➢ Periprosthetic shoulder infections often differ from those of the hip and knee because of the indolent nature of the offending organisms in most shoulder infections.
➢ Current guidelines for the diagnosis and treatment of periprosthetic joint infections of the hip and knee are less effective when applied to periprosthetic shoulder infections.
➢ Due to the indolent nature of many periprosthetic shoulder infections, especially those associated with Propionibacterium acnes (P. acnes), preoperative diagnostic laboratory values (erythrocyte sedimentation rate [ESR], C-reactive protein [CRP] levels, and serum interleukin-6 [IL-6] levels) may not be elevated.
➢ Diagnosis should be based on clinical suspicion, patient factors, and intraoperative findings when the results of perioperative testing are equivocal.
➢ Empiric antibiotic therapy for suspected infection after operative intervention should be considered to avoid delay in treatment, as P. acnes may take up to twenty-one days to grow on culture medium.
➢ Newer diagnostic studies, such as synovial fluid analysis, lowered thresholds for frozen-section analysis, culturing of P. acnes variants, polymerase chain reaction (PCR) analysis, and implant sonication hold potential for increasing accuracy and sensitivity in the diagnosis of indolent infections.
One of the most serious complications following shoulder arthroplasty is periprosthetic infection. Infection is associated with poor outcomes, increased cost, and technically difficult revision surgery. The prevalence of infection following primary shoulder arthroplasty has been reported to be between 0.7% and 4%, and even higher following revision surgery1-6. Two recent, large retrospective reviews noted infection rates of 0.7% (nineteen of 2540 cases)2 and 1.2% (thirty-two of 2588 cases)6. Bohsali and colleagues reported that deep periprosthetic infections accounted for 4.6% of all complications of total shoulder arthroplasty2. At the time of long-term follow-up, Singh et al. observed five, ten, and twenty-year periprosthetic infection-free rates of 99.3%, 98.5%, and 97.2%, respectively6.
Despite the low incidence of infection following shoulder arthroplasty, periprosthetic infection continues to be an important concern for hospitals, surgeons, and patients. The number of shoulder arthroplasties performed in the United States is growing faster than ever. According to the Agency for Healthcare Research and Quality Database, more than 53,000 patients underwent shoulder arthroplasty in 2011, with a projected increase to 75,000 per year by 20207. Despite adequate preventative measures, some of these patients will develop a deep periprosthetic infection requiring extensive medical and surgical management8. The treatment algorithms for established infection after shoulder arthroplasty are similar to those after hip and knee replacements, and include irrigation and debridement with one or two-stage exchange, resection arthroplasty, antibiotic suppression, and arthrodesis.
Most data on periprosthetic infection are derived from the extensive hip and knee literature, with much less information reported for the shoulder. Hence, the established algorithms for the diagnostic tests are based on lower-extremity-infection cohorts. While similarities exist, certain aspects of periprosthetic shoulder infection are unique and require specific attention. One of the most important differences relates to the variation in offending organisms, with an increased prevalence of less virulent organisms in the shoulder, particularly Propionibacterium acnes (P. acnes)9-11. As a consequence of the unique characteristics of shoulder infections, the standard preoperative and intraoperative diagnostic tests as well as the criteria for infection diagnosis used for the hip and knee may not be as relevant to the shoulder. The purpose of this review is to examine the current and future tools for diagnosis of infection and discuss their relevance to periprosthetic infection after shoulder arthroplasty.
