using this technique. Polymerase chain reaction can only detect 1 organism at a time, although this has been partially mitigated by multiplex and broad-range PCR.37,38 DNA from dead bacteria can be identified, making diagnosis of active infection after a course of antibiotics difficult. Furthermore, the test does not have the ability to distinguish pathologic vs nonpathologic strains. Another limitation of concern is the reported high false-positive rates. An extreme example by Jacovides et al38 demonstrated an 88% false-positive rate. This may be due to contamination from the sample prior to or during transport of the sample to the laboratory as PCR can detect as few as 10 cells.35
Next generation sequencing is a catch-all term for a variety of new techniques that allow rapid and relatively inexpensive sequencing compared with the traditional Sanger technique, which relied on chain-termination using dideoxynucleotides.39 This process, although efficacious, is most useful for short sequences because it is more time consuming and expensive than newer techniques. The process of NGS starts with quantitative PCR to determine the bacterial burden present in the sample. The amplicons from the PCR are then all sequenced. These sequences are then compared with the NIH/Genbank database, with a 90% agreement needed.40
Two case reports outside of the orthopedic literature include the diagnosis of culture-negative endocarditis and neuroleptospirosis using this method. Both of these cases involved patients who were taking antibiotics prior to presentation at the hospital.41,42 In the endocarditis case, the organism was found to be Abiotrophia defectiva, a fastidious organism and variant of streptococci.41 These cases demonstrate the utility of NGS for uncommon pathogens, as well as cases in which antibiotics have already been given, a common scenario in the presentation of PSI.
Tarabichi et al40,43 performed 2 prospective studies evaluating the role of NGS for PJI. The first study, published in January 2018, involved collecting samples from patients undergoing primary and revision hip or knee arthroplasties.43 Samples were classified as either positive or negative according to the MSIS criteria. Of the positive group, 68% were culture positive and 89% were NGS positive. Interestingly, of the 39 MSIS negative group, 10 (26%) were NGS positive and only 1 (3%) was culture positive. The sensitivity and specificity of detecting bacteria using NGS was 89% and 73%, respectively, whereas it was 61% and 97% for culture, respectively. However, when detecting a single organism (representing >59.5% of the bacteria present), the sensitivity of NGS decreased to 71% and the specificity increased to 95%.43
In the second study, published in February 2018, synovial fluid samples were collected from 86 patients undergoing aspiration of the hip or knee as part of the workup for a suspected PJI.40 Participants were divided into 3 groups: group I (n=30) was the alpha defensin and culture-positive group; group II (n=24) was the alpha defensin positive and culture-negative group; and group III (n=32) was the alpha defensin and culture-negative group. Next generation sequencing detected an organism in 96.1% of the samples for group I; in 3 of these, NGS was able to detect the mecA gene for S aureus, whereas the standard culture was not. Furthermore, for groups I, II, and III, NGS detected several organisms in a sample, as well as several species of fungi and less virulent organisms (C acnes), that the standard cultures did not detect.40
Recently, Namdari et al44 evaluated the correlation between NGS and routine cultures in revision shoulder arthroplasty. Tissue samples from 44 patients undergoing revision shoulder arthroplasty for any reason were taken for cultures and for NGS. The most commonly identified cultured bacteria were C acnes (56.5%) and CoNS (39.1%), and the most commonly identified bacteria by NGS were C acnes (70.6%), Acinetobacter radioresistens (35.3%), and CoNS (29.4%). The concordance (kappa) between using culture in the criteria and NGS in the criteria for defining infection was 0.333, which was considered fair.44 In this study, more than 90% of NGS results were polymicrobial,44 similar to what was found by Tarabichi et al.40
Because NGS is able to generate thousands of sequences from a single tissue sample, it gives information on all the organisms that inhabit the shoulder joint. Thus, a drawback of NGS is determining which organisms are pathologic and which organisms are commensal. As with PCR, there is a risk of a high false-positive rate associated with NGS. More research is needed to understand the shoulder's natural microbiome before the results of NGS become truly meaningful.