 |
 |
 |
|
| home | CME | help | feedback | subscriptions | archive |
|
The Journal of Bone and Joint Surgery (American). 2006;88:869-882. doi:10.2106/JBJS.E.01149 © 2006 The Journal of Bone and Joint Surgery, Inc.
Diagnosis of Periprosthetic Infection
1 Departments of Pathology and Orthopaedic Surgery, The Cleveland Clinic
Foundation, L25, 9500 Euclid Avenue, Cleveland, OH 44195. E-mail address for
T.W. Bauer:
osteoclast{at}aol.com
Investigation performed at the Departments of Pathology and Orthopaedic
Surgery, The Cleveland Clinic Foundation, Cleveland, Ohio
Periprosthetic infections are rare, but there is evidence to suggest that their frequency may be underestimated. No single laboratory test has perfect sensitivity and specificity for diagnosing infection. Most tests have better specificity when they are performed for patients in whom infection is suspected clinically rather than when they are used as screening tests. Screening test results that may suggest the possibility of infection include elevation of the erythrocyte sedimentation rate and/or serum C-reactive protein level more than three months after an arthroplasty. Most serologic tests are difficult to interpret when the patient has an underlying inflammatory arthropathy. Cultures of aspirated joint fluid can be especially helpful for patients who have symptoms suggestive of infection, but their results are best interpreted two weeks after administration of antibiotics has been discontinued. Joint fluid cell counts may also be helpful, but Gram stains of joint fluid have poor sensitivity and specificity. Criteria for diagnosing infection on the basis of frozen sections of implant membranes have not yet been standardized, but in many laboratories more than five neutrophils per high-power field in five or more fields (excluding surface fibrin) has been found to be suggestive of infection. Most polymerase chain reactions that detect the universal 16S rRNA bacterial gene have problems with false-positive results, but combining a universal polymerase chain reaction with subsequent bacterial sequencing can help improve specificity. Polymerase chain reactions can detect necrotic bacteria, so the clinical importance of positive results of this analysis in the absence of other features of infection remains to be determined.
There have been important improvements in total joint arthroplasty in terms of implant design, fixation, and control of periprosthetic infection. The use of prophylactic antibiotics, body exhaust systems, laminar airflow, and other precautions has helped reduce the prevalence of clinically recognized periprosthetic infection from nearly 10% in the early years in which arthroplasty was performed1 to <1% in some series2,3. Despite this decline, periprosthetic infection remains one of the most challenging complications of joint arthroplasty and is associated with immense physiological, psychological, and financial costs. Furthermore, several recent observations have suggested, but have not proven, that some arthroplasty failures that were interpreted as being due to aseptic loosening might in fact have represented the consequence of inflammatory reactions to bacteria or bacterial products. These observations include (1) the finding that antibiotic-containing bone cement protects against so-called aseptic loosening4,5, (2) evidence of bacteria on a surprisingly high proportion of implants that had been revised because of aseptic loosening6,7, (3) occasional cases in which implant membranes showed acute inflammation but intraoperative cultures were negative8,9, and (4) emerging data suggesting that bacterial endotoxin and related molecules may have a role in particle-induced bone resorption10-14. The purpose of this article is to review our current understanding of periprosthetic infection with particular focus on the efficacy of various tests to help make the diagnosis.
A fundamental issue in determining the prevalence of a disease is defining the criteria with which the disease can be diagnosed with certainty. Currently, periprosthetic infection is most frequently diagnosed by isolation of one or more organisms from the periprosthetic tissue or fluid with use of conventional microbiologic culture techniques, and the results of microbiologic culture are usually considered the standard with which other diagnostic tests are compared. However, organisms are not always isolated from areas that ultimately prove to be infected, and sometimes positive cultures of specimens of periprosthetic tissue may not represent clinically important infections15 (Table I) because specimens can become contaminated when the tissue is being harvested, being transported, or in the laboratory. In addition to microbiologic culture of tissue or fluid, other tests are used to help diagnose periprosthetic infection. However, all diagnostic tests have limitations, and the sensitivity, specificity, and predictive value of positive and negative test results are usually calculated with respect to an existing reference standard (the "gold standard") (Fig. 1)16-18. Because of the aforementioned limitations of diagnostic tests, clinicians often utilize a combination of tests to confirm or exclude the diagnosis of periprosthetic infection. Developing a definition of infection that is robust enough to serve as a gold standard is an ongoing challenge that influences our perception of the value of any diagnostic test that is compared with that gold standard. The prevalence of infection in a cohort of patients also influences the predictive value of positive and negative test results. Recognizing the limitations of using a reference standard for comparison, many investigators have attempted to evaluate the efficacy of various tests for diagnosing periprosthetic infection, as discussed below.