A comprehensive analysis of the diagnosis of infection following shoulder arthroplasty requires a basic understanding of the offending organisms and the mechanism of infection around the prosthetic shoulder. While both coagulase-negative Staphylococcus species (CNSS) and the more serious Staphylococcus aureus (S. aureus) are common offending organisms, the high prevalence of P. acnes is unique to the shoulder10. Singh and colleagues isolated S. aureus, CNSS, and P. acnes in 31% (ten), 16% (five), and 19% (six) of thirty-two cases of periprosthetic shoulder infection, respectively6. Interestingly, they noted a trend toward an increased prevalence of P. acnes, almost equaling the prevalence of the Staphylococcus species, in the more recent years. Indeed, in a recent study from our institution, of forty-eight patients who underwent revision shoulder arthroplasty and had positive intraoperative cultures between 2002 and 2009, 38% (eighteen) were infected with P. acnes compared with 35% (seventeen) infected with CNSS and 13% (six), with S. aureus12. Piper et al. noted a similar distribution for patients categorized as having definite periprosthetic shoulder infection: 39% (seven) of eighteen were infected with P. acnes; 28% (five), with CNSS; and 22% (four), with S. aureus4. This trend toward increased rates of P. acnes infections is likely related to the increased awareness of the organism as a true pathogen and of its longer incubation time in cultures. Previously thought to be a culture contaminant because of its indolent nature and identification as part of normal skin flora, P. acnes is now recognized as a pathogenic organism capable of leading to devastating complications in the shoulder13-17. Since P. acnes often requires longer culture times (ten to twenty-one days) to grow compared with other pathogenic bacteria15, microbiology laboratories that discard culture plates after a few days may fail to identify a P. acnes infection. This may lead to implant failure being incorrectly attributed to aseptic loosening.
P. acnes is a relatively slow-growing, anaerobic, gram-positive organism. As noted, it is considered part of the normal skin flora in adults, with particular prevalence in areas with a high density of sebaceous glands and hair follicles, such as the axilla. An increased prevalence of P. acnes infections in the shoulder compared with the hip and knee correlates with colonization density of these anatomic locations10. While patients with clear signs of infection and growth of P. acnes on culture are clearly considered to have periprosthetic infection, the literature has yet to conclusively define the diagnosis for patients who have positive P. acnes cultures after shoulder arthroplasty but no other clinical signs of infection3,12,18. The slow-growing nature of P. acnes and its propensity for the shoulder may affect the sensitivity of current diagnostic tests. While it is clear that P. acnes can cause clinically relevant infections in the shoulder, its course may differ from that of the other frequently encountered organisms. This indolent course may translate to a muted response to preoperative and intraoperative diagnostic testing, such as measurement of the erythrocyte sedimentation rate (ESR) and C-reactive protein (CRP) level19, fluid and tissue cultures4, and frozen-section analysis20. Because of the increasing recognition of P. acnes as an offending pathogen, and because of the challenges it presents for diagnosis, recent investigators have examined the efficacy of prophylactic antimicrobial measures in reducing the prevalence of P. acnes infections, although consensus recommendations are not currently available21,22. Recently, a unique hypothesis of a relationship between P. acnes and osteoarthritis has been proposed23. Levy and colleagues found a high prevalence of P. acnes (42%; twenty-three of fifty-five) during primary shoulder arthroplasty in arthritic joints and proposed a potential role for P. acnes in glenohumeral arthropathy23. This was an uncontrolled study, however, and further investigation is needed to clarify these findings.
Despite differences in clinical presentation between the offending species, S. aureus, P. acnes, and CNSS have a shared overlap in the mechanism of orthopaedic implant bacterial adherence. These organisms are all capable of forming microcolonies surrounded by hydrated polymeric matrices of their own synthesis along the prosthetic surface. These complex bacterial structures are termed biofilms, and are recognized to play a key role in the pathogenesis and resistance to diagnosis and treatment of periprosthetic joint infections. These bacteria are able to convert from free-floating, planktonic organisms into sessile communities of slowly dividing bacteria along an implant surface, which are resistant to antibiotics and host defenses and thus difficult to detect by traditional fluid and tissue cultures24. Many of the experimental and future diagnostic tests discussed in this review represent attempts to address these biofilm-related issues.
