Clinical Review (insufficient quality of evidence to enable a Clear Recommendation)
Based on observational studies, we do not recommend the routine use of plain X-rays (inadequate sensitivity, specificity) or CT scans (inadequate sensitivity) for all patients with a possible diagnosis of osteomyelitis (Table 1) as they may result in unnecessary radiation and use of resources. However, these studies may be helpful if a fracture or other non-infectious cause of bone pain (e.g., tumor, foreign object, etc.) is prioritized on the differential diagnosis, and/or the pre-test probability of osteomyelitis is lower (e.g., ≤15%). Magnetic resonance imaging (MRI) and certain tagged white cell scans are the most accurate imaging modalities for diagnosing osteomyelitis. Inflammatory biomarkers are insufficiently accurate, and we do not recommend their routine use for osteomyelitis diagnosis. Blood cultures have variable sensitivity but if the patient has systemic symptoms or risk factors for bacteremia (e.g., intravenous drug use), isolating likely pathogens (e.g., Staphylococcus aureus) can be helpful to target therapy, and potentially obviate the need for bone biopsy. If available, bone biopsy for histopathology is highly accurate if positive, but cannot rule out osteomyelitis if negative. Culture of biopsy specimens of the affected bone may help identify etiology and target antimicrobial therapy.
Based on observational studies, plain X-rays have low sensitivity and specificity for diagnosing DFO (Table 1). The probe-to-bone (PTB) test is simple, non-invasive, and has reasonable sensitivity and specificity as a diagnostic method for DFO, which may preclude the need for imaging in some settings. MRI and certain tagged white cell scans are the most accurate imaging modalities for diagnosing DFO, although their specificities are lower than their sensitivities. Inflammatory biomarkers are insufficiently accurate, and we do not recommend their routine use for diagnosis. If available, percutaneous bone biopsy for deep microbiological cultures may help target antimicrobial therapy; surface cultures are not accurate and not recommended.
There is no established, accurate referent standard diagnostic test for PJI. Certain tagged white cell scans are the most accurate imaging studies for PJI (Table 1), however given the limitations of individual tests, published algorithms are sometimes recommended to establish the diagnosis. Data are limited and inadequate to compare the relative accuracies of competing algorithms. Practically, the diagnosis is typically made from a combination of history, physical exam, imaging studies to assess alternate causes of pain and instability, inflammatory markers, synovial fluid analysis, and/or operative specimens. Molecular diagnostic testing is a promising approach, but data are mixed and inadequate to recommend for or against its use as of 2022.
No RCTs have been published that define optimal diagnostic strategies for osteomyelitis. However, numerous observational studies of various designs, sizes, and quality have been published that evaluated the accuracy of various diagnostic modalities.
Imaging Studies (Table 1)
Plain Films
The primary utility of plain X-rays in evaluation of patients with osteomyelitis is to exclude other diagnoses, such as fractures, metallic foreign bodies, or malignancies.(Zhou, Girish et al. 2021) Periosteal elevation, the most common X-ray finding suggesting osteomyelitis, is neither sensitive nor specific for osteomyelitis.(Lee, Sadigh et al. 2016, Zhou, Girish et al. 2021) It can be caused by any condition in which there is inflammation of the tissue layers above the bone, as well as tumors, or even stress fractures.(Lee, Sadigh et al. 2016, Zhou, Girish et al. 2021) Furthermore, the earliest bony change in osteomyelitis is marrow edema, which cannot be detected by X-rays. As such, plain X-rays are typically normal during the earlier phases of osteomyelitis and X-ray findings such as cortical destruction and bony erosions are rare, and typically found after prolonged periods of infection.(Lee, Sadigh et al. 2016, Fattore, Goh et al. 2020, Zhou, Girish et al. 2021) However, the sensitivity of X-rays does likely increase with time in untreated infections due to progressive bony erosion.
Llewellyn et al. conducted the most recent and comprehensive, systematic review and meta-analysis of various imaging modalities at diagnosing osteomyelitis across all body sites.(Llewellyn, Jones-Diette et al. 2019) Eighty-one studies were included. Plain X-rays had a pooled (95% CI) sensitivity and specificity of only 70% (62%-79%) and 82% (70%-90%), respectively, resulting in poor positive and negative likelihood ratios (3.9 and 0.4, respectively). There was no noted variation in study accuracy by type of osteomyelitis or body site. Thus, X-rays are neither sensitive nor specific for the diagnosis of osteomyelitis.
Computerized Tomography (CT) Scans
CT scans may detect cortical disruption of bone and play a role in identifying target sites for needle biopsies in osteomyelitis. However, CT scans cannot detect marrow edema, making them insensitive for the diagnosis of osteomyelitis, particularly in earlier phases.(Pugmire, Shailam et al. 2014, Fattore, Goh et al. 2020, Zhou, Girish et al. 2021) Llewellyn et al. found that CT scans had a pooled (95% CI) sensitivity and specificity of 70% (40%-89%) and 90% (58%-98%), respectively, for diagnosing osteomyelitis.(Llewellyn, Jones-Diette et al. 2019) These results indicate that a positive CT scan may be helpful to diagnose osteomyelitis (positive likelihood ratio of 7.0), but a negative CT scan is less helpful in ruling it out (negative likelihood ratio of 0.3).
Nuclear Medicine Studies
Similarly to plain X-rays and CT scans, Llewellyn et al. reported that nuclear medicine scintigraphy studies of various types (excluding certain tagged white cell scans, which are considered separately, below) had relatively poor pooled (95% CI) sensitivity and specificity at 84% (72%-91%) and 71% (58%-81%), yielding positive and negative likelihood ratios of 2.9 and 0.2 respectively, for diagnosing osteomyelitis.(Llewellyn, Jones-Diette et al. 2019) Thus, such scans may be modestly helpful if negative, depending on pre-test probability, but do not substantively alter pre-test probability if positive. Similarly, in a study of 30 patients with possible osteomyelitis, 15 of whom were subsequently confirmed to have osteomyelitis affecting a range of bones, and 15 with other inflammatory, malignant, or traumatic/degenerative injuries, bone scans were positive in all nine patients with osteomyelitis in whom they were obtained.(Howard, Einhorn et al. 1994) However, they were also falsely positive in 11 of 12 (specificity 8%) patients who were ultimately diagnosed with other conditions.(Howard, Einhorn et al. 1994) These results are consistent with findings from a more recent study in which the accuracy of triple phase bone scanning and single photon emission computerized tomography (SPECT) scanning were poor to mixed for diagnosing osteomyelitis (ranging from 60-90% sensitivity/specificity, likelihood ratios <5).(Lee, Won et al. 2021)
Llewellyn et al. reported superior accuracy of certain tagged white cell scans, positron emission tomography (PET) scans, and SPECT scans for diagnosing osteomyelitis.(Llewellyn, Jones-Diette et al. 2019) Specifically, tagged white cell scans had a pooled (95% CI) sensitivity and specificity of 87% (75%-94%) and 95% (85%-98%), yielding positive and negative likelihood ratios of 17.4 and 0.1, respectively. PET scans had pooled (95% CI) sensitivity and specificity of 85% (72%-93%) and 93% (83%-97%), yielding positive and negative likelihood ratios of 12.1 and 0.2, respectively.(Llewellyn, Jones-Diette et al. 2019) SPECT scans had a pooled (95% CI) sensitivity and specificity of 95% (88%-98%) and 82% (62%-93%), yielding positive and negative likelihood ratios of 5.3 and 0.06, respectively. Thus, tagged white cell and PET scans are more specific than SPECT scans, and the latter are more sensitive.