Infections at the sites of total joint arthroplasties are sometimes categorized on the basis of the presumed mechanism and timing of the infection. So-called acute postoperative infections are thought to result from organisms that gained access to the joint during the operation, or soon after it, from the overlying skin or a draining wound. Infections of this type generally become symptomatic within a few days or weeks after the arthroplasty. So-called late chronic infections may result from organisms introduced during the operation, either from the air, from surgical instruments, or from the implant itself. The lag period is the time needed for the organisms to proliferate and induce symptoms that prompt recognition of the infection. Hematogenous infections are the result of the seeding of an arthroplasty site by organisms carried by the bloodstream from a different site (e.g., a urinary tract infection or a cutaneous or mucosal ulcer). The distinction between these types of infection may be difficult and is somewhat arbitrary. While early reviews suggested that the majority of arthroplasty-related infections were the consequence of wound contamination19, more recent studies have suggested that late infections are much more common. For example, in a retrospective review of more than 6000 total knee replacements, Peersman et al.3 reported an overall deep infection rate of 0.39% following primary arthroplasties and 0.97% following revision operations. One-third of the deep infections occurred within the first three months after the operation, and the remaining cases were considered late infections. In a study of more than 3000 total hip arthroplasties performed over a sixteen-year period, Schmalzried et al.20 noted that the incidence of hematogenous arthroplasty-related infection increased during the time that the cohort was followed. This change from acute to chronic infections presumably reflects changes in surgical practice during recent decades, including the use of prophylactic antibiotics, the use of antibiotic-impregnated cement, and alterations in the operating room environment2,21-23.
Clinical Factors A detailed clinical history and physical examination constitute the most important ways to recognize a potential periprosthetic infection. The type and duration of symptoms, details of the postoperative course, the presence of comorbidities, and the types of treatments rendered should be discussed in detail. Periprosthetic infection may be diagnosed with reasonable certainty on the basis of the history and clinical presentation when there are classic signs of infection such as severe joint pain, fever, chills, or a draining periarticular sinus. In such cases, laboratory tests are used simply to confirm the diagnosis of the periprosthetic infection. However, periprosthetic infection has an innocuous presentation in most patients and may be difficult to diagnose on the basis of the history and physical findings alone. Many of the symptoms and signs of infection overlap with those of other clinical conditions such as intra-articular hematoma, instability, and aseptic loosening. It is under these circumstances that additional diagnostic modalities play a critical role in the confirmation or exclusion of the diagnosis of periprosthetic infection.
Radiographic Studies After a physical examination, evaluation of a patient with a loose or painful prosthetic joint commences with radiographic studies. There are a few nonspecific changes suggestive of infection that may be apparent on plain radiographs. These include periosteal reaction, scattered foci of osteolysis, or generalized bone resorption in the absence of implant wear (Fig. 2-A). In general, however, the majority of patients with periprosthetic infection, especially those with an acute presentation, do not have obvious radiographic findings suggestive of infection or may show features indistinguishable from those seen in association with aseptic loosening24. The main role of conventional radiographic evaluation of these patients is to rule out other conditions such as wear and osteolysis or fractures.
