Periprosthetic infection occurs at a rate of approximately 0.5% or higher after primary total knee arthroplasty and 1% or higher after revision total knee arthroplasty, and this complication leads to marked morbidity and mortality for affected patients1. Treatment costs for revision arthroplasty due to infection were estimated to be over $80,000 per knee in a 1996 study and will be even higher today2. Establishing the proper role of preoperative serum tests and intraoperative tests such as Gram stains, frozen sections, and cultures in the diagnosis of infection represents a challenge3-5. Despite all efforts, some patients remain infected despite negative diagnostic testing during an attempt at definitive reimplantation.
Intraoperative pathologic evaluation of freshly frozen tissue is one of the tests with good specificity used in algorithmic approaches to diagnosing periprosthetic joint infection, but the wide range of sensitivity values reported for this test raises concerns about the effects of intraobserver and interobserver variability among pathologists on the consistency of diagnosis6. The proper technique for assessing the polymorphonucleocyte (PMN) count in a fresh-frozen sample has been investigated previously7-10, but little is known about how to utilize results from paraffin-embedded permanent samples. Several studies involving a limited number of patients have indicated concordance of over 95% between frozen and permanent samples9,11, whereas other studies have indicated lower concordance rates12. When the result of the permanent sample disagrees with that noted intraoperatively, questions arise regarding the proper course of treatment. It is unknown whether such discrepancies in diagnosis have an impact on the patient’s eventual outcome. Furthermore, some previous studies have grouped initial revision and reimplantation procedures, as well as arthroplasty procedures performed on hips and knees. We believe that these should be analyzed separately, as the effectiveness of frozen section analysis may be different for the various procedures in a staged revision.
Because of these issues, we investigated discrepancies between frozen and permanent samples in the diagnosis of infection during total knee arthroplasty revision. We hypothesized that the rate of discrepancy between frozen and permanent sections in the diagnosis of periprosthetic infection was low for both initial revision and reimplantation procedures, and that this low rate of discrepancy had little overall effect on patient management. The purposes of this study were (1) to determine the rate of discrepancy between diagnoses made with frozen sections and with permanent sections at both the initial revision procedure and reimplantation, and (2) to investigate the clinical effect of such discrepancies on the length of time to reimplantation during a staged revision.
A review of prospectively collected data from all total knee arthroplasties performed by a single surgeon was undertaken to identify patients who underwent staged revision due to periprosthetic infection after total knee arthroplasty, had intraoperative frozen tissue samples obtained for histopathologic analysis at both initial revision and reimplantation, and had more than twenty-four months of follow-up after the time of the index revision procedure.
A flowchart showing patient selection and treatment is provided in Figure 1. Between 2002 and 2008, a total of 598 revision total knee arthroplasties were performed. According to the preoperative evaluation described below, these were classified into three categories: presumed aseptic (n = 455), questionable infection (n = 99), and confirmed infection (n = 44). In the patients with a preoperatively confirmed infection, no frozen sections were obtained at the time of the index revision. In the patients whose infection status was questionable, frozen sections were obtained intraoperatively; thirty-one had no evidence of acute inflammation on evaluation of these samples and were reimplanted uneventfully, whereas sixty-eight had positive confirmation of acute inflammation on frozen sections. Additionally, twenty-nine patients who had been presumed to be not infected had intraoperative findings that were suspicious for infection (e.g., discoloration or fragility of the tissue, boggy fluid from the wound, necrotic appearance of the tissue) and had frozen sections obtained for evaluation. Nineteen of these twenty-nine patients who had initially been classified in the aseptic category had no evidence of acute inflammation and were reimplanted uneventfully, whereas ten had positive findings of acute inflammation.
A total of seventy-eight patients were determined to have acute inflammation on the basis of the frozen sections and underwent staged revision of the prosthesis. Two of these seventy-eight patients were lost to follow-up, leaving seventy-six patients who comprised the final study group. After implant removal and a course of parenteral antibiotics, reimplantation was attempted and frozen sections were again obtained. Negative frozen sections were noted in sixty-one patients during reimplantation. Positive frozen sections were noted in fifteen patients, who did not undergo reimplantation; these patients underwent further irrigation and debridement, placement of a new antibiotic cement spacer, and another course of intravenous antibiotics, after which reimplantation was performed only when there was no evidence of acute inflammation. A total of 304 tissue samples were obtained for intraoperative histopathologic analysis. Ninety-two of these samples were obtained during removal of the components because of infection and 212 were obtained during attempted reimplantation. All but one patient (who underwent arthrodesis and subsequently remained free of infection) underwent successful reimplantation.