MRI
MRI is the most accurate, generally available radiographic method to identify osteomyelitis, with sensitivities in excess of 90-95% and specificities of 80-90%, resulting in positive likelihood ratios of 5-10 and negative likelihood ratios of 0.056-0.125.(Pugmire, Shailam et al. 2014, Lee, Sadigh et al. 2016, Llewellyn, Jones-Diette et al. 2019, Gornitzky, Kim et al. 2020, Llewellyn, Kraft et al. 2020, Zhou, Girish et al. 2021) Furthermore, MRIs do not expose the patient to ionizing radiation, and become positive for osteomyelitis substantially earlier than X-ray and CT imaging.(Fattore, Goh et al. 2020)
In their recent review of 81 studies, Llewellyn et al. reported that the pooled (95% CI) sensitivity and specificity of MRI for diagnosing osteomyelitis were 96% (92%-98%) and 81% (71%-88%), respectively (positive and negative likelihood ratios of 5.1 and 0.05, respectively).(Llewellyn, Jones-Diette et al. 2019) MRI has also been found to be accurate in diagnosing skull-base osteomyelitis. In one study of patients with malignant otitis externa, the sensitivity and specificity of MRI using specific diffusion weighted imaging cut-offs was 86-90% and 79-90%, respectively (positive and negative likelihood ratios of ~6 and ~0.1 respectively).(Abdel Razek and Mahmoud 2020) Similarly, in a case series of patients with skull base osteomyelitis or tumors, MRIs were able to distinguish infection from malignancy.(Lesser, Derbyshire et al. 2015)
Summary of Radiographic Studies for Osteomyelitis without PJI
In summary, a consistent and substantial body of literature has found that plain X-rays, CT scans, and various forms of scintigraphy (excepting tagged white cell scans) are relatively inaccurate for diagnosing osteomyelitis. X-rays and CT scans can be specific when bony destruction is encountered without an alternative explanation, but these findings are late, may be rarely encountered, and cannot be distinguished from other destructive bony processes without further investigation (e.g., biopsy, surgery). Bone scans can be sensitive but are highly non-specific for osteomyelitis.
While plain X-rays are often recommended to be routinely obtained as an initial diagnostic tool in various guidelines sources, we emphasize that this practice may be rooted in historical inertia rather than published data. Indeed, one review of guidelines has acknowledged that this recommendation by others is based on low quality evidence.(Tran and Mierzwinski-Urban 2020) As plain X-rays are not accurate for diagnosing osteomyelitis, their primary role may be in diagnosing a patient for whom there is a prioritized concern for non-infectious causes of disease, such as fracture or other mechanical causes of bone pain. In a patient with a low initial pre-test probability of osteomyelitis (e.g., 15% or less), a negative X-ray can preclude the need for a diagnostic MRI (shifting post-test probability to ≤ 5%). However, one might question the need for an MRI in such a situation, irrespective of X-ray results.
We emphasize that obtaining numerous relatively low-cost tests adds considerable overall cumulative cost to healthcare. More importantly, obtaining a high volume of such tests occupies a considerable amount of radiology technician time, which can create backlogs and delays in care for other patients. Thus, the notion that X-rays are less expensive or resource intensive than other options may be a false conclusion given the high volume of such studies, and lack of adequate consideration regarding the overall time required for radiology technicians to obtain and process such films, the clinician time taken to interpret them, and importantly, their low overall diagnostic value.
Overall, MRIs appear to be the most sensitive of all radiographic studies, and they have the advantage of not exposing patients to ionizing radiation. PET scans and tagged white cell scans may be more specific than MRIs, but are less sensitive, do expose patients to ionizing radiation, and may be more expensive depending on the care setting. Thus, MRIs are a reasonable, primary radiographic modality to diagnose osteomyelitis. PET and tagged white cell scans may be alternatives in circumstances where patients cannot receive an MRI, presuming that completing an empiric course of antimicrobial therapy is deemed a less acceptable option than establishing a definitive diagnosis of osteomyelitis.
Biomarkers: ESR, CRP, Procalcitonin, Others
Most studies evaluating accuracy of inflammatory biomarkers in the modern era have used MRI as the referent standard for identifying osteomyelitis patients. Many of these studies have reported limited accuracy of ESR and CRP at diagnosing osteomyelitis. Finally, given the accuracy of MRI, virtually no study has defined a role for inflammatory biomarkers to further improve the accuracy of diagnoses compared to MRI—i.e., how does the inflammatory biomarker add to diagnosis when an MRI is already done, or will be done irrespective of the biomarker result?
For example, in an observational study of 133 patients with vertebral osteomyelitis, Ghassibi et al. evaluated the impact of specific pathogens on ESR, CRP, and other biomarker values.(Ghassibi, Yen et al. 2021) The reference standard for confirming osteomyelitis diagnosis was MRI. The mean ESR and CRP were substantially greater than the normal value cut-off, but did not achieve meaningful accuracy (e.g., likelihood ratios all <5). The mean white blood cell (WBC) count in peripheral blood was only 12,300 per microliter, slightly greater than normal. The mean percent neutrophil count was also unhelpful. Biomarker values were higher for pyogenic causes, and S. aureus in particular, than for culture-negative, fungal, or tuberculous (TB) osteomyelitis. Only S. aureus and streptococci caused mean WBC counts to rise above normal levels. CRP was normal in half of patients and remained unhelpful to dichotomize patients into those with or without osteomyelitis. WBC count was normal in 86% of patients with culture negative, and 100% of fungal osteomyelitis. A variety of methods were used to calculate receiver operating curves (ROCs) for each of these biomarkers. None led to meaningful biomarker cut-offs. Corroborating these results with MRI as the referent standard, Scharrenberg et al. reported that CRP levels had limited ability to distinguish vertebral osteomyelitis from non-infectious causes, including erosive spinal degeneration (e.g., erosive osteochondrosis), particularly post-operatively.(Scharrenberg, Yagdiran et al. 2019)
Similarly, in a study of 30 patients with suspected osteomyelitis, 15 subsequently were determined to have osteomyelitis of various body parts, and the remaining patients were ultimately diagnosed with arthritis, myositis, trauma, sarcoma, or other inflammatory disorders.(Howard, Einhorn et al. 1994) Temperature, WBC count, and ESR did not distinguish infectious from non-infectious, inflammatory bone disorders.
In a larger study of 163 patients with traumatic extremity osteomyelitis, ESR and CRP were again poorly accurate at diagnosing osteomyelitis.(Wu, Lu et al. 2020) Finally, these results are reinforced in a study of 102 patients with pedal osteomyelitis who did not have diabetes.(Ryan, Ahn et al. 2019) Using a combination of MRI and SPECT scanning, combined with bone culture and histopathology as the referent standards, the optimal cut points of ESR and CRP achieved sensitivity/specificities of only 49%/79% and 45%/71%, respectively.(Ryan, Ahn et al. 2019)
Thus, overall, studies have not found that typical inflammatory biomarkers are accurate in diagnosing osteomyelitis. None have demonstrated that they can be used to avoid imaging studies or enhance diagnostic accuracy compared to imaging studies.