Radionuclide Imaging Gallium-67 is bound in serum to iron-transporting molecules such as transferrin. It is transported to tissues on the basis of vascularity, inflammation, and other factors. Gallium-67 scans alone have a low sensitivity for diagnosing infection27,34. The demonstration of congruent patterns by gallium-67 and technetium-99 scans often reflects aseptic changes around implants, but a lack of congruence (i.e., positive scans with different spatial distributions) can be seen when there is an infection35. Technetium-99m-polyclonal IgG (immunoglobulin G) scintigraphy has been reported to have a high sensitivity for recognizing infections around hip and knee prostheses, but like many types of scans it has a low specificity36. The role of fluorodeoxyglucose-positron emission tomography (FDG-PET) scans in the diagnosis of infections at the sites of arthroplasties has been evaluated at some centers. Inflammatory cells metabolize predominantly glucose, and the uptake of glucose is enhanced when such cells are stimulated. Activated macrophages and neutrophils express high concentrations of glucose transporters, which facilitate the movement of FDG (as well as glucose) through the cell membrane. Deoxyglucose is phosphorylated to deoxyglucose-6-phosphate, which is not a substrate for glucose-6-phosphate dehydrogenase so it becomes trapped in tissue long enough to allow PET imaging. Thus, FDG reflects glucose utilization and can indicate areas of inflammation. Studies have shown combined FDG-PET imaging to have variable sensitivity and specificity for diagnosing periprosthetic infection30,37. One study, for example, demonstrated approximately 91% sensitivity and 72% specificity for diagnosing infections around knee prostheses and 90% sensitivity and 89% specificity for diagnosing infections around hip prostheses37. Although FDG-PET scans may have greater specificity than leukocyte-labeling bone scans, false-positive results may occur as a result of uptake of FDG in particle-induced inflammation around implants with aseptic loosening38.
Serologic Tests Another serologic test that has shown promise for diagnosing infection is measurement of the serum level of interleukin-6 (IL-6), a factor produced by monocytes and macrophages. In a recent study, the serum level of IL-6 was found to be consistently elevated (>10 pg/mL [>10 ng/L]) in patients with periprosthetic infection, and it had a higher predictive value than most other serologic markers51. A potential advantage of measuring the IL-6 level is that the level returns to normal soon (within forty-eight hours) after the operation and is not likely to be elevated in patients with aseptic loosening. However, it may be elevated in patients with an underlying inflammatory arthropathy.
Culture of Aspirated Joint Fluid There are several possible reasons for the difference in the predictive values between the above studies52,53. One possible reason is that one study dealt with hips and the other, with knees. False-positive test results may be more common in fluids aspirated from hips than in those aspirated from knees. On the other hand, the prevalence of infection in the second study (29%) was much higher than that in the first (2%), presumably because the test was applied to all patients undergoing revision arthroplasty in the first study but was limited to patients with "symptomatic" knee replacements in the second. The important effect of prevalence on calculations of predictive values is illustrated by using the Bayesian equation to calculate the positive predictive value54 (see Appendix). Including prevalence in the calculation yields a positive predictive value of only 15% in the 1993 study of hip fluid aspirations but a value of 72% in the 1997 study of knee aspirations. These calculations illustrate that the predictive value of a positive result of a culture of joint fluid is higher if the study is not used as a screening test for infection but is used instead as a confirmatory test for patients in whom clinical findings (or prior laboratory test results) have already raised the suspicion of infection. Very similar findings were described by Spangehl et al.45, who also recommended culture of aspirated fluid when a prior screening test, such as measurement of the erythrocyte sedimentation rate or C-reactive protein level, is positive. The sensitivity of cultures of aspirated fluid is increased by repeating the test for patients who had a negative result on prior culture of aspirated fluid but for whom there is a strong clinical suspicion of periprosthetic infection53. The sensitivity is greatly reduced when the test is performed for patients receiving antibiotic treatment53. To minimize the influence of antibiotics, joint aspiration is best performed at least two weeks after the last dose of antibiotics has been given. Although aspiration of the knee can be performed without the use of fluoroscopy, the hip joint cannot be aspirated accurately unless fluoroscopy is utilized. Radiographic confirmation of appropriate needle placement is essential for joint aspiration of the hip and sometimes for aspiration of the knee.
