Abstract
Background:
The accuracy of intraoperative periprosthetic frozen section histologic evaluation in predicting a diagnosis of periprosthetic joint infection prior to microbiologic culture results is unknown.
Methods:
We performed a systematic review and meta-analysis of all longitudinal studies that compared frozen section histologic results with simultaneously obtained microbiologic culture at the time of revision total hip or total knee arthroplasty. The data sources were Ovid MEDLINE, Ovid EMBASE, the Cochrane Library, ISI Web of Science, and SCOPUS, from the inception of each database to January 2010.
Results:
Twenty-six studies involving 3269 patients undergoing revision hip or knee arthroplasty met the inclusion criteria. A culture-positive periprosthetic joint infection was confirmed in 796 (24.3%) of the patients. Frozen section results, using any of the diagnostic criteria chosen by the investigating pathologist, had a pooled diagnostic odds ratio of 54.7 (95% confidence interval [CI], 31.2 to 95.7), a likelihood ratio of a positive test of 12.0 (95% CI, 8.4 to 17.2), and a likelihood ratio of a negative test of 0.23 (95% CI, 0.15 to 0.35) for the diagnosis of periprosthetic joint infection. Fifteen studies utilizing a threshold of five polymorphonuclear leukocytes (PMNs) per high-power field to define a positive frozen section had a diagnostic odds ratio of 52.6 (95% CI, 23.7 to 116.2), and six studies utilizing a diagnostic threshold of ten PMNs per high-power field had a diagnostic odds ratio of 69.8 (95% CI, 33.6 to 145.0). There was no significant difference between the diagnostic odds ratio or likelihood ratios associated with these thresholds. The moderate to high heterogeneity among the included studies was unexplained by variability in the study design, diagnostic criteria for acute inflammation, reference standard for periprosthetic joint infection, or prevalence of infection. This heterogeneity could be due to differences in the inclusion and exclusion criteria, tissue sampling error, experience or technique of the pathologists, number of microscopic fields visualized, and field diameter examined.
Conclusions:
Intraoperative frozen sections of periprosthetic tissues performed well in predicting a diagnosis of culture-positive periprosthetic joint infection but had moderate accuracy in ruling out this diagnosis. Frozen section histopathology should therefore be considered a valuable part of the diagnostic work-up for patients undergoing revision arthroplasty, especially when the potential for infection remains after a thorough preoperative evaluation. The optimum diagnostic threshold (number of PMNs per high-power field) required to distinguish periprosthetic joint infection from aseptic failure could not be discerned from the included studies. There was no significant difference between the diagnostic accuracy of frozen section histopathology utilizing the most common thresholds of five or ten PMNs per high-power field.
Every year, 750,000 people in the United States undergo hip or knee replacement surgery, and this number is projected to increase to 4 million per year by 20301,2. Lower-extremity total joint reconstruction is quite durable, with a cumulative revision burden of <20% in the United States. The leading aseptic causes of failure include loosening, wear (osteolysis), and instability. However, infection is one of the primary modes of failure, and it is the dominant mode of early failure of a hip or knee prosthesis3-6. Furthermore, septic failure results in substantial morbidity and huge costs of care. The cost of treating one episode of periprosthetic joint infection is estimated to be in excess of $50,0007,8. The distinction between joint infection and aseptic failure of total hip or knee arthroplasty is important to make preoperatively but can be clinically difficult because of the low virulence and biofilm-forming ability of the pathogens7. Clinical symptoms and signs are not always reliable, currently available preoperative inflammatory markers in serum and cell counts in synovial fluid have limitations, and intraoperative Gram staining is associated with low sensitivity8-10.
When preoperative results are equivocal, orthopaedic surgeons must rely on a combination of the gross and microscopic appearance of the periprosthetic tissue at the time of surgery to determine whether or not the joint is infected. Failure to accurately recognize a joint infection may lead to the unintended implantation of a new prosthesis into an infected surgical site. Without the appropriate debridement of the joint or adequate local or systemic antibiotic treatment, this may lead to persistence of the infection and possibly early failure of the revision arthroplasty. On the other hand, erroneous diagnosis of a joint infection where there is no infection may result in the patient undergoing unnecessary multiple surgical procedures and inappropriate treatment with a prolonged course of parenteral antibiotics. This places the patient at increased risk of complications from repeated anesthesia and surgery, an extended duration of hospital stay, adverse effects of antimicrobial agents and central venous catheters, prolonged immobilization and rehabilitation, and increased overall costs7.