Blood Cultures
Numerous observational studies have described a wide range of sensitivity of blood cultures for establishing the microbial etiology of osteomyelitis of numerous types. A 2016 review described blood culture positive rates of 40-89% across 11 studies of patients with vertebral osteomyelitis.(Nickerson and Sinha 2016) Similarly, more recent case series reported 60%(Amsilli and Epaulard 2020) or 31%(Avenel, Guyader et al. 2021) blood culture positivity for vertebral osteomyelitis. While negative blood cultures are not helpful, positive blood cultures that identify a likely pathogen may obviate the need to proceed to more invasive microbiological testing (e.g., bone biopsy, discussed below). Furthermore, blood cultures are relatively non-invasive and inexpensive. They are therefore reasonable to obtain in patients with systemic signs of infection and a high suspicion for pyogenic osteomyelitis.
Biopsy & Culture
Several small observational studies have evaluated the diagnostic accuracy of bone histopathology and/or microbial culture for osteomyelitis.
A meta-analysis of 7 prior observational studies including 482 patients evaluated the overall diagnostic accuracy of imaging-guided biopsy to diagnose vertebral osteomyelitis.(Pupaibool, Vasoo et al. 2015) The overall sensitivity and specificity of biopsy was 52% and >99%, respectively, resulting in a remarkably high positive likelihood ratio of 52, and a poor negative likelihood ratio of 0.5. Thus, positive biopsy histopathology is extremely accurate for ruling in the diagnosis of osteomyelitis, but a negative biopsy is very poor at ruling out osteomyelitis.
Histopathology can also help identify osteomyelitis caused by atypical pathogens (e.g., via granulomas, special stains such as AFB and silver stains, and even particular pathogen-specific findings, such as Coccidioides spherules). Furthermore, even a positive Gram stain on histopathology, in the absence of a positive culture, can enable targeted antibiotic de-escalation or escalation when appropriate.
Histopathology might have a higher sensitivity than culture. For example, in a study of 30 patients with suspected osteomyelitis at various body sites, the authors conducted what they referred to as “fine needle bone biopsies”.(Howard, Einhorn et al. 1994) However, the needle size used for the biopsies was 11 gauge, and hence the biopsies performed might be more accurately described as core biopsies using modern vernacular. Fifteen patients were ultimately diagnosed with osteomyelitis; the referent standard in this study was triple phase bone scan, radiographic changes on serial plain X-rays, or finding of pus on biopsy under the periosteum. Thirteen of 15 patients diagnosed with osteomyelitis had positive histopathology on biopsy (sensitivity 87%). Of the 15 patients who did not have osteomyelitis, histopathology was accurately negative in 14 (specificity 93%). Of note, 1 patient had a false positive diagnosis of osteomyelitis on histopathology due to severe acute and chronic inflammation with bone necrosis, in what was subsequently determined to be an Ewing’s sarcoma lesion. Nevertheless, this study reinforced the potential superiority of histopathology over culture as means to establish a diagnosis of osteomyelitis. Similarly, in a study of 29 CT-guided bone biopsies, only 21% resulted in positive cultures.(Beroukhim, Shah et al. 2019)
Additionally, in a study of 84 patients who underwent CT-guided vertebral bone biopsy, histopathology was positive in 41% of biopsy samples, whereas culture was only positive in 19%.(Garg, Kosmas et al. 2014) Furthermore, among control patients who were biopsied due to suspicion of non-infectious causes (primarily cancer), 77% of the biopsies established an alternative, non-infectious diagnosis by histopathology. This study reinforces that the benefits of histopathology lie in its ability to diagnose both infectious and alternate, non-infectious etiologies.
However, in other studies of 88, 46, 142, 111, and 64 patients with vertebral, other axial, or extremity osteomyelitis who underwent biopsy, the positivity rate of culture was higher at 60%, 36%, 43%, 36%, and 31%, respectively.(Chang, Simeone et al. 2015, Kasalak, Wouthuyzen-Bakker et al. 2018, Diffre, Jousset et al. 2020, Lee, Pukenas et al. 2020, Schiro, Foreman et al. 2020) While histopathology may be more sensitive, and enable diagnosis of alternative diseases, the benefit of culture results is that they frequently enable more appropriate targeting of antimicrobial therapy.(Kasalak, Wouthuyzen-Bakker et al. 2018) Finally, the site of biopsy may be relevant, given that another study found that biopsies of paravertebral soft tissues resulted in a higher diagnostic yield than bone biopsy (68% vs. 38% for endplate-disk biopsies).(Chang, Simeone et al. 2015)
Whether all patients in whom osteomyelitis is a concern should be subjected to bone biopsy cannot be established from the literature. Ultimately, the low yield of bone biopsies may suggest that they are not routinely required in all patients, particularly, as is true for any procedure, because there are procedural risks for the patient. For example, in a study of 78 patients with vertebral osteomyelitis who underwent bone biopsy, only 10 patients had positive histopathology, 14 had positive culture, and eight were positive both by histopathology and culture.(Lim, Walter et al. 2021) Only 19 of the biopsies altered treatment; 15 patients underwent de-escalation, and antimicrobial therapy was expanded in four cases.
Overall, the advantages of bone biopsy include confirming the diagnosis, evaluating for alternative diagnoses (such as malignancy), helping to identify atypical cases (e.g., TB, fungal), providing initial guidance to adjust antimicrobial therapy based on Gram stain, and enabling targeted therapy if cultures are positive. Disadvantages include that bone biopsies require considerable resources and may be difficult to obtain in many patient settings, the procedure is invasive with risks of harm to the patient, it may establish the diagnosis in half or less cases, and even histopathology can rarely lead to incorrect diagnoses.(Howard, Einhorn et al. 1994) Nonetheless, in patients for whom a microbiological diagnosis has not been otherwise achieved, it may be reasonable to pursue a biopsy in patients who are not felt to be adequately responding to appropriate empiric therapy in order to identify resistant, unusual, or non-bacterial pathogens and to exclude alternative diagnoses.
Thus, obtaining a bone biopsy must be individualized on a case-by-case basis and may rationally vary by practice setting and patient characteristics. It is also unclear if multiple biopsy specimens would increase or alter sensitivity or specificity diagnostically, and this is an area that warrants study in the future.
Special Methods and Molecular Diagnostics
Limited studies are available to evaluate novel molecular diagnostic methods for osteomyelitis without PJI. Choi et al. reported that among 45 patients with vertebral osteomyelitis who underwent a bone biopsy or aspirate, 21 patients had true positive 16S RNA PCR tests, identifying pathogens, and three patients had false positive PCR tests.(Choi, Sung et al. 2014) In contrast, culture was true positive in 12 patients and false positive in one. Other case reports have also used PCR to identify unusual pathogens in osteomyelitis without PJI, but only in small numbers and without controls.(Shibata, Tanizaki et al. 2015, Alkhatib, Younis et al. 2017, Hase, Hirooka et al. 2018) Although data are not available to confirm accuracy, it may be rational to attempt such molecular diagnostic methods particularly for culture-negative infections when patients are clinically not responding to empiric therapy and this testing is available.