Gram Stains of Aspirated Joint Fluid
Joint Fluid Leukocyte Counts
In 2003, Mason et al. retrospectively reviewed data on 440 revision total knee arthroplasties and identified eighty-six patients who had presented with clinical features suspicious for infection and had therefore undergone joint fluid aspirations60. The mean white blood-cell count for the fifty knees that were found to be uninfected was 645 ± 878/mL (about 6/µL), whereas the mean count for the thirty-six infected knees was 25,951/mL (260/µL). There was a mean of 72.8% ± 28.6% neutrophils in the infected knees and 27% ± 24% in the uninfected ones. The authors suggested that the optimum criteria for diagnosing infection included a white blood-cell count of >2500/mL and >60% neutrophils60. Trampuz et al.61 prospectively evaluated synovial fluid specimens from ninety-nine patients with aseptic failure of a total knee prosthesis and from thirty-four patients with an infection at the site of a total knee arthroplasty. Using receiver operator characteristic curves, the authors estimated that a synovial fluid leukocyte count of 1.7 x 103/µL or a differential count of >65% neutrophils was the optimum cutoff for a diagnosis of infection61. As seen in Table II, the disparity in reported cell concentrations suggests that some authors may not have reported the correct units of volume. Setting aside the inconsistencies in units, there are still discrepancies with regard to the level at which the cell count in fluid from the site of a prosthetic joint may be considered abnormal. From a practical standpoint, we consider a white blood-cell count of >500/µL as suggestive of periprosthetic infection.
Efficacy of Analysis of Frozen Sections for Diagnosis
Perhaps the first study of the use of frozen sections to diagnose an infection at the site of an arthroplasty was reported by Charosky et al. in 197362. Those authors described the results of analysis of frozen sections of implant membranes obtained from twenty patients, ten of whom had intraoperative cultures that were positive for organisms and ten of whom had negative cultures. Of the ten with positive cultures, five had acute inflammation that was "2+ or greater" (not otherwise defined) and the other five had chronic inflammation that was "2+ or greater." The authors concluded that acute inflammatory changes or "severe chronic inflammation" were presumptive evidence of infection. Another early study, and probably the most frequently quoted (and misquoted), on this topic was performed by Mirra et al. and published in slightly different forms in 197663 and 198264. In the first publication63, the authors noted that, of more than 550 total joint arthroplasties performed between 1970 and 1974 at a single center, an unspecified number were revision arthroplasties. The authors retrospectively reviewed the histologic findings in membranes around twenty-four failed hip prostheses and ten failed knee prostheses and attempted to correlate those findings with the presumed mechanism of failure. There was no single gold standard for diagnosing infection; instead, the diagnoses of septic and aseptic loosening appear to have been based on a combination of radiographic features and culture results. The authors did not describe the criteria that they used to select the thirty-four cases for review. The extent of inflammation was quantified as the average number of cells in five different microscopic fields obtained from areas of maximal inflammation. Interestingly, the high-power microscopic field used in the study was a net magnification of 500x. Although 60x lenses are also available, the majority of microscopes in use today have a 40x objective lens and a 10x ocular lens, yielding a final magnification of 400x—i.e., 20% lower than the magnification used in the study by Mirra et al. In the original publication by Mirra et al.63, acute inflammation was graded as absent, 1+ (one to five cells per high-power field), 2+ (six to forty-nine cells per high-power field), or 3+ (fifty or more cells per high-power field). Lymphocytes and plasma cells were quantified similarly. All fifteen patients with positive cultures had 2+ or 3+ acute inflammation, although one of them did not have clinical evidence of deep infection. Neutrophils were not present (at least not at the 2+ level) in patients for whom the cultures were negative. The authors noted that patients with rheumatoid arthritis can have up to ten neutrophils per high-power field, but apparently two infections in patients with coexisting rheumatoid arthritis still could be diagnosed on the basis of frozen sections. In 1982, Mirra et al.64 expanded their original series to include the results of biopsies from 1970 to 1978, including those done during fifty-four revision hip operations, thirty-nine revision knee operations, and one revision of a silicone toe implant. Ninety-four cases were studied, including the thirty-four that had been previously described63. Of those ninety-four biopsies, twenty-two demonstrated areas of acute inflammation with more than five neutrophils per high-power field in five fields. Twenty-one of the joints with a positive biopsy result had a positive culture and one had a negative culture but