Demographic information, preoperative laboratory values, intraoperative data, and postoperative clinical and radiographic evaluations were collected from inpatient and clinic charts. Appropriate institutional review board approval was obtained for this study.
The diagnosis of infection was made on the basis of multiple factors including clinical symptoms (e.g., fevers, erythema, drainage), joint aspiration cultures, serum laboratory markers (elevated erythrocyte sedimentation rate and C-reactive protein level), and/or intraoperative cultures. Diagnostic criteria were based on new guidelines from the Musculoskeletal Infection Society workgroup13, which defines three grounds on which a joint may be considered to be infected: (1) a sinus tract in communication with the implant is noted; (2) two separate fluid or tissue cultures from the joint yield a pathogen; or (3) at least four of the following six criteria are met: (a) elevated serum erythrocyte sedimentation rate (>30 mm/hour) or C-reactive protein level (>10 mg/L), (b) any elevated leukocyte count in the synovial aspirate, (c) any increased PMN percentage in the synovial aspirate, (d) gross purulence in the synovial aspirate, (e) one fluid or tissue culture yielding a pathogen, or (f) frozen tissue sections with more than five PMNs per high-power field in at least five fields. (Note that the criterion for the present study was an average of more than five PMNs per high-power field in at least three fields.)
Staged revision was attempted for each infection; this included removal of the femoral, tibial, and patellar components; placement of a custom, antibiotic-impregnated spacer (2 g of tobramycin and 2 g of either vancomycin or gentamicin per 40 g of bone cement); and a minimum six-week course of parenteral antibiotics. Patients returned to the operating room for attempted reimplantation after they had completed the prescribed antibiotic course and subsequently been off antibiotics for a minimum of two weeks. The mean time between commencement of postoperative antibiotics and attempted reimplantation was ninety-two days (range, fifty-six to 337 days); the time varied because of difficulty in scheduling the time for the surgery or delay of the surgery secondary to medical comorbidities. In cases of persistent infection, the knee was thoroughly debrided and irrigated, with replacement of the cement spacer, extension of the parenteral antibiotic course, and further attempts at reimplantation.
Between one and five tissue samples (mean, 1.9 samples) were obtained from various locations (joint capsule, synovium, soft tissue, femur, tibia, and patella) at both the index revision procedure and subsequent reimplantation attempts and were sent immediately for frozen section analysis. Typically, one sample was obtained from the first area of joint capsule encountered. If this sample was positive for acute inflammation, no further samples were obtained. Otherwise, one sample was obtained from the femur and another was obtained later from the tibia, and these were analyzed separately. Thereafter, additional samples were collected from areas suspicious for infection and were analyzed later during the procedure. Samples were also saved for paraffin-embedded permanent sections used for the final analysis.
Sections were cut to 5-μm thickness and stained with hematoxylin and eosin. A team of three pathologists and senior orthopaedic surgeons worked closely together to agree on the histopathologic diagnosis of periprosthetic infection and to correlate these findings with intraoperative observations. Multiple slices from each sample were evaluated, and more than 40 high-power (×400) fields were scanned in each slice scanned for each slice. A section was designated as positive if more than five PMNs per high-power field were present in at least three fields. Indeterminate sections contained at least five fields with between one and five PMNs per high-power field. Sections that had less than one PMN per high-power field averaged over all fields and did not meet criteria for positive or indeterminate status were designated as negative. Infrequently, sections were found to contain a single field with a focal collection of more than five PMN, in which case a search of the section for other focal collections was performed before declaring the section negative for acute inflammation. In each case, the remainder of the excised membrane or pseudocapsule surrounding the implant was held temporarily by the pathology department in case of an indeterminate frozen or paraffin section analysis, and it was discarded when a definitive histopathologic diagnosis was made.