Probe-to-Bone (PTB) Test
A systematic review included seven studies with 1,025 patients and compared the PTB test to reference standards (primarily bone biopsy).(Lam, van Asten et al. 2016) The PTB test had a pooled (95% CI) sensitivity of 87% (75%-93%) and specificity of 83% (65%-93%), resulting in reasonable positive and negative likelihood ratios of 5 and 0.2, respectively. In a patient with low to moderate pre-test probability, this accuracy may be sufficient to exclude DFO without requiring an MRI or other imaging studies (e.g., 33% pre-test probability shifts to <10% post-test probability with a negative probe-to-bone test).
Some experts have suggested that the combination of a negative plain X-ray and PTB test enhances sensitivity compared to PTB alone. However, there are very limited data assessing the accuracy of the combination vs. either test alone. One study of the combination of PTB and plain x-ray did find an impressive sensitivity and specificity of 97% and 92%, respectively in diagnosing DFO.(Aragon-Sanchez, Lipsky et al. 2011) However, in that study, the sensitivity and specificity of both tests alone were also unusually high (95% and 93% for PTB and 82% and 93% for plain X-ray, respectively). Thus, combining PTB with plain X-ray did not appreciably enhance accuracy compared to PTB alone. Furthermore, the authors note that there was an unusually high pre-test probability of DFO in the patients because they were performed at a referral center for this disease.
In a second study, which also had a high pre-test probability of DFO, combining the PTB test with plain X-rays reduced sensitivity and specificity compared to PTB alone. Specifically, Morales Lozano et al. evaluated the sensitivity and specificity of the PTB test with or without plain X-rays in a prospective study of patients with diabetic foot ulcers.(Morales Lozano, Gonzalez Fernandez et al. 2010) The PTB test had an impressive 98% sensitivity and 78% specificity for DFO. However, combining the PTB test with plain X-rays lowered the sensitivity to 89% and lowered the specificity to 77%. Thus, data appear insufficient to support use of combination plain X-rays and PTB test for DFO and suggest that PTB may perform adequately by itself. Further study is needed.
Imaging Studies (Table 1)
In a comprehensive systematic review of the accuracy of imaging studies specifically for the diagnosis of DFO, Llewellyn et al. determined that the pooled (95% CI) sensitivity and specificity of plain X-rays was only 62% (51%-72%) and 78% (63%-89%).(Llewellyn, Kraft et al. 2020) These results yielded positive and negative likelihood ratios of 2.8 and 0.5, respectively.(Llewellyn, Kraft et al. 2020) Thus, the accuracy of plain X-rays for diagnosing DFO is poor, and they are of low value for this purpose. Nuclear medicine studies were more sensitive but less specific, resulting in poor overall accuracies. Specifically, the pooled (95% CI) sensitivity and specificity of various forms of scintigraphy (excluding tagged white cell scans) were only 85% (77%-90%) and 68% (56%-77%), resulting in positive and negative likelihood ratios of 2.7 and 0.2, respectively.
PET scans are more accurate, although they are also more resource intensive and less available. They also expose the patient to considerable ionizing radiation. Llewellyn et al. found their pooled (95% CI) sensitivity and specificity to be 84% (53%-96%) and 93% (76%-98%), respectively for DFO.(Llewellyn, Kraft et al. 2020) Based on only three studies, SPECT had superior sensitivity but much inferior specificity at 96% (76%-99%) and 55% (19%-86%), respectively. In a smaller systematic review of six studies of PET scans for diagnosing DFO, the sensitivity of the test ranged from 75% to 90% and specificity ranged from 75% to 98% in individual studies.(Llewellyn, Kraft et al. 2020) Similarly, Lauri et al. systematically reviewed the literature for diagnostic accuracy of PET scans and tagged white cell scans for DFO.(Lauri, Tamminga et al. 2017) They found that PET scan sensitivity/specificity were 89% and 92%, respectively, and tagged white cell scan sensitivity and specificity were 91-92% and 75-92%, respectively, depending on the nature of the white cell label.
Overall, MRIs remain the most accurate routinely available imaging test for DFO while avoiding ionizing radiation. Llewellyn et al. reported a pooled sensitivity and specificity of 96% and 84%, respectively, resulting in positive and negative likelihood ratios of 6 and 0.05, respectively.(Llewellyn, Kraft et al. 2020) In their review, Lauri et al. reported a similar sensitivity and specificity of MRI for DFO at 93% and 75%, respectively.(Lauri, Tamminga et al. 2017) Finally, in a third systematic review of 36 studies evaluating various imaging studies for diagnosing DFO, MRI was the most accurate study overall, and the authors reaffirmed its role as the preferred diagnostic modality due to accuracy and the avoidance of ionizing radiation.(Llewellyn, Kraft et al. 2020)
Inflammatory Biomarkers
Xu et al. evaluated a combination of PTB test plus inflammatory biomarkers for diagnosing 111 cases of DFO from among 204 patients with diabetic foot infections (DFI).(Xu, Cheng et al. 2020) The referent standard for diagnosis was a positive bone biopsy. The authors found that WBC count, percent neutrophils in peripheral blood, and CRP were all inaccurate at distinguishing DFO from non-osteomyelitis cases. While ESR was the most accurate biomarker, it achieved an 83% sensitivity and 71% specificity at its best cut-point, which would not substantively alter post-test probabilities in individual patients unless the pre-test probability was already quite low (a 15% pre-test probability would become a 4% post-test probability).(Xu, Cheng et al. 2020)
Moallemi et al. also found limited accuracy (sensitivity and specificities of 60%-70%, positive and negative likelihood ratios 2 and >0.5, respectively) for ESR and CRP in diagnosing DFO.(Moallemi, Niroomand et al. 2020)
Lavery et al. evaluated 353 patients with DFI, of which 176 had DFO.(Lavery, Ahn et al. 2019) Of note, they excluded patients with comorbid conditions that could have falsely elevated ESR and CRP results in an attempt to optimize testing accuracy. The referent standard was MRI or SPECT scan and bone biopsy. The investigators evaluated multiple cut-points of ESR and CRP and found none that yielded good discriminatory results. At their optimal cut points, sensitivity, and specificity of ESR were only 74% (95% CI, 67%-80%) and 56% (95% CI, 48%-63%), respectively; sensitivity and specificity of CRP were only 49% (95% CI, 41%-57%) and 80% (95% CI, 74%-86%), respectively. Especially given that these accuracies are inflated by exclusion of patients with comorbidities associated with inflammation, these results cast considerable doubt on the utility of ESR and CRP as a diagnostic tool for DFO.
Finally, a study of 90 patients determined that procalcitonin had greater accuracy than ESR or CRP at diagnosing DFO, with MRI again the referent standard.(Soleimani, Amighi et al. 2021) However, the overall levels of procalcitonin (PCT) were 0.13 +/- 0.02 ng/ml in patients with osteomyelitis, whereas typically levels of >0.5 are used to distinguish the need for continuation of antibiotics. Hence, the procalcitonin levels were quite low, and within the range that would not typically be used to support antibacterial therapy.
