With the aging of the population and the growing incidence of injuries, the number of individuals receiving joint prosthesis replacement and internal fixation is increasing sharply. Currently, approximately 800,000 prostheses are implanted and 2,000,000 fracture fixations are performed annually in the United States, with infection rates of approximately 2% and 5%, respectively.1–3 It is estimated that periprosthetic joint infection (PJI) costs the US health care system $1.6 billion annually, resulting in an economic burden and leading to higher rates of disability and mortality.2,4 Effective treatment of implant-associated infections relies on their early identification. However, this is sometimes difficult owing to multiple factors, including low virulence of bacteria, non-typical clinical signs and symptoms, and confounding postoperative status.2,5 To facilitate the diagnosis of implant-associated infections, the Musculoskeletal Infection Society and the International Consensus Meeting have devised an integrated criterion including serology, microbiology, histology, and physical examination.6
However, the Musculoskeletal Infection Society criterion has limited sensitivity and specificity in distinguishing PJI from aseptic inflammation.7 In contrast, novel serum biomarkers such as interleukin-6 (IL-6), leukocyte esterase (LE), and defensins produced in the host response against infection have been shown to have promising diagnostic accuracy. Additionally, synovial fluid biomarkers have also been useful in the diagnosis of local infection related to implants. Despite the discovery of new biomarkers and testing methods, their true diagnostic values remain inconsistent and under debate. Therefore, it would be beneficial to review and summarize the current understanding and related clinical studies for clinicians.
In this article, the authors discuss different kinds of serum and synovial biomarkers for the diagnosis of implant-associated infections (Figure 1). Their biological origin, clinical use, and strengths and weaknesses are reviewed.
As a plasma protein produced by hepatocytes and regulated by IL-6,8 C-reactive protein (CRP) is essential in activating the complement system of innate immunity. In clinical practice, CRP is considered one of the most critical indicators of inflammation. Thus, CRP has been widely used to monitor the body's acute-phase responses, such as trauma, infection, and connective tissue disorders.9 Serum CRP has a high sensitivity (more than 90%) in the diagnosis of bone infection. However, its specificity ranges from 20% to 80%.10 Currently, both the American Academy of Orthopaedic Surgeons and the Musculo-skeletal Infection Society recommend that CRP and erythrocyte sedimentation rate (ESR) be used as screening tools in the diagnosis of PJI.6,11 However, screening for infection with CRP has several disadvantages. First, the specificity of CRP is weak because it also responds to auto-immune disorders and trauma (including surgery) apart from infection. Second, in diagnosing low virulent infections, the sensitivity of CRP is also disappointing. It was reported that the false-negative rate of CRP combined with ESR exceeded 10%, with a sensitivity of lower than 50%, in the diagnosis of Propionibacterium acnes infection of shoulder prostheses.12,13 Finally, patient demographics (age, sex, antibiotics, and disease status) along with testing methods will also affect CRP levels.13,14
As intra-articular CRP level is related to local complement activation and phagocytosis, the diagnostic value of synovial fluid CRP is also being investigated.15,16 It was postulated that intra-articular CRP originated from serum CRP, which diffused into the joint when there was a change in synovial permeability under inflammation.15,16 In 2012, Parvizi et al16,17 first reported that synovial CRP had better accuracy than serum CRP and could achieve a sensitivity of 85% and a specificity of 95% in the diagnosis of periprosthetic knee infection when using a cutoff value of 9.5 mg/L. In a meta-analysis containing 6 studies, synovial fluid CRP was shown to provide higher accuracy than serum CRP, with pooled sensitivity and specificity of 92% and 90%, respectively.18 Nonetheless, there were also studies indicating that the differences in diagnostic values between synovial fluid CRP and blood serum CRP were not statistically significant.15,19 Moreover, extracted synovial fluid could also become too viscous or hemolytic for analysis. These studies were limited by size and immunoassay methods.15,19,20
In general, serum CRP remains one of the simplest and most effective methods to screen for PJIs. The usefulness of synovial fluid CRP is unclear; however, it may have limitations similar to serum CRP because of the nature of CRP as a marker of inflammation rather than local infection.