The diagnosis made on the basis of the frozen section was considered a false negative or false positive diagnosis if there was a discrepancy between the diagnosis based on the frozen section and the diagnosis based on the permanent section. The rate of discrepancy was calculated on the basis of the number of samples rather than patients (i.e., by comparing the diagnosis based on each individual frozen sample with the diagnosis based on the permanent section). The clinical outcome was assessed on the basis of the eventual success in clearing the infection and of the need for further surgical revision after what was intended to be a final reimplantation.
Follow-up examinations were performed at one, two, six, and twelve months after the reimplantation procedure and annually thereafter. At each follow-up visit, the patient was assessed for physical signs and symptoms of infection, and anteroposterior and lateral radiographs of the knee were made for evaluation of component loosening and development of new radiolucencies.
The rate of discrepancy between the diagnosis based on the frozen section and the diagnosis based on the permanent section was compared between initial revision procedures and reimplantation procedures with use of the Fisher exact test. The mean time to reimplantation for patients who had a discrepancy between frozen and permanent sections was compared with that for patients without a discrepancy with use of the two-tailed Student t test.
Source of Funding
No external source of funding was sought or utilized in the completion of this study.
Seven discrepancies were noted between the frozen and permanent sections, with concordance in the other 297 samples (98%) (Table I).Six of these discrepancies occurred in four patients in whom positive permanent sections were obtained after negative frozen sections had been obtained intraoperatively. However, none of these four patients underwent reimplantation because other frozen sections obtained during the procedure were positive. Three of these patients eventually went on to reimplantation. The infection in the remaining patient failed to clear after multiple revision operations and spacer exchanges; this patient eventually underwent arthrodesis of the knee joint and remained infection-free thereafter.
One other patient had an indeterminate frozen section during the initial revision and was later diagnosed with infection on the basis of positive permanent sections. One other frozen section obtained from this patient during the initial revision procedure had been positive, resulting in a decision to perform a staged revision for infection. Thus, although one of the frozen sections had a discrepancy, the appropriate surgical management decision was again made because multiple samples were obtained. This patient underwent a successful staged revision, rather than reimplantation at the time of the initial revision, and the reimplantation was uneventful.
The rate of discrepancy for the index revision procedures (one of ninety-two, 1%) was lower than that for the reimplantation procedures (six of 212, 3%), although this difference did not reach significance (p = 0.679). The mean time to reimplantation for patients who had a discrepancy between the frozen and permanent sections (185 days; range, 103 to 247 days) was similar to that for patients who had no discrepancy (157 days; range, forty-six to 443 days). This difference was not significant (p = 0.789).
A false negative intraoperative diagnosis can lead to marked morbidity during reimplantation of components during staged revision of a total knee arthroplasty for apparent infection. Diagnosing periprosthetic infections is of great importance, but the reliability of intraoperative histopathologic analysis of frozen tissue sections has been unclear. The management of these infections relies on proper diagnosis and characterization of the infection14,15, but the effect of inaccurate or misleading intraoperative frozen section results on patient outcomes has not been characterized.
The present study showed that the rate of discrepancy between the results of intraoperative frozen sections and permanent paraffin-embedded sections was low overall; the rate of discrepancy was also lower for initial revision procedures than for reimplantation, although this difference was not significant. The observed discrepancies were all false negative results, and in each patient the presence of infection was confirmed on the basis of other tissue samples. This emphasizes the importance of collecting more than one sample for frozen section analysis as a matter of protocol in appropriate cases of suspected infection to avoid obtaining only a single false negative sample and thereby failing to diagnose an infection. Reimplantation was unfortunately not possible in one of the patients with a discrepancy, but this outcome was unaffected by the presence of the discrepancy, confirming our hypothesis that such discrepancies had little effect on patient management.
There are several limitations to the present study. The location and number of samples collected were variable. The authors attempted to collect tissue in a standardized fashion; however, if a suspicion of localized infection existed, devitalized tissue was sometimes obtained from locations that would not typically have been harvested. The final diagnosis of infection relies on multiple factors, and patients can be misdiagnosed as infected or as uninfected even with use of the stated criteria. The number of samples collected at the index revision procedure was less than the number collected in subsequent procedures, but this was because a larger proportion of first samples were positive at this stage. Therefore, we do not believe that this led to an increase in the potential for false negatives at the index revision procedure, and this should not be considered to be a study limitation. Despite our attempts to create a cohesive team of pathologists and surgeons with a uniform approach to the diagnosis of periprosthetic infection, sampling errors in the high-power fields examined and differences in diagnosis between pathologists could have biased the results. Rare samples that contained focal collections of PMNs were designated as negative for acute inflammation as previously described. Of note, none of the patients who had such samples experienced reinfection or had a discrepancy between the frozen section with the focal collection and the permanent section, but the proper management of such results remains undetermined.