Several systematic reviews have also sought to define the accuracy of inflammatory biomarkers at diagnosing osteomyelitis, with conflicting findings. Victoria van Asten et al. reviewed eight studies of patients with DFO and found that only ESR was accurate at diagnosing osteomyelitis; CRP, procalcitonin, and various inflammatory cytokines (e.g., interleukins IL-2, IL-6, IL-8, and tumor necrosis factor (TNF)) were not.(Victoria van Asten, Geradus Peters et al. 2016) They reported a higher sensitivity and specificity of ESR in this setting than other studies cited above, with sensitivity and specificity of 81% (95% CI, 71%-88%) and 90% (95% CI, 75%-96%), respectively. In contrast, in their review of the literature, Markanday found lower accuracies, reporting sensitivities and specificities for both ESR and CRP in the 70-80% range for DFO.(Markanday 2015)
Percutaneous Bone Biopsy (PBB)
PBB may be useful to help make the diagnosis of DFO and guide antimicrobial therapy when surgical intervention is not planned. An 11-study systematic review of patients with DFO found that the pooled proportion of culture-positive PBB was 84% (95% CI, 73%-91%).(Schechter, Ali et al. 2020) There was extensive heterogeneity, however after excluding two studies with very high proportions of positive culture results, the pooled proportion of culture positive PBBs was more conservatively estimated at 77% (95% CI, 68%-85%).
While culture yields may be relatively higher for patients with suspected DFO compared to other sites of osteomyelitis,(Wu, Gorbachova et al. 2007) the systematic review highlighted limitations that should be considered. First, studies seldom reported the technical aspects of biopsied procedures (e.g., needle gauge size). Second, ulcer severity scores were under-reported. Third, only one study provided methods for identifying or defining contaminants. Finally, concerns have been raised about lower culture yields when received antibiotics prior to the biopsy occurring. The relative timing of antibiotic exposure to biopsy is discussed in Section 2.
As for biopsies for osteomyelitis outside the context of DFO (see above), PBB specimens tend to have higher rates of positive histopathology than cultures. For example, Tardaguila-Garcia et al. compared the sensitivity and specificity of bone histopathology and culture in 52 patients for whom clinicians had a suspicion for DFO.(Tardaguila-Garcia, Sanz-Corbalan et al. 2021) A limitation of this study was that no specific referent standard was used to identify confirmed osteomyelitis. Suspicion was based on a positive PTB test with serial plain films. Biopsies were obtained after cessation of antibiotics for 48-72 hours, and then after surgical debridement of surface materials. Thirty-six patients had a positive microbial culture from the bone biopsy, compared to 47 patients who had histopathological evidence of osteomyelitis. Thus, cultures were less sensitive than histopathology, but it is not possible to determine the precise accuracy of histopathology from a focal biopsy in the absence of a referent standard.
Furthermore, there may be considerable variation between pathologists when diagnosing osteomyelitis. In one study of four pathologists independently reviewing 39 cases of suspected osteomyelitis, the kappa coefficient for concordance was only 0.3 (1/3 correspondence rate), which is indicative of only fair agreement.(Meyr, Singh et al. 2011) Consistency may be improved when pathologists use a standardized framework for diagnosis.(Cecilia-Matilla, Lazaro-Martinez et al. 2013)
Similarly, concordance between surface swabs and deep bone biopsies are poor and underscore the lack of utility of surface swabs for diagnosing DFO. For example, Senneville et al. reported the results of bone biopsy from 76 patients with DFO.(Senneville, Melliez et al. 2006) The concordance of surface swab culture compared with a culture of a percutaneous bone biopsy specimen was only 23%. They subsequently published a review of two other studies comparing superficial swabs to bone or deep tissue cultures in which concordance was only 19% and 38%, respectively.(Senneville, Lipsky et al. 2020) The investigators concluded that superficial swab cultures should not be performed. Similarly, concordance between surface swabs and deep bone biopsies are poor and underscore the lack of utility of surface swabs for diagnosing DFO.
Finally, non-culture-based molecular diagnostics are promising but cannot be routinely recommended as part of a biopsy panel at this time. A study by Malone et al. evaluating peptide nucleic acid fluorescence in situ hybridization (PNA FISH) of proximal tissue margins of surgical resections found 8/14 (57%) specimens without growth had a positive result.(Malone, Fritz et al. 2019) Despite finding bacteria in proximal clean margins, there was no data suggesting worse outcomes. Also, contamination at the time of specimen collection or during subsequent handling could lead to false positive results. Time to results, cost, and interpretations are barriers to using these newer diagnostics.
As for other types of osteomyelitis, the number of biopsies to optimize diagnostic sensitivity and specificity is of interest for the future study.
Overview
There is no uniformly accepted diagnostic criteria for PJI. There have been multiple attempts to develop diagnostic criteria for PJI, including the Musculoskeletal Infection Society (MSIS) initial definition in 2011, followed by the modified International Consensus on Musculoskeletal Infection (ICM) criteria initially in 2013, revised in 2018, and more recently, the European Bone and Joint Infection Society (EBJIS), now endorsed by MSIS as well.(Parvizi, Zmistowski et al. 2011, Parvizi, Gehrke et al. 2014, Parvizi, Tan et al. 2018, Schwarz, Parvizi et al. 2019, McNally, Sousa et al. 2021) These definitions are based on clinical characteristics, blood biomarkers, synovial fluid studies, microbiology tests, and histology results with various cut off values to define PJI. In these various classifications, there are a few criteria that, if present alone, are proposed to confirm the presence of PJI. Combinations of findings and various cutoff values are considered for probable infection or to rule out infection. Furthermore, in the absence of a referent standard for diagnosis, it is difficult to determine the true accuracy of any of the proposed schema for diagnosing PJI.
Additional caveats include potential variation in accuracy of diagnostic lab criteria based on the timing of the laboratory studies relative to surgery, because the diagnostic lab criteria can be influenced by the post-operative state itself or variations in surgical management. Furthermore, while PJI of the hips and knees are the most common and widely studied, the utility of the diagnostic criteria may differ among other arthroplasties.
Clinical Signs
PJIs can present in many ways, ranging from asymptomatic loosening of the joint, to fever, joint redness, and systemic sepsis. However, the presence of a sinus tract directly communicating with the joint, or external visualization of the prosthesis, is considered by diagnostic algorithms to be definitive for the diagnosis of PJI. Aside from intraoperative cultures obtained for the purpose of guiding antimicrobial therapy, further diagnostic studies are not considered necessary.(Parvizi, Zmistowski et al. 2011, Tsaras, Osmon et al. 2012, Parvizi, Gehrke et al. 2014, Gomez-Urena, Tande et al. 2017, Parvizi, Tan et al. 2018, McNally, Sousa et al. 2021) We emphasize that there are no good data to establish the accuracy of the presence or absence of a sinus tract, nor to validate it as a referent standard for diagnosis. However, it is a commonly accepted standard in clinical practice.
Imaging Studies
Published sensitivity and specificity of individual studies for diagnosing PJI (Table 1) should be considered cautiously given the absence of an optimal referent standard. Thus, these numbers are uncertain estimates, which may also account for the wide variations in reported sensitivity and specificity.
Several reviews have concluded that WBC scintigraphy and MRIs may have the best overall accuracy of the various radiologic techniques for detecting PJI.(Sconfienza, Signore et al. 2019, Romano, Petrosillo et al. 2020) Plain X-rays are not accurate for the diagnosis, with sensitivity and specificities reportedly as low as 14% and 70%, respectively.(Sconfienza, Signore et al. 2019) X-rays may be indicated, however, in the evaluation of a painful prosthetic knee joint to exclude non-infectious pathology and hardware complications. CT scans may have superior accuracy, although data are limited, and the scatter from the prosthetic material may affect interpretability.