Produced by C cells in the thyroid gland, procalcitonin (PCT) is a precursor of the hormone calcitonin and is stimulated by pro-inflammatory stimuli, especially bacteria.21 Generally, PCT is upregulated within 2 to 4 hours after infection, peaks in 6 to 24 hours, and lasts for 7 days.22 Because of its high level in systemic bacteremia, PCT has been recommended for use as an indicator in monitoring sepsis.23,24 In orthopedics, PCT is used to identify infection under inflammatory, rheumatic, or post-surgery conditions, and the level of PCT is unaffected by immunosuppressants.21,25–28 Despite the promising traits of PCT, its diagnostic accuracy is undesirable.29 At a cutoff value of 0.3 ng/mL, Bottner et al30 showed that serum PCT had a low sensitivity of 33% and a high specificity of 98% in the diagnosis of periprosthetic knee or hip infection. More recently, at a cutoff value of 0.025 ng/mL, Ettinger et al5 demonstrated a high sensitivity of 80% and a low specificity of 37%. In a meta-analysis consisting of 6 studies, the pooled sensitivity and specificity regarding serum PCT were 53% and 92%, respectively.31 Most researchers have concluded that the diagnostic value of serum PCT is inferior to that of CRP or alpha-defensin and that it should not be used as a single test in the diagnosis of PJI.32–35 Regarding synovial fluid PCT, the number of studies is limited. At a cutoff value of 0.5 ng/mL, Saeed et al36 concluded that the sensitivity and specificity of synovial PCT in the diagnosis of septic arthritis were 88% and 57%, respectively. At the same cutoff value, Wang et al37 demonstrated that synovial PCT had both high sensitivity and specificity of 87% and 94%, respectively, in the diagnosis of septic arthritis (including prosthetic joints), outperforming serum PCT.33 However, Wang et al37 excluded patients who underwent joint surgery, which may have led to variability in these results.
In summary, it remains unclear whether the diagnostic value of synovial fluid PCT is better than that of serum PCT.31 Based on current evidence, it is unlikely that serum or synovial PCT will become an ideal biomarker for the diagnosis of PJI. However, PCT might be useful for diagnosing infection in patients with rheumatoid, inflammatory, or surgery conditions, for which CRP is less reliable, and for screening suspected cases at initial evaluation because of its high sensitivity.25,28,38,39
Leukocyte esterase is a special enzyme secreted by neutrophils recruited locally in response to bacterial infection and has been investigated for detecting infections of the urinary tract, peritoneum, and various bodily fluids.40 In orthopedics, the role of synovial fluid LE in distinguishing septic arthritis or periprosthetic infection from aseptic conditions has been explored.41,42 To measure this intracellular enzyme, a colorimetric strip test was developed, which included using a detergent for lysis of synovial polymorphonuclear cells, leaking of LE, and combining LE with dye to yield a purple color.40,43 The intensity of color change (read as negative, trace, +, or ++) directly reflects the number of neutrophils. In 2011, Parvizi et al40 reported that LE had a sensitivity and a specificity of 80.6% and 100%, respectively, at a ++ reading for the diagnosis of knee arthroplasty infection. Later studies showed that the sensitivity and specificity of LE for the diagnosis of PJI ranged from 66% to 100% and from 86.6% to 100%, respectively.40,44–47 In 2015, a meta-analysis revealed that the pooled sensitivity and specificity of LE were 81% and 97%, respectively.48 Because of its relatively high specificity and negative predictive value, some have concluded that LE would be a good marker for ruling out aseptic conditions in patients suspected of having infection, especially under trace or negative conditions.44,45,47 Other advantages of LE include its relatively low cost (US $0.17 per test) and its rapid detection (<2 minutes).47,48 The disadvantages of LE mainly relate to the colorimetric testing method. First, viscous aspirate or too much cell debris can render an LE test result un-readable, and this can affect up to one-third of all samples.44,46 Second, because neutrophils exist in many inflammatory conditions, false-positive results can be observed with conditions such as aseptic loosening or inflammatory diseases.45,49–51 Finally, LE has poor accuracy in detecting periprosthetic infection of the shoulder.52
In summary, LE is a quick, efficient, and cost-effective test with good accuracy that is worth further exploration.