Clinical manifestations of periprosthetic shoulder infections are often less overt than those of infections after knee and hip arthroplasty16. This has been attributed to the extensive vascular-rich soft-tissue environment around the shoulder and the indolent nature of the offending organisms, particularly P. acnes25. Since clinical symptoms of local or systemic infection may be subtle, a high degree of suspicion is indicated when there are unexpected features in the history, on physical examination, or on radiographic studies, such as unexplained shoulder pain within the first few years after surgery and/or component loosening. It is crucial to obtain a thorough history of the postoperative course following a shoulder replacement, including wound drainage, antibiotic use, preoperative risk factors, and duration of pain relief before the recurrence of symptoms. The general medical health of the patient is also an important consideration, since periprosthetic shoulder infections occur more commonly in patients with chronic systemic illness or those with a decreased immune response and/or immunosuppression1-3,5,14,26-29.
Periprosthetic shoulder infections have been classified on the basis of the timing of presentation after surgery. This classification system refers to deep wound infections. Superficial infections are of less concern; they typically occur during the early postoperative period and can be treated successfully with oral antibiotics. Deep infections are classified as acute when they occurred less than three months after surgery, subacute when they occurred between three and twelve months following surgery, and late or chronic when they occurred more than one year following surgery2,14,30. In the hip and knee literature, authors have utilized a comprehensive periprosthetic infection classification based on the time of onset following surgery: type 1 indicates the presence of positive cultures at the time of revision arthroplasty in a patient in whom infection was not suspected on the basis of clinical findings; type 2, an acute infection detected within thirty days after the index arthroplasty; type 3, an acute hematogenous infection that occurred at any time; and type 4, a chronic infection31. This classification system does not account for potential differences in the presentation of periprosthetic shoulder infections, and a modified version reflecting the differences would be beneficial for guiding further treatment. The majority of periprosthetic infections in the shoulder, however, are deep chronic infections that developed at the time of the primary arthroplasty. These patients will often state that their shoulder never felt right postoperatively and that they never had a period of symptomatic relief. Conversely, patients presenting with acute-onset infection typically report a sudden increase in pain level or decrease in function after a relatively normal initial postoperative course. The majority of these infections are attributed to seeding from other areas in the body, and the source should be identified before further surgical planning is considered. Such patients may have good pain relief after the initial surgery but later develop atraumatic, otherwise unexplained new-onset shoulder pain.
When symptoms are present they can be variable, but the most common and prominent clinical complaint is shoulder pain. This pain is often generalized, usually constant, and worst at night. Although the pain is present at rest, it is often worse with use of the shoulder. This mechanical or activity-related pain is often associated with component loosening. In addition to pain, stiffness is common with an infection at the site of a shoulder arthroplasty1,14. Patients may have difficulty gaining an adequate range of motion postoperatively, and their attempts at gaining motion can further exacerbate pain and increase stiffness despite their continued effort in postoperative rehabilitation. Range of motion, both active and passive, will elicit stiffness, which is typically in all planes. Pain at the terminal range of motion is often present. Since the majority of periprosthetic shoulder infections are low-grade or subclinical, patients rarely present with fevers, chills, swelling, purulent drainage, a sinus tract, or sepsis. If present, these features are typically associated with a more aggressive pathogen, such as S. aureus. However, a non-blanching red-to-brown shaded rash around the axilla or around the shoulder area is often associated with a P. acnes infection. Hematoma formation following shoulder arthroplasty has also been associated with positive cultures and the development of a deep infection, although use of drains postoperatively has not been proven to decrease the risk of infection in these patients32.
Physical examination of the shoulder is also performed to rule out other causes of pain, such as rotator cuff pathology, nerve injuries, dislocation, or component loosening. After the initial examination of the surgical incision site, attention should be paid to the contour of the shoulder and any evidence of muscle atrophy, as this may indicate rotator cuff pathology or nerve injury. Palpation of the glenohumeral joint as well as the acromioclavicular joint can also help differentiate the source of pain. Strength examination should also be performed to determine the status of the rotator cuff muscles. Finally, the cervical spine and elbow should be examined to rule out other extrinsic causes of pain.