Nuclear medicine scintigraphy and tagged white cell scans have had sensitivities and specificities for PJI ranging from 69%-94% across numerous studies.(Sconfienza, Signore et al. 2019) Tagged white cell scans had the highest accuracy among nuclear medicine studies, with sensitivities and specificities >90%.(Erba, Glaudemans et al. 2014, Sconfienza, Signore et al. 2019, Teiler, Ahl et al. 2020) PET and SPECT scans may have promise in diagnosing PJI; however, the data are limited and somewhat mixed. In one study of 130 patients with painful prosthetic hips, PET scan was 95% sensitive but only 39% specific for detecting PJI.(Kiran, Donnelly et al. 2019) Plate et al. reported a sensitivity of 78% and specificity of 94% for SPECT for diagnosing osteomyelitis, including PJI among 26 cases.(Plate, Weichselbaumer et al. 2020) In contrast, Wenter et al. reported a sensitivity of 86% but a specificity of only 67% for diagnosing 101 cases of PJI among 215 patients with prosthetic complications.(Wenter, Muller et al. 2016)
In a systematic review of 13 studies evaluating the accuracy of various diagnostic tests for PJI, PET scans had a sensitivity and specificity ranging from 80%-90% each.(Ahmad, Shaker et al. 2016) In another systematic review of 11 studies, PET scans had a sensitivity and specificity of 82% and 87%, respectively, but with statistical heterogeneity between the included studies.
In a review of four studies of MRI for PJI, the sensitivity ranged from 65%-92% and specificity ranged from 85%-99% for knee PJI, and while sensitivity and specificity were 94% and 97% for hip PJI.(Sconfienza, Signore et al. 2019) Other studies have been concordant, with sensitivities of 78%-86% and a specificity of 73%-90%.(Galley, Sutter et al. 2020, Schwaiger, Gassert et al. 2020)
Inflammatory Biomarkers
Ahmad et al. conducted a meta-analysis from 278 clinical studies comprising 27,754 patients with PJI and found that the pooled sensitivity and specificity for diagnosing PJI was only 75% and 70% for ESR and 88% and 74% for CRP.(Ahmad, Shaker et al. 2016) IL-6 accuracy was higher, with a sensitivity and specificity of 97% and 91%. However, IL-6 levels are not available in most hospital laboratories. Finally, Berbari et al. conducted a meta-analysis of 30 studies (n = 1,270 patients with PJI) and reported pooled sensitivity and specificities of 75% and 87% for ESR and 97% and 74% for CRP.(Berbari, Mabry et al. 2010) In other studies of PJI, CRP has been well described to be falsely negative, particularly for patients with more indolent pathogens.(Perez-Prieto, Portillo et al. 2017)
D-dimer, a fibrin degradation product, has recently been evaluated as a potential biomarker for PJI. For example, a single-center prospective study of 245 patients total undergoing primary or revision arthroplasty for aseptic or septic failure were included, and all had D-dimer, ESR, and CRP drawn pre-operatively.(Shahi, Kheir et al. 2017) Using a cutoff of 850 ng/ml, D-dimer was found to have a sensitivity of 89% and a specificity of 93%, significantly higher than either ESR or CRP in this study. While this biomarker may be promising, the fact that D-dimer is known to be elevated in many conditions, including various types of thrombosis and hematoma, raises concerns about the generalizability of this single report. Further studies are needed to determine the validity of these findings.
Synovial Fluid Studies
Included in all of the major PJI algorithmic definitions are synovial fluid WBC count, and percentage of polymorphonuclear cells (PMNs) with varying diagnostic cut-offs proposed, based on numerous observational studies.(Parvizi, Zmistowski et al. 2011, Parvizi, Gehrke et al. 2014, Parvizi, Tan et al. 2018, Schwarz, Parvizi et al. 2019, McNally, Sousa et al. 2021) It is important to note that the various synovial WBC cutoffs for PJI are much lower than those for native septic arthritis.(Margaretten, Kohlwes et al. 2007) Additionally, the synovial fluid WBC count and differential have been shown to change over time, with synovial fluid WBC count and neutrophil percentages significantly elevated early in the post-operative period, which may lead to false positive results depending on the cutoff used.(Christensen, Bedair et al. 2013) Qu et al. performed a meta-analysis including 15 studies and 2,787 patients and found pooled sensitivity/specificity for diagnosing PJI using synovial fluid WBC count was 88%/93%, and for synovial fluid PMN% the sensitivity/specificity were 90%/88%, respectively.(Qu, Zhai et al. 2014)
At the time of arthrocentesis, synovial fluid is often sent for microbiologic culture. There is significant heterogeneity in culture techniques across institutions, microbiology labs, and internationally, and thus the diagnostic utility of culture can vary. However, in a meta-analysis by Lee et al. of five studies including 509 patients evaluating the utility of culture in diagnostic arthrocentesis, the sensitivity was found to be poor at 62%, but the specificity was high at 94%.(Lee, Koo et al. 2017) It is unknown whether direct inoculation into blood culture bottles might improve diagnostic yield as it does for other sterile sites.
Synovial Fluid Leukocyte Esterase (LE)
LE is an enzyme secreted by activated neutrophils at the site of an infection and can be detected on a colorimetric test strip similar to that used in the detection of urinary tract infections, with the advantage of being inexpensive and providing real-time results if used in a point-of-care fashion.(Parvizi, Jacovides et al. 2011) Li et al. recently published an updated meta-analysis evaluating the diagnostic accuracy of LE for PJI.(Li, Zhang et al. 2020) They identified 17 studies involving a total of 1,963 patients (including 571 PJIs) and obtained a pooled sensitivity and specificity of 90% and 96%, respectively, for diagnosing PJI. A small study of 61 patients demonstrated the test retained its specificity even in the setting of adverse local tissue reactions seen after metal-on-metal total hip arthroplasty, which is known to potentially generate purulent synovial fluid.(Tischler, Plummer et al. 2016) One caveat is that the test is invalidated by blood contamination in synovial fluid without centrifugation prior to use.(Gomez-Urena, Tande et al. 2017) It has been included in the recent MSIS and ICM definitions for PJI.