Interleukin-6 is a cytokine produced by almost all stromal cells and immunocytes.53 It plays an essential role in promoting plasma cell differentiation and eliciting hepatic acute-phase responses, which will cause high fever, anemia, and release of CRP.54–56 The normal serum IL-6 level is approximately 1 to 5 pg/mL, elevating to 30 to 430 pg/mL during menstruation, with cancer, and after surgery.53,57 Following arthroplasty implantation without infection, serum IL-6 reaches its peak in 6 to 12 hours and drops below 50 pg/mL in 48 hours. Because it responds and declines more quickly than CRP or ESR, IL-6 is considered a better indicator of postoperative inflammation.58,59 However, IL-6 does not perform better than traditional bio-markers in the diagnosis of PJI. In 2005, Di Cesare et al57 first reported that, using a cutoff value of 10 pg/mL, serum IL-6 had a sensitivity and a specificity of 100% and 95%, respectively, for diagnosing PJI, outperforming ESR and CRP. In studies concerning the diagnosis of periprosthetic hip or knee infection, the sensitivity and specificity of serum IL-6 vary greatly, ranging from 36% to 100% and from 68% to 95%, respectively.5,30,32,34,57,60,61 The difference in these results, as some have stated, is partly because of the lack of a “gold standard” to diagnose PJI, the rather small number of included patients, and changes in cutoff value and testing method.60,61 In a meta-analysis containing 3 early studies, Berbari et al62 showed that serum IL-6 was more accurate than CRP, ESR, and peripheral white blood cell count. However, later clinical studies showed that IL-6 could not outperform CRP in sensitivity or specificity.5,32,34,60,61 Xie et al63 concluded in a meta-analysis that the pooled sensitivities of serum and synovial fluid IL-6 were 72% and 91%, respectively, and that the pooled specificities of serum and synovial fluid IL-6 were 89% and 90%, respectively. The combination of serum CRP and IL-6 demonstrated a sensitivity of 57% to 100% and a specificity of 68% to 100%.30,32,60,61 Serum IL-6 also had a low sensitivity of 12.5% to 14% in the diagnosis of indolent shoulder infection caused by P acnes.12,59 Further, the level of serum IL-6 could be affected by inflammatory conditions such as sepsis, trauma, and arthritis.57,60
Synovial fluid had a higher concentration of IL-6 than serum and thus acquired better accuracy in the diagnosis of PJI.64,65 In 2007, Nilsdotter-Augustinsson64 demonstrated that synovial IL-6 had a sensitivity and a specificity of 69% and 93%, respectively, at a threshold of 10,000 pg/ mL in the diagnosis of periprosthetic hip infection. Deirmengian et al,66,67 Gollwitzer et al,68 and Jacovides et al69 separately examined and identified synovial fluid biomarkers but reached the same conclusion—synovial IL-6 has great potential in diagnosing prosthetic infection after total knee arthroplasty (TKA) or total hip arthroplasty (THA). In a meta-analysis containing 7 studies, Frangiamore et al7 showed that the overall sensitivity and specificity of synovial IL-6 were 87% and 89%, respectively, sharing the same accuracy as CRP (95%). Interestingly, in another study, Frangiamore et al70 demonstrated that synovial fluid IL-6 had high sensitivity and specificity (87% and 90%, respectively) in diagnosing periprosthetic shoulder infection, which has always been an obstacle for CRP, PCT, and serum IL-6. The disadvantage of synovial IL-6 is the same as that of CRP: being an inflammatory cytokine rather than a specific infection marker. Synovial IL-6 is upregulated in osteoarthritis, rheumatoid arthritis, and cartilage defects, leading to a diagnostic challenge in these cases.71,72
In conclusion, because of its relatively high accuracy and fast response, IL-6 is showing potential as a valuable predictor of PJI.