To our knowledge, the overall performance of intraoperative frozen sections in the diagnostic algorithm for periprosthetic joint infection has not been systematically reviewed. It is unclear whether the diagnostic accuracy differs significantly according to the underlying joint disease or prosthesis location, or according to the selected threshold value for acute inflammation, which is the number of polymorphonuclear leukocytes (PMNs) observed per high-power field11-16.
The purpose of this study was to determine the utility of intraoperative frozen sections in distinguishing between aseptic failure and culture-positive joint infection in patients undergoing revision surgery for failure of a total hip or total knee arthroplasty.
The protocol of this review adheres to methodological guidelines regarding the conduct of systematic reviews of diagnostic accuracy studies17 and the standards proposed by the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) statement18.
Eligibility Criteria
We included peer-reviewed longitudinal studies that assessed the accuracy of intraoperative frozen sections (the index test), in comparison with simultaneously obtained microbiologic cultures from at least two periprosthetic tissue samples (the reference standard), for the diagnosis of joint infection at the time of revision hip or knee arthroplasty. We included studies that evaluated multiple preoperative or intraoperative markers of joint infection if they had primary data for a head-to-head comparison of frozen sections with the reference standard. Studies that involved patients who presented for reimplantation surgery (with no prosthesis in place at the time of tissue sampling) were excluded. No language restriction was applied.
Search Strategy
The literature search was performed by an expert librarian with use of Ovid MEDLINE, Ovid EMBASE, the Cochrane Library, ISI Web of Science, and SCOPUS (from inception of each database to January 2010). The search terms used were “intra operative frozen section,” “frozen section biopsy,” “frozen section analysis,” “histology,” “histological evaluation,” “histopathology,” “pathology of joint tissues,” or “presence of polymorphonuclear leukocytes” and “revision arthroplasty,” “total knee replacement,” “total hip replacement,” “total knee arthroplasty,” “total hip arthroplasty,” “prosthetic joint infection,” “peri implant infection,” “aseptic loosening of joint prosthesis,” “resection arthroplasty,” “sepsis in revision total joint arthroplasty,” “diagnosis of prosthetic joint infection,” or “failed total joint arthroplasty” (see Appendix). In addition, one reviewer manually reviewed the bibliographies of retrieved articles for additional citations.
Two reviewers independently screened the titles and abstracts of all identified reports. The full published reports corresponding to all abstracts selected by the reviewers were retrieved, and a second round of selection on the basis of the inclusion criteria was performed by the same two reviewers independently. At both stages of selection, we estimated chance-adjusted agreement statistics among reviewers (i.e., the kappa statistic); disagreements were resolved by consensus or by arbitration.
We circulated the list of included articles to content experts to judge the completeness of the list and to identify completed but unpublished studies.
Validity Assessment
Two reviewers independently assessed the methodological quality of each included study with use of the QUADAS (Quality Assessment of Diagnostic Accuracy Studies) tool19. This tool includes fourteen questions in three domains (generalizability, clarity, and validity; see Appendix). Disagreements were resolved by consensus, and agreement between the two reviewers was measured by the kappa statistic.
Data Abstraction
A standardized data extraction form was developed, and two reviewers independently extracted data from each included report. Extracted data included the number of participants in the study, location of joints involved, and criteria for a positive intraoperative frozen section (threshold number of PMNs per high-power field, minimum number of high-power fields required, and microscopic magnification that constituted a high-power field). Other data included the criteria for the reference standard, which were microbiologic growth from solid media in at least two specimens with or without any additional criteria for joint infection diagnosis (clinical symptoms and signs, a sinus tract from the prosthesis to the skin, preoperative inflammatory markers in serum, synovial fluid leukocyte count, purulence noted at the time of surgery, or permanent histologic sections from biopsy specimens), and any explicit exclusion criteria.