Other Synovial Fluid Biomarkers
Alpha defensin is a small peptide that is also secreted by activated neutrophils in the setting of infection, exhibiting antimicrobial effects against a spectrum of pathogens.(Gomez-Urena, Tande et al. 2017) In the same meta-analysis mentioned above by Li et al., the diagnostic validity of alpha defensin in PJI was examined.(Li, Zhang et al. 2020) The review identified 21 studies with a total of 1,928 patients (and 650 PJIs), of which eight studies used a lateral flow assay, 12 studies used a laboratory-based immunoassay, and one study did not report on the testing methods. The pooled sensitivity and specificity of alpha defensin for diagnosing PJI were 89% and 96%, though significant heterogeneity was observed between samples due to differences in patient sample size and method of detection. False positives can occur in metallosis or in acute gout.(Sayan, Kopiec et al. 2021) This test is included in the updated MSIS and ICM criteria for PJI, though its use in arthroplasty at sites other than the hip or knee, or its cost effectiveness are not yet known.(Gomez-Urena, Tande et al. 2017)
While serum and synovial fluid levels of CRP have been shown to correlate, a recent meta-analysis by Wang et al. included six studies comprising a total of 456 participants, and found a pooled sensitivity and specificity of 92% and 90% of synovial fluid CRP for PJI, which was superior to serum CRP.(Wang, Wang et al. 2016) However, there was heterogeneity in the platforms and cutoff values used, and larger studies are needed to confirm the utility prior to its implementation routinely in the diagnosis of PJI. Similarly, synovial fluid IL-6 has been to be more specific than serum IL-6 levels. A meta-analysis of 17 studies describing PJI diagnosis using serum and synovial fluid IL-6 demonstrated that synovial fluid IL-6 had a sensitivity and specificity of 91% and 90%, which was notable for a higher sensitivity than serum IL-6 and comparable specificity.(Xie, Dai et al. 2017) IL-6 is likely not available for use routinely in most clinical laboratories, but may be in the future if further studies evaluate the optimal cutoff for use. While many of these synovial fluid biomarkers show promise, it is not yet known whether they improve the diagnosis of PJI compared to more conventional tests such as synovial WBC count, PMN%, and histopathology, or whether they will prove to be cost-effective.
Intraoperative Testing: Histopathology
Histologic exam of intraoperative frozen section to assess for acute inflammation is another diagnostic tool used by surgeons, particularly when pre-operative results are equivocal for PJI, or at time of revision surgery to avoid implanting a new joint into an infected site. In 2013, Tsaras et al. performed a systematic review and meta-analysis of studies comparing the performance of frozen section histology to simultaneously obtained microbiologic culture at the time of revision hip or knee arthroplasty.(Tsaras, Osmon et al. 2012) The review of 26 studies, including 3,269 patients of which 796 (24.3%) had a culture-positive PJI, found that the positive likelihood ratio was an impressive 12.0 for ruling in PJI, whereas the negative likelihood ratio was a less impressive 0.23. They reported no difference when comparing studies using thresholds of five vs. ten PMNs per high-power field. There was significant heterogeneity among pooled studies, which is at least partially reflective of the highly operator-dependent nature of frozen section sample preparation and interpretation. Another meta-analysis by Zhao et al. also found no statistical difference in the diagnostic odds ratio when comparing a cutoff threshold of five vs. ten PMNs per high-power field.(Zhao, Guo et al. 2013) Thus, data suggest that a diagnostic threshold of either five or ten PMNs per high-power field in each of five high-power fields can help diagnose or rule out PJI at the time of revision arthroplasty. It is unknown whether these thresholds apply to joints other than the hip or knee, or the performance other than at the time of revision arthroplasty.(Barrack, Bhimani et al. 2019)
Additionally, lower virulence organisms, such as Cutibacterium acnes (formerly Propionibacterium acnes), may fail to induce a neutrophil response or acute inflammation, and consequently, the sensitivity of this method in these situations is likely lower. However, neutrophilic infiltrates and, thus false positive results, can also be seen in the setting of periprosthetic fracture or inflammatory arthritis in the absence of infection.(Gomez-Urena, Tande et al. 2017)
Culture and Gram Stain
A prospective study of 117 patients who underwent revision hip or knee arthroplasties, performed for septic or aseptic reasons, compared the performance of tissue cultures vs. swabs in diagnosing PJI. The study reported a higher accuracy for tissue cultures relative to swab cultures (sensitivity 93% vs. 70%, and specificity 98% vs. 89%, respectively).(Aggarwal, Higuera et al. 2013) In a prospective study by Atkins et al. evaluating 297 patients who underwent revision hip or knee replacement at a single institution, three or more positive cultures were reported to have a sensitivity of 66% and specificity of 99.6% when compared to the presence of acute inflammatory cells in specimens examined histologically.(Atkins, Athanasou et al. 1998) Through the use of mathematical modeling, they suggested that obtaining five or six intraoperative tissue specimens for culture would result in a sensitivity of >80% and a specificity of >90% for detecting PJI with two or more specimens positive for the same organism. They also found that Gram stains had a very low sensitivity of only 6%, though with a specificity of >99%. Thus, negative Gram stains or culture results are not recommended to be used to rule out PJI.
Special Methods and Molecular Diagnostics
Few observational studies have assessed the role of special methods and molecular diagnostics for PJI. Sonication of device material removed or debrided during PJI surgical management may be used in microbiology laboratories to culture etiologic pathogens.(Gamie, Karthikappallil et al. 2021) Stephan et al. evaluated 90 patients with PJI to determine if prior antibiotics affected the yield of sonication-based culture methods.(Stephan, Thurmer et al. 2021) They found that cultures were positive in 86%, 81%, and 87% of patients who received peri-operative antibiotic prophylaxis, therapeutic antibiotics for ≥1 day prior to surgery, or no antibiotics prior to surgery, respectively. Thus, they reported no impact of prior antibiotics on sonication-based culture yield for PJIs. A more recent method to liberate bacteria from beneath biofilms in lieu of sonication involves addition of dithiothreitol to the prosthetic material.(Gamie, Karthikappallil et al. 2021) That method is less established than sonication but may result in similar diagnostic yield, which may be higher than culture results without sonication.(Gamie, Karthikappallil et al. 2021) Indeed, either sonication and addition of dithiothreitol has been shown to result in higher positive culture rates compared to cultures without these biofilm disruption methods.(Ahmad, Shaker et al. 2016, Prieto-Borja, Aunon et al. 2018, Sambri, Cadossi et al. 2018, Sebastian, Malhotra et al. 2018, Tani, Lepetsos et al. 2018)
However, studies are not uniform, and some have indicated that sonication does not increase culture yield compared to adequate culture of periprosthetic tissue.(Dudareva, Barrett et al. 2018, Yan, Karau et al. 2018) Furthermore, there is extra cost and technician time required for sonication and dithiothreitol methods, and this extra cost and time may or may not meet cost-effectiveness thresholds in various clinical settings.(Romano, Trentinaglia et al. 2018)
A recent systematic review of more modern concepts discussed PCR, sequencing, and metagenomics methods for establishing the microbial etiology of PJI.(Gamie, Karthikappallil et al. 2021) In individual studies, molecular diagnostics have been able to achieve higher rates of microbial identification than traditional culture. As reviewed,(Gamie, Karthikappallil et al. 2021) some studies have reported that multiplex PCR and next generation metagenomic sequencing not only to have superior sensitivity and high specificity compared to traditional culture methods, they were also faster than traditional microbiological methods.(Stylianakis, Schinas et al. 2018, Cai, Fang et al. 2020, Huang, Li et al. 2020, Lausmann, Kolle et al. 2020, Suren, Feihl et al. 2020) However, the data are mixed, as multiple other studies have found that traditional culture performed similarly to molecular methods.(Bemer, Plouzeau et al. 2014, Ryu, Greenwood-Quaintance et al. 2014, Larsen, Khalid et al. 2018, Malandain, Bemer et al. 2018, Sebastian, Malhotra et al. 2018, Yan, Karau et al. 2018, Lane, Ganeshraj et al. 2019, Wang, Huang et al. 2020, Gamie, Karthikappallil et al. 2021)
In a meta-analysis, the sensitivity and specificity of 16s RNA PCR was pooled across 15 observational studies of patients with PJI.(Zhang, Feng et al. 2020) The pooled (95% CI) sensitivity and specificity were 70% (67%-73%) and 93% (91%-94%). Sonication of the culture material before application of 16s RNA slightly increased accuracy, with a sensitivity and specificity of 76% and 93%, respectively. In a second meta-analysis of 12 studies of 16s RNA PCR, the pooled (95% CI) sensitivity was 81% (73%-87%) and specificity was 94% (94%-97%).(Li, Zeng et al. 2019) Antecedent antibiotics reduced the sensitivity of the PCR assay (71% vs. 94%). Furthermore, the study found that sensitivity varied based on the method used, with Illumina sequencing achieving higher specificity than other methods (96% vs. 83%). In a third meta-analysis of nine studies of sonication plus PCR, pooled sensitivities and specificities were 75% (95% CI, 71%-81%) and 96% (95% CI, 94%-97%).(Liu, Fu et al. 2018) Thus, the sensitivity and specificity of such molecular methods for establishing the microbial etiology of PJI appears to be approximately 70%-75% and 90%-95%, respectively.