was thought to be infected on the basis of clinical findings. Five joints had positive intraoperative cultures (with growth of Corynebacterium in four and Micrococcus in one) but no substantial acute inflammation, and the organisms were thought to have been either contaminants or as causing a "low-virulence" infection. The two publications by Mirra et al. are the origin of the commonly quoted criterion of five neutrophils per high-power field. It should be noted that the original articles describe five neutrophils in each of five microscopic fields from the area of highest cellularity, excluding superficial fibrin, in a patient who does not have rheumatoid arthritis. To our knowledge, the influence of the variability in magnification (with 500x used by Mirra et al. compared with the more commonly used 400x) has not been previously noted. Other authors have attempted to validate histologic criteria for the diagnosis of infection. For example, Fehring and McAlister65 performed a study of 107 consecutive total joint revisions in which all patients had analysis of frozen sections of tissue obtained from multiple surgical sites. Intraoperative cultures were performed for all patients, and at least two tissue blocks representing four sites were evaluated in each case. Unfortunately, the results of the frozen-section analysis were somewhat compromised by the authors' exclusion of ten patients, in part because their cases were difficult to classify on the basis of the extent of inflammation. The authors did not try to determine the concentration of inflammatory cells that was predictive of infection. Instead, cases were interpreted as positive if there was "evidence of acute inflammation characterized by the presence of polymorphonuclear leukocytes." The authors emphasized the importance of an overall histologic interpretation, rather than relying solely on a count of neutrophil concentration. Using the results of intraoperative cultures as the reference standard, Fehring and McAlister calculated the sensitivity and specificity of the frozen-section interpretation as well as of an overall histologic diagnosis based on analysis of frozen and permanent sections. Of ninety-seven cases that were retained in the study, eleven were found to be infected and eighty-six were not infected. There were nine false-positive and nine false-negative frozen sections, yielding a specificity of 89.5% and a sensitivity of only 18.2%. On the basis of the complete histologic analysis, there were twelve false-positive and two false-negative results, yielding a sensitivity of 82% and a specificity of 86%. Interestingly, there was ultimately a high clinical suspicion of infection in six patients with negative intraoperative cultures: two had draining sinuses, one had a positive culture of fluid obtained with joint aspiration, and three had had prior resection arthroplasties because of infection. Thus, this study could be interpreted as showing that frozen-section analysis has relatively poor sensitivity, especially if one considers the ten cases that were excluded. On the other hand, it also illustrates the problem of using intraoperative cultures as the reference standard instead of the final clinical diagnosis based on a combination of tests. Lonner et al.66 performed a prospective study similar to the one reported by Fehring and McAlister65. Frozen sections were obtained from at least two areas in each of 175 consecutive patients undergoing revision arthroplasty. The five most cellular fields were evaluated, and an infection was considered to be present if there was an average of five or more polymorphonuclear leukocytes in at least five high-power fields. The authors also recorded the cases with ten or more polymorphonuclear leukocytes per high-power field. An average of four or fewer polymorphonuclear leukocytes per high-power field was interpreted as indicating the absence of infection. Nineteen patients had positive intraoperative cultures. With the culture results used as the reference standard, there were three false-negative and seven false-positive histologic interpretations (a sensitivity of 84% and a specificity of 96%). Of the seven patients with a false-positive result, five had five to nine polymorphonuclear leukocytes per high-power field. If the authors had used ten cells per high-power field as the cutoff, there would have been only two false-positive histologic interpretations (specificity, 98%). Of note, seven of the positive intraoperative cultures were considered by the treating physicians to be probably due to contaminants. All of the patients with those cultures had negative histologic findings, and all were treated as if they did not have an infection. No signs of infection had developed in these seven patients after an average duration of twenty months of follow-up, a finding that illustrates the problem of using intraoperative culture results as the reference standard. In 1995, Athanasou et al.67 reported on a prospective study in which frozen sections from several different sites were obtained during each of 106 hip and knee revision arthroplasties performed between 1991 and 1993, and the results were compared with those of intraoperative cultures. In