Serum and Synovial Fluid Evaluation
As with periprosthetic infections in other joints, additional diagnostic workup should include measurement of serum CRP levels and the ESR as well as joint aspiration for synovial fluid analysis. A meta-analysis by Piper et al. showed measurement of serum CRP levels and ESR to have sensitivities of 42% and 16%, respectively, for diagnosing shoulder infections, compared with 88% and 75% for diagnosing lower-extremity infections19. This discrepancy is related to the mild inflammatory response elicited by P. acnes and must be taken into account when applying guidelines for diagnosis of periprosthetic joint infection, which were based on lower-extremity studies. Synovial fluid cell counts have commonly been used to detect the presence of periprosthetic infection. While white blood cell (WBC) counts of >50,000 cells/mm3 and polymorphonuclear neutrophil (PMN) percentages of >75% are highly suggestive of infection11, lower thresholds do not necessarily rule out the possibility of infection. Cutoff values as low as >1100 cells/10−3 cm3 for WBC count and >64% for PMN percentage have been used to diagnose periprosthetic infection in patients undergoing revision knee arthroplasty33, although similar lowered thresholds have not been proposed for the shoulder. In a study of revision hip arthroplasties, Schinsky et al. also found that elevated WBC count and PMN percentage provided high specificity and sensitivity when combined with elevated serum CRP level and ESR34. A work group of the Musculoskeletal Infection Society highlighted the importance of synovial cell count analysis by incorporating WBC count and PMN percentage into their new definition of periprosthetic joint infections33.
There are unique challenges with joint aspiration and synovial fluid analysis for the shoulder when compared with such analyses for the hip and knee. A decreased synovial inflammatory response typically results in less synovial fluid production; thus, attempts to aspirate shoulder synovial fluid preoperatively are frequently unsuccessful. Sperling et al. noted that preoperative aspiration was possible for only 56% of their patients who had an infection following shoulder arthroplasty14. Although infection was identified in the majority of these cases, most were CNSS infections and <30% were due to P. acnes14. In the series reported by Codd et al., joint aspiration could be obtained from only 38.8% of the shoulders, and pathogens were identified in only 29% of those aspirates35.
Synovial fluid cytokines and other inflammatory proteins show some potential for use in early diagnosis of periprosthetic infection, but validation with larger cohorts and more efficient sample acquisition methods is still needed to determine their value in diagnosing infections after shoulder arthroplasty.
In the acute infection phase, radiographs of the shoulder demonstrate unremarkable findings. However, when a patient has a chronic infection, radiolucencies around one or both components or gross loosening of the implants may be seen. Therefore, in the absence of a clear reason for aseptic loosening, a change in radiolucent lines, osseous resorption, periosteal reaction, or component loosening seen on radiographs is highly suggestive of a chronic infection36. Our clinical experience has shown that an otherwise unexplained finding of humeral stem loosening within the first five years following surgery should raise a high suspicion for indolent or chronic infection.
Ultrasonography or magnetic resonance imaging (MRI) with intravenous contrast medium can be used to identify loculated fluid collections and track contiguous spread of the infection or osteomyelitis, but artifact from the prosthesis may distort the images11. Computed tomography (CT) may not provide any further information to facilitate the diagnosis of a periprosthetic infection, but it may detect more subtle signs of component loosening or shift not seen on radiographs. Radionuclide studies are another imaging method for the diagnosis of infection. Technetium-99m sulfur colloid bone marrow scintigraphy combined with indium-111-labeled WBC scanning has been considered to be the imaging gold standard for diagnosis of some periprosthetic infections; however, use for diagnosing shoulder infections has produced mixed results1,30,37-39. In addition, the in vitro labeling process is labor intensive, is not always available, and involves direct handling of blood products. Recently, the utility of 18F-labeled fluorodeoxyglucose-positron emission tomography (FDG-PET) in diagnosing periprosthetic hip infection was evaluated in a multicenter study40. FDG-PET is less time-consuming than technetium-99m sulfur colloid bone marrow scintigraphy combined with indium-111-labeled WBC scanning and does not entail any invasive procedures. PET findings suggestive of an infection were established as increased uptake at the stem-prosthesis interface41, and the test was reported to have a sensitivity of 85% with a specificity of 93% for diagnosis of periprosthetic hip infections. We are not aware of any studies that evaluated FDG-PET in the context of shoulder infections. Overall, advanced imaging modalities, such as MRI, technetium-99m bone scanning, and indium-111-labeled WBC scanning, have not been as reliable for diagnosis of infection after shoulder arthroplasty as they have been for other periprosthetic infections1,30,38,39.