Raw data from the articles were used to independently construct 2 × 2 contingency tables from the absolute numbers of true positives, true negatives, false positives, and false negatives. When studies reported the sensitivity and specificity for multiple thresholds, we included each set of 2 × 2 results in the appropriate subgroup, and we chose the threshold with the best accuracy value (estimated by dividing the sum of true-positive and true-negative cases by the total number of cases) for the whole-group analysis. Disagreements between the two reviewers were resolved by consensus.
Author Contact
We attempted to contact all authors and provide them with a form that included data extracted from their published work. They were asked to verify the extracted data and to fill in data that were missing or could not be discerned from the report.
Outcomes of Interest
The primary end point for the evaluation of the diagnostic accuracy of intraoperative frozen sections was the diagnostic odds ratio. Other outcomes of interest included the likelihood ratio of a positive test and the likelihood ratio of a negative test. The likelihood ratio is defined as the probability of a given test result in patients with the disease divided by the probability of that same result in patients without the disease. The positive likelihood ratio equals sensitivity/(1 − specificity); the negative likelihood ratio equals (1 − sensitivity)/specificity.
The diagnostic odds ratio, which equals the ratio of the positive likelihood ratio to the negative likelihood ratio, expresses how much greater the odds of having the disease (a joint infection) are for a patient with a positive test result (frozen section) than for a patient with a negative test result. It is a single measure of diagnostic test performance that combines the strengths of the sensitivity and specificity parameters and does not depend on the prevalence of the disease that the test is used to diagnose. Unlike specificity and sensitivity, the diagnostic odds ratio is often reasonably constant regardless of the diagnostic threshold20.
Positive likelihood ratios of >10 and negative likelihood ratios of <0.1 have been noted to provide convincing diagnostic evidence, whereas those of >5 and <0.2 provide strong diagnostic evidence21.
Summary receiver operating characteristic (SROC) curves were also drawn, with use of the linear models of Moses and Shapiro, to summarize the accuracy of frozen sections in each subgroup. The Cochran Q value and the area under the curve were calculated22. We avoided pooling the sensitivity and specificity values separately because such analyses ignore their interrelationship.
Statistical Analysis
We decided a priori that the included studies would likely be heterogeneous and chose the random-effect model of DerSimonian and Laird23 for all pooled analyses. The difference in sample size among the studies was taken into account by weighting each observation by the reciprocal of the variance of the study and performing weighted regression. We used the inconsistency index (I2) statistic to assess heterogeneity. The I2 statistic estimates the percentage of the variability in results across studies that is explained by true differences in patients, tests, outcomes, and design rather than by chance; values of 25%, 50%, and 75% indicate low, moderate, and high inconsistency, respectively24.
We visually inspected the SROC curves for a “shoulder and arm pattern” that might suggest diagnostic threshold bias, and we also calculated a Spearman correlation coefficient. A strongly positive rank-correlation coefficient and a p value of <0.05 would indicate a significant threshold effect.
Subgroup analyses were planned a priori to investigate the diagnostic accuracy according to the index test threshold (five or ten PMNs per high-power field) and the reference test used (microbiologic culture alone or in combination with other criteria). Subgroup analyses were also planned with study design (historical cohort compared with prospective), independence of the index test from the reference standard (yes compared with no), prosthesis location (hip compared with knee), prevalence of joint infection in the patients undergoing revision surgery (≤20%, >20% to 40%, or >40%), and underlying joint disease (inflammatory compared with degenerative arthritis) as independent variables.
A statistical test of interaction was conducted for each subgroup analysis, with a p value of <0.05 being considered significant25. The relative diagnostic odds ratio of the corresponding covariates was calculated.
A Fagan nomogram was constructed to illustrate our findings. This nomogram shows the effect of the pooled likelihood ratios on the post-test probability of joint infection26.
Source of Funding
No outside funding was used for this study.
Study Selection
Two hundred and twenty-seven studies were retrieved through the primary search, and seventy-six of these were considered potentially eligible (Fig. 1). The full text of each of those articles was reviewed in detail, and twenty-six studies met the inclusion criteria. Two additional studies were identified from the bibliography search, and no additional studies were identified by the content experts. Agreement between the