an evaluation of ten high-power fields with maximal inflammation, the authors quantified inflammatory cells into four tiers (absent, one, one to five, and more than five cells per field). Of note, lymphocytes and plasma cells were included along with neutrophils, but neutrophils entrapped in fibrin adherent to the surface of the membrane were excluded. Intraoperative cultures were considered positive if organisms grew on direct plating or if a similar strain grew on enrichment in more than one culture; single isolates from only one culture were considered to be negative findings. On the basis of the culture results, twenty-four arthroplasty sites were determined to be infected and eighty-four were considered to be not infected. Compared with these culture results, the frozen-section analysis yielded two false-negative and three false-positive results—a sensitivity of 90%, a specificity of 96%, and positive and negative predictive values of 88% and 98%. The authors noted that there were occasional lymphocytes in the thirty-six uninfected cases. These cells were often perivascular and were not regarded as suspicious for infection. In addition, three patients with underlying rheumatoid arthritis had numerous lymphocytes and plasma cells, and five patients with aseptic loosening and abundant metal particles also had moderate numbers of lymphocytes. While these patients were recognized as probably not having an infection, the authors noted that: "in the absence of rheumatoid disease, plasma cells were a good marker of infection, being noted in eight of the infected cases." Of the two patients who were considered to have a "false-positive" frozen section on the basis of a negative intraoperative culture, one had loosening eighteen months later and was found to have an infection at the repeat revision arthroplasty. The second patient also had a clinical course suggestive of infection, which again emphasizes the limitation of using intraoperative culture results as a reference standard. In 2000, Pandey et al.68 reported a study that appears to have overlapped, in part, with the study by Athanasou et al.67. Pandey et al. retrospectively reviewed the results of histologic tissue analysis and intraoperative cultures of specimens from 617 revision arthroplasties performed between 1992 and 1996 at several hospitals affiliated with the Oxford Skeletal Infection Research and Intervention Service. Although there was overlap among the authors of the two studies67,68, different criteria were used for the histologic diagnosis of infection. At least ten high-power fields were evaluated, and an average score for the various inflammatory cells was calculated68. One inflammatory cell per high-power field in at least ten fields was considered to be consistent with infection. For the intraoperative cultures, isolation of the same organism from three or more culture specimens was considered diagnostic of infection. Organisms were considered contaminants if different strains grew in different broths and there was no growth on direct plating. A single isolate was considered to be unimportant. Of the 617 revision arthroplasty sites, 526 were clinically suspected to be aseptic and ninety-one were suspected to be infected. Eighty-one were proven to be infected according to the microbiologic criteria noted above. Five hundred and twenty-one cases had no growth on culture and had negative histologic findings as only scattered lymphocytes were present (true-negative histologic findings). Both the cultures and the histologic analysis showed features of infection in seventy-nine cases (true-positive histologic findings). Two cases had "significant growth of organisms" on culture but negative histologic findings (false-negative histologic findings), and ten cases had negative cultures but acute inflammation in the peri-implant membrane. Seven of the ten patients had received preoperative antibiotics, and all ten were treated clinically as if they had an infection. Finally, five cases showed inflammation in the tissue but negative cultures. Two of these patients had rheumatoid arthritis and loosening developed within two years. As described above and in additional studies summarized in Table III57,69,70, criteria for interpreting microscope slides of frozen sections are not yet uniform. Considering a low number of neutrophils (for example, one cell per high-power field68) or even lymphocytes or plasma cells67 to be diagnostic of infection will provide maximum sensitivity but will be associated with false-positive diagnoses and hence decreased specificity. Use of more stringent criteria (for example, ten polymorphonuclear leukocytes per high-power field in at least ten high-power fields66) will improve specificity at the expense of sensitivity (Table III). Numeric criteria are complicated even more by differences in the visual field size of different microscopes. While most authors have used 10x ocular and 40x objective lenses (yielding a nominal net magnification of 400x), other differences in microscope and camera configurations can vary the visual field by as much as twofold. Therefore, the number of inflammatory cells per high-power field should be recognized as only an approximation.