Equivocal results of a preoperative workup may require the surgeon to depend on intraoperative findings to determine a definitive surgical plan. Gram staining has commonly been performed on operative tissue specimens as a means of screening for the presence of infection, but several studies have raised questions about its value42-44 and, due to low sensitivity, this test is no longer recommended during revision arthroplasty. A possible exception is for a known infection, for which recognition of gram-positive versus gram-negative species may help guide the immediate choice of antibiotics.
Frozen-section histological analysis appears to be a sensitive and specific tool when used in an algorithmic approach for diagnosing periprosthetic infections in the hip and knee38,45. Sensitivity and specificity values ranging from 77% to 95% and 92% to 96%, respectively, have been reported46,47. For example, Stroh et al. reported a 97.7% rate of concordance between diagnoses of infection based on frozen sections and those based on permanent sections in patients undergoing revision total knee arthroplasty48. The efficacy of frozen-section interpretation for diagnosis of infection in patients with shoulder arthroplasty is less clear. One study showed a 92% rate of negative intraoperative histologic evaluations for patients with culture-positive infection18. Another report demonstrated no correlation between culture-positive infection and the results of frozen-section histological analysis, with the exception of S. aureus infection49. A retrospective review with reevaluation of all microscope slides demonstrated that frozen-section analysis had a sensitivity of 61% for diagnosis of infection in revision shoulder arthroplasty on the basis of criteria endorsed by the American Academy of Orthopaedic Surgeons (AAOS) for diagnosing periprosthetic infections of the hip and knee20. However, when a lower threshold of ten PMNs, in total, in the five most densely inflamed high-powered fields was applied to the same group of patients, sensitivity was increased to 72% while 100% specificity was maintained20. A larger prospective study with frozen sections reviewed by multiple pathologists is necessary to verify these results and further evaluate this lower threshold.
At the time of revision surgery, the surgeon must make intraoperative decisions regarding prosthesis removal or retention, usually before culture results become available. Furthermore, extended culture incubation of up to twenty-one days may be necessary to maximize the recovery of P. acnes from tissue specimens15,49,50. Pottinger et al. sought to correlate several preoperative and intraoperative factors with positive intraoperative cultures for P. acnes in an effort to predict culture results and help guide treatment51. Their analysis revealed that male sex, radiographic evidence of humeral component loosening, membrane formation, and cloudy fluid were each independent and significant predictors of positive P. acnes cultures for patients undergoing revision shoulder arthroplasty for presumed aseptic failures. Whether unexpected positive cultures represent a definite infection or a contamination remains a diagnostic dilemma52. In one study, the rate of unexpected positive cultures was as high as 56%52. This high rate of positive P. acnes cultures appears to correlate with the duration of culture observation and the number of cultures of specimens obtained during surgery. Foruria et al. found unexpected positive cultures for 15% of patients in whom revision was performed with no previous suspicion of infection53. Of these patients, 10% resulted in symptomatic infection requiring revision arthroplasty and >25% had no clinical relevance. Grosso et al. found that only 5.9% of patients with unexpected positive cultures had a symptomatic recurrence of infection within twenty-four months12. These findings suggest that intensive antimicrobial therapy may not be necessary to reduce the risk of recurrent infection in the presence of positive intraoperative cultures without clinical signs of infection, particularly if all components and foreign material were removed at the time of revision surgery and bone cement containing antibiotics was employed. However, additional data are needed to guide decision-making.
Additional Studies to Consider
A number of potential serum and synovial biomarkers are currently being evaluated for use as preoperative diagnostic tests for periprosthetic infections. In particular, tests of serum and synovial fluid interleukin-6 (IL-6) levels have shown promising results in studies of total hip and knee arthroplasty54,55. A study by Frangiamore et al. showed promising results for measurements of synovial fluid IL-6 levels for the diagnosis of periprosthetic shoulder infection56. A number of other cytokines, including IL-8 and IL-1β, identified in periprosthetic synovial fluid may also prove to be useful in the diagnosis of infection55,57. Larger prospective studies are necessary to explore these synovial fluid markers.