Produced by primitive innate immunity, defensins are small, ubiquitous, and multifunctional polypeptides, acting as effectors in inflammation modulation, chemotaxis, and bactericidal activities.73,74 Defensins play a critical part in the pathogenesis of osteomyelitis and arthritis.75–77 There are three subtypes of defensins: alpha, beta, and theta. Only six alpha-defensins (HD5 and HD6; HNP-1, -2, -3, and -4) and seven beta-defensins (HBD-1, -2, -3, -4, -26, -27, and -28) are identified in the human body.78,79 Alpha-defensins are present mainly in the granules of neutrophils, while beta-defensins are secreted by mucosal cells, which may explain the preference for alpha-defensin in studying synovial fluid defensins.78,80 Recent studies have found that the level of synovial fluid alpha-defensin is unaffected by the type of organism or the preoperative use of antibiotics, making it a promising bio-marker for the diagnosis of implant-associated infection.81,82 In 2013, Gollwitzer et al68 first demonstrated a marked elevation of cathelicidin LL-37 and human beta-defensin-3 (HBD-3) in joint aspirates of 35 PJI patients, with areas under the curve of 0.745 and 0.875, respectively. In a study involving 95 revision THA or TKA patients, some of whom had systemic inflammatory disease, Deirmengian et al67 reported that alpha-defensin had an excellent performance of 100% sensitivity and 100% specificity in the diagnosis of infection. Later studies all reached the same conclusion—alpha-defensin has a sensitivity and a specificity of more than 95% in the diagnosis of PJI, outperforming the LE test.46,82–85 A meta-analysis containing 6 studies calculated that the pooled sensitivity and specificity of alpha-defensin were 100% and 96%, respectively.48 In a more recent study, Yuan et al86 showed that the pooled diagnostic sensitivity of alpha-defensin for PJI was 96% and the specificity was 95%. However, the sensitivity and the specificity of synovial alpha-defensin decrease to 63% and 95%, respectively, for the diagnosis of peri-prosthetic shoulder infection related to P acnes.87 A similar decrease in accuracy has also been reported for the diagnosis of infection in second-stage revision THA or revision TKA (sensitivity, 67%; specificity, 97%).85 Changing the testing method from immunoassay to lateral flow test also deceases the sensitivity to 67% for the diagnosis of PJI.88
Generally, based on its high accuracy, synovial alpha-defensin is now the most promising biomarker for the diagnosis of implant infection and has the potential to be “the orthopedic mirror of the human chorionic gonadotropin test.”89 However, testing for alpha-defensin is expensive (approximately US $760). A standard testing method and the diagnostic value against less virulent bacteria should be determined for alpha-defensin to become a practical biomarker.48,89 The meta-analysis results and the characteristics of current major biomarkers in diagnosing implant-associated infection are summarized in Table 1.