Partly on the basis of the studies described above, we currently interpret a frozen section as being suggestive of infection if it contains at least five neutrophils in each of three 400x high-power microscopic fields located beneath the surface of the membrane (Figs. 2-A through 4-B). In the appropriate clinical setting, even fewer neutrophils should raise the suspicion of infection. Neutrophils entrapped in superficial fibrin (Fig. 5) or adherent to endothelial cells (marginating) are not thought to be diagnostic of infection, but neutrophils in fibrous tissue between the capillaries that compose granulation tissue may be predictive of infection. Frozen sections of tissue from a patient with an underlying inflammatory arthropathy such as rheumatoid arthritis are especially difficult to interpret because, in these patients, acute inflammation involves peri-implant membranes even in the absence of infection. Lymphocytes and plasma cells have been seen in biopsy specimens from patients who have been treated with antibiotics for infection, but these cells are currently thought to be nonspecific and in general not predictive of active infection. Inflammation is not uniformly distributed around the prosthesis, so frozen-section analysis of biopsy specimens taken from several different sites increases the sensitivity compared with that of an analysis of a single biopsy specimen. It is also important for the tissue submitted for frozen-section analysis to adequately represent the fibrous membrane and not contain only superficial fibrin. Although we continue to use the same histologic criteria for diagnosing active infection at the second stage of a two-stage revision arthroplasty done because of infection, the predictive value of these observations in this clinical context (after the use of local and systemic antibiotics) requires further study (as described below). Communication and feedback between the surgeon and pathologist are key to help both physicians to determine the clinical importance of inflammation in any given case.
Microbiologic Cultures of Tissue
Diagnosing Infection at the Time of Reimplantation
Endotoxin
Molecular Techniques Other new techniques that may have a role in diagnosing infection include the use of microarray89 and proteomics technologies. A microarray allows isolation and evaluation of numerous mRNA genes with a single test. Proteomics allows simultaneous isolation and evaluation of numerous proteins. The premise of these techniques is to identify organism-specific genes or proteins. The challenge for all of the new molecular tests will be to distinguish clinically important infections from trace levels of necrotic bacteria or contaminants and to provide that information quickly enough to be of practical help in guiding patient care.
The diagnosis of periprosthetic infection remains a challenging problem, as there is no single diagnostic modality with absolute sensitivity and specificity. Accurate diagnosis often requires the use of combinations of tests and a strong clinical suspicion. Serologic tests (measurements of white blood-cell count, erythrocyte sedimentation rate, and C-reactive protein level) represent the first-line investigation and generally have good sensitivity but lower specificity. Imaging, such as with a labeled white-blood-cell scan, may be used to further support a diagnosis of an infection when serologic findings are abnormal or in equivocal cases. Aspiration of the joint has high specificity and is especially valuable for diagnosing suspected infections of the knee. Intraoperative cultures should be performed for all patients suspected of having a periprosthetic infection. Extreme care should be exercised to prevent contamination of these samples. Analyses of intraoperative frozen sections have limitations, mostly related to the experience of the pathologist who interprets the sections and the sampling methods of the surgeon. In institutions with adequate pathology resources, interpretation of frozen sections can be very helpful at revision arthroplasty as well as at the time of reimplantation in a two-stage revision of an arthroplasty complicated by infection. Close communication between the surgeon and pathologist, with follow-up of borderline cases, helps the team of physicians to establish their own decision thresholds. Intraoperative cultures, although considered the gold standard, may be negative for some patients with clinically proven periprosthetic infection, and clinical acumen should be employed to override the negative or equivocal findings of diagnostic modalities in some cases. New molecular diagnostic methods will help to diagnose infections in the future.
A "predictive value calculator" is available on our web site at jbjs.org (go to the article citation and click on "Supplementary Material").
|
Top
Abstract
Introduction