A method proposed to increase the sensitivity of microbiological cultures is to sonicate excised implants and then culture the sonicate fluid58. The process of implant sonication involves the use of ultrasonic pressure waves to dislodge biofilms from the surface of implants into culturable medium in an attempt to increase the sensitivity of microbiological studies4. For example, we have found that sonication for one to five minutes increases the sensitivity of either microbiological cultures or polymerase chain reaction (PCR) to detect biofilm-formative bacteria adherent to metal implants59. Piper et al.4 found sonication cultures to be more sensitive than standard tissue cultures for the detection of periprosthetic shoulder infections (66.7% and 54.5%, respectively; p = 0.046). Sonication equipment in the microbiology laboratory is inexpensive, but the work flow required to transmit large implants in clean, sterile containers from the operating room to the microbiology laboratory requires close collaboration between the operating room facilities and the microbiology laboratory. Additional studies are needed to determine if this is a cost-effective test with respect to shoulder implants.
The use of PCR for detection of bacterial rRNA as an indicator of the presence of bacteria at the sites of prosthetic joints has also been tested in revision hip and knee arthroplasty60-62. The sensitivity and specificity of broad-range PCR from synovial fluid for the diagnosis of periprosthetic infections have been reported to range from 50% to 92% and 65% to 94%, respectively63-65. Achermann et al. evaluated a real-time multiplex PCR test for the diagnosis of periprosthetic infection from sonication fluid, but did not find PCR to be significantly more sensitive than culture of sonication fluid66. Other studies have demonstrated higher sensitivity for several different variations of PCR testing67-69. However, because the majority of studies have shown equivocal results in the setting of revision arthroplasty, especially with respect to the clinical relevance of PCR-positive culture-negative cases, PCR is not widely accepted as a standard of care in diagnosing periprosthetic infection.
Future studies focused on identifying the pathogenicity of P. acnes subtypes could help delineate true infection from contaminants.
Infection after shoulder arthroplasty remains a diagnostic challenge due to the indolent host response to the major offending organism, P. acnes, and its subtle clinical presentation. We recommend an evaluation of several patient factors and perioperative diagnostics when evaluating patients for periprosthetic shoulder infection (Table I). Comorbid conditions, immune status, history of infection, radiographic findings, and the results of standard serum and synovial fluid preoperative testing are important to evaluate preoperatively. Intraoperatively, multiple tissue specimens should be obtained for frozen-section histological analysis and culture, and clinical suspicion should be raised by the presence of cloudy fluid and membrane formation when the results of other tests are negative. Empiric treatment with antibiotics should be considered if the diagnosis of infection is still unclear at the time of revision arthroplasty, as tissue culture results may take up to twenty-one days to become positive for P. acnes infection. A lower threshold for intraoperative frozen-section analysis, study of synovial fluid markers, evaluation of pathogenic strains of P. acnes, and implant sonication all have the potential to increase the efficacy of diagnosis of periprosthetic infections. Further work is needed in these areas to determine efficacy, as well as in the development of guidelines for diagnosis and management of periprosthetic infections specific to the shoulder.
Source of Funding: No external funding was utilized for this investigation.
Investigation performed at the Orthopaedic and Rheumatologic Institute, Cleveland, Ohio
Disclosure: None of the authors received payments or services, either directly or indirectly (i.e., via his or her institution), from a third party in support of any aspect of this work. One or more of the authors, or his or her institution, has had a financial relationship, in the thirty-six months prior to submission of this work, with an entity in the biomedical arena that could be perceived to influence or have the potential to influence what is written in this work. No author has had any other relationships, or has engaged in any other activities, that could be perceived to influence or have the potential to influence what is written in this work. The complete Disclosures of Potential Conflicts of Interest submitted by authors are always provided with the online version of the article.
- Copyright © 2013 by The Journal of Bone and Joint Surgery, Incorporated