The primary advantage of broader, non-biased molecular sequencing methods may be to identify unusual or fastidious organisms that are difficult to culture by traditional methods.(Ivy, Thoendel et al. 2018, Thoendel, Jeraldo et al. 2018) However, a complication of these results is that it can be difficult to determine if the detected organism is an etiologic pathogen, and no reference standard is available to clarify this issue.
Test |
Sensitivity |
Specificity |
+LR* |
-LR* |
Reference |
|
Osteomyelitis without PJI | ||||||
X-rays |
70% |
82% |
3.9 |
0.4 |
(Llewellyn, Jones-Diette et al. 2019) |
|
CT Scans |
70% |
90% |
7.0 |
0.3 |
(Llewellyn, Jones-Diette et al. 2019) |
|
MRI |
96% |
81% |
5.1 |
0.05 |
(Llewellyn, Jones-Diette et al. 2019) |
|
Nuclear Medicine Scintigraphy† |
84% |
71% |
2.9 |
0.2 |
(Llewellyn, Jones-Diette et al. 2019) |
|
White Cell Tagged Scans |
87% |
95% |
17.4 |
0.1 |
(Llewellyn, Jones-Diette et al. 2019) |
|
PET |
85% |
93% |
12.1 |
0.2 |
(Llewellyn, Jones-Diette et al. 2019) |
|
SPECT |
95% |
82% |
5.3 |
0.06 |
(Llewellyn, Jones-Diette et al. 2019) |
|
ESR |
49%-79% |
50-80% |
1.6-3.8 |
0.3-0.4 |
(Ryan, Ahn et al. 2019, Wu, Lu et al. 2020, Ghassibi, Yen et al. 2021) |
|
CRP |
45%-76% |
59%-71% |
1.1-2.6 |
0.3-0.8 |
(Ryan, Ahn et al. 2019, Wu, Lu et al. 2020, Ghassibi, Yen et al. 2021) |
|
Biopsy (histopathology) |
52% |
>99% |
>50 |
0.5 |
(Pupaibool, Vasoo et al. 2015) |
|
DFO |
||||||
X-rays |
62% |
78% |
2.8 |
0.5 |
(Llewellyn, Kraft et al. 2020) |
|
MRI |
93%-96% |
75%-84% |
3.7-6.0 |
0.05-0.09 |
(Lauri, Tamminga et al. 2017, Llewellyn, Kraft et al. 2020) |
|
Nuclear Medicine Scintigraphy† |
85% |
68% |
2.7 |
0.2 |
(Llewellyn, Kraft et al. 2020) |
|
White Cell Tagged Scans |
91%-92% |
75%-92% |
3.6-11.5 |
0.09-0.1 |
(Lauri, Tamminga et al. 2017) |
|
PET |
84% |
93% |
12.0 |
0.2 |
(Llewellyn, Kraft et al. 2020) |
|
ESR |
60%-81% |
56%-90% |
1.4-8 |
0.2-0.7 |
(Victoria van Asten, Geradus Peters et al. 2016, Lavery, Ahn et al. 2019, Moallemi, Niroomand et al. 2020, Xu, Cheng et al. 2020) |
|
CRP |
49%-76% |
55%-80% |
1.1-3.8 |
0.3-0.9 |
(Markanday 2015, Lavery, Ahn et al. 2019, Moallemi, Niroomand et al. 2020, Xu, Cheng et al. 2020) |
|
Probe-to-bone |
87% |
83% |
5.1 |
0.2 |
(Lam, van Asten et al. 2016) |
|
PJI‡ |
||||||
X-rays |
14% |
70% |
0.5 |
1.2 |
(Sconfienza, Signore et al. 2019) |
|
MRI |
65%-94% |
73%-99% |
2.4->50 |
0.06-0.5 |
(Sconfienza, Signore et al. 2019, Galley, Sutter et al. 2020, Schwaiger, Gassert et al. 2020) |
|
Nuclear Medicine Scintigraphy† |
83%-94% |
69%-90% |
2.7-9.4 |
0.07-0.2 |
(Nagoya, Kaya et al. 2008, Ikeuchi, Okanoue et al. 2013, Ouyang, Li et al. 2014) |
|
White Cell Tagged Scans |
93%-100% |
91%-100% |
10->50 |
0.08-<0.01 |
(Erba, Glaudemans et al. 2014, Teiler, Ahl et al. 2020) |
|
PET |
82%-95% |
39%-87% |
1.3-7.3 |
0.06-0.5 |
(Kwee, Kwee et al. 2008, Jin, Yuan et al. 2014, Kiran, Donnelly et al. 2019) |
|
ESR |
75% |
70%-87% |
2.5-5.8 |
0.3-0.4 |
(Berbari, Mabry et al. 2010, Perez-Prieto, Portillo et al. 2017) |
|
CRP |
88%-97% |
74% |
3.4-3.7 |
0.04-0.2 |
(Berbari, Mabry et al. 2010, Perez-Prieto, Portillo et al. 2017) |
|
IL-6 |
97% |
91% |
10.8 |
0.03 |
(Berbari, Mabry et al. 2010) |
|
Synovial WBC Count |
88% |
93% |
12.6 |
0.1 |
(Qu, Zhai et al. 2014) |
|
Synovial PMN% |
90% |
88% |
7.5 |
0.1 |
(Qu, Zhai et al. 2014) |
|
Synovial Culture |
62% |
94% |
10.3 |
0.4 |
(Lee, Koo et al. 2017) |
|
PJI, prosthetic joint infection; LR, likelihood ratio; CT, computerized tomography; PET, positron emission tomography; SPECT, single photon emission computed tomography; MRI, magnetic resonance imaging; ESR, erythrocyte sedimentation rate; CRP, C-reactive protein rate; DFO, diabetic foot osteomyelitis; IL-6, Interleukin-6; WBC, white blood cell; PMN, polymorphonuclear *A positive LR ≥5 is helpful and ≥10 is very helpful at shifting post-test probabilities; a negative LR ≤ 0.2 is helpful and ≤ 0.1 is very helpful at shifting post-test probabilities. †Excluding tagged white cell studies, which are considered separately. ‡Because there is no identified optimal referent standard for the diagnosis of PJI, sensitivity, specificity, and LRs for tests for PJI should be considered to be uncertain estimates. |