The results of the present study are in accordance with previously reported results that describe good correlation between frozen and permanent sections (Table II). Wong et al. analyzed frozen and permanent samples from thirty-three patients (twenty-five hips, eight knees) with a mean age of sixty-nine years (range, forty-seven to eighty-seven years) who underwent revision hip or knee arthroplasty for infection11. Using a cutoff of five PMNs per high-power field, the concordance between frozen and permanent sections was 95% (nineteen of twenty sections). Feldman et al. reported on thirty-three patients (twenty-three hips, ten knees) with a mean age of sixty-two years (range, thirty-five to eighty-three years) who underwent revision joint arthroplasty for infection9. Using a cutoff of five PMNs per high-power field, frozen sections had a sensitivity of 100% and a specificity of 96% compared with intraoperative culture results, and there was 100% concordance between the frozen and permanent sections. Della Valle et al. analyzed frozen sections from sixty-four patients (thirty-three knees, thirty-one hips) with a mean age of sixty-four years (range, thirty-two to eighty-five years) who underwent reoperation after resection arthroplasty for infection16. They noted two discrepancies (one false negative, one false positive) in the sixty-four patients. The authors made no mention of any change in treatment or management on the basis of the discrepancy between frozen and permanent sections. Raab et al. reviewed the frequency of discrepancies between diagnoses based on frozen and permanent section with use of the Q-Tracks program. This service, which is offered by the College of American Pathologists and stores histopathologic data on any surgical case (from both private and teaching hospitals) for which an intraoperative consultation was requested, has the end goal of improving laboratory benchmarks17. Out of a total of 251,161 frozen samples, 248,512 (98.9%) were concordant. Although this is similar to the rate observed in the present study, a direct comparison is confounded by the inclusion of a variety of operations other than joint arthroplasty.
The present study contrasts with some other existing literature that shows different concordance rates of frozen to permanent sections. Lonner et al. reviewed 175 patients (142 revision total hip arthroplasties, thirty-three revision total knee arthroplasties) with a mean age of sixty-six years (range, thirty-six to eighty-eight years) for evidence of deep infection after joint revision12. They found twelve discrepancies (7% of patients, eight false negatives and four false positives) between frozen and permanent sections. The authors of that study focused on the concordance between frozen sections and intraoperative cultures; thus, no mention was made of any change in management on the basis of discrepancies between frozen and permanent sections. Tohtz et al. described the results of one or two-stage revision hip arthroplasty in sixty-four patients (sixty-four hips) with a mean age of sixty-seven years (range, twenty-four to eighty-nine years) with or without a joint infection18. They noted two discrepancies (3%) between frozen and permanent sections and twelve (19%) were indeterminate, resulting in an overall concordance rate of 78%.
Our results suggest that the reliability of frozen sections for histopathologic diagnosis of periprosthetic joint infections in total knee arthroplasty, as judged by comparison with permanent sections, is high and that frozen sections are approximately as reliable for the diagnosis of periprosthetic infection during initial revision procedures as during subsequent reimplantation attempts. We recommend the practice of obtaining multiple tissue samples intraoperatively from both standardized and suspicious regions as several of the patients in the present study would have been diagnosed as being infection-free with the use of only one section. Following this practice should result in a low rate of cases that are adversely affected by discrepancies between frozen and permanent sections. We also recommend working in concert with a team of trained pathologists who are familiar with the histopathologic diagnosis of periprosthetic infection.
Disclosure: None of the authors received payments or services, either directly or indirectly (i.e., via his or her institution), from a third party in support of any aspect of this work. One or more of the authors, or his or her institution, has had a financial relationship, in the thirty-six months prior to submission of this work, with an entity in the biomedical arena that could be perceived to influence or have the potential to influence what is written in this work. No author has had any other relationships, or has engaged in any other activities, that could be perceived to influence or have the potential to influence what is written in this work. The complete Disclosures of Potential Conflicts of Interest submitted by authors are always provided with the online version of the article.