Meta-Analysis and Characteristics of Different Biomarkers for the Diagnosis of Implant-Associated Infection
Bacterial Genetic Markers
Diagnostic biomarkers based on innate immune response are convenient and ready to use. However, in some circumstances, it is difficult for these biomarkers to distinguish between aseptic inflammation and low virulence infections. Also, testing protein-based biomarkers requires tissue, making early detection difficult.90 As a result of advances in molecular technologies, new methods testing bacterial genetic markers are also being investigated. Typically, with these methods, fluid or tissue samples are collected adjacent to the implant for DNA extraction. Sonication of implants has been shown to improve the specificity of the polymerase chain reaction (PCR) results of the samples.91 Currently, molecular diagnosis of implant-associated infection mainly involves conventional broad-range PCR, specific PCR assays, and sequencing technology. The broad-range PCR tests bacteria-specific genes, including the 16S rDNA gene, the 23S rDNA gene, the rpoB gene, and the 16S–23S intergenic spacer located specifically in bacterial genomes.92 However, only 16S rDNA amplification, when yielding a positive result, allows precise identification of the organisms.93 Specific PCR assays are designed to test pathogen-specific genes and have been shown to be more sensitive than 16S rDNA PCR in testing for Staphylococcus aureus and Mycobacterium tuberculosis infection in bone.93,94 The sensitivity and the specificity of PCR for diagnosing implant-associated infection vary greatly, ranging from 16% to 96% and from 45% to 100%, respectively.90 In a meta-analysis consisting of 9 studies, Liu et al95 concluded that the sensitivity and the specificity of PCR of sonicated fluid for periprosthetic infection were 75% and 96%, respectively, equivalent to or better than those of intra-operative tissue culture. The potential of high-throughput sequencing for diagnosing infection is also being explored. The use of high-throughput sequencing allows for the establishment of a complete library of DNA within the aspirated fluid or tissue, thus providing better resolution than broad-range PCR for identifying possible bacteria. In a recent study, the sensitivity and the specificity of next-generation sequencing were 89.3% and 73.0%, respectively, while the sensitivity and specificity of tissue culture were 60.7% and 97.3%, respectively.96 Although sequencing cannot outperform tissue culture regarding diagnostic value, use of this macrogenomic method provides more thorough information on the whole picture of the organisms within the implant and offers a promising way to identify rare bacterial infections.96,97 Despite the benefits of early detection and organism identification, molecular methods have varied sensitivity and the potential for contamination, posing challenges to their use.92 Although the true value of diagnosis based on bacterial genetic markers remains unclear, they are a promising way to supplement tissue culture and may provide information to target therapy.98
Many other inflammatory cytokines and bactericidal mediators have also been explored in pursuit of an optimal bio-marker for implant-associated infections. Gollwitzer et al68 showed that serum IL-4 and IL-6 had a specificity of 90.0% and 95.0% and a sensitivity of 60.0% and 46.7%, respectively, outperforming synovial IL-4 and IL-6. Gollwitzer et al68 also demonstrated that serum HBD-2 had a sensitivity of 80.0% and a specificity of 70%. Serum tumor necrosis factor-alpha and lipopolysaccharide-binding protein are small molecules induced by cytokines. Studies have found that tumor necrosis factor-alpha has a low sensitivity of 43% and that lipopolysaccharide-binding protein has both low sensitivity and specificity.30,99 In a recent study, Shahi et al100 demonstrated that serum D-dimer had a high sensitivity and specificity of 89% and 93%, respectively, in diagnosing PJI. This is encouraging because D-dimer is frequently tested, convenient, and inexpensive. Regarding synovial fluid biomarkers, Deirmengian et al66,67 identified that IL-1a, IL-1b, IL-6, IL-8, IL-10, IL-17, granulocyte colony-stimulating factor, elastase 2, bactericidal permeability-increasing protein, neutrophil gelatinase-associated lipocalin, lactoferrin, resistin, and thrombospondin all exhibit an area under the curve of more than 0.9, outperforming serum CRP. In addition to these markers, researchers also found that skin-derived antileukoproteinase and vascular endothelial growth factor were markedly elevated after infection and acquired good diagnostic accuracy.66,69 There were also studies focusing on the infected tissue as a whole. Toll-like receptors (TLRs) are a family of transmembrane receptors that take part in the reorganization of microbial components or pathogen-associated molecular patterns. In 2014, Cipriano et al101 investigated the expression of TLR messenger RNA in periprosthetic tissue using real-time PCR and concluded that TLR1 (sensitivity, 95.2%; specificity, 100%) and TLR6 (sensitivity, 85.7%; specificity, 82.8%) could be used to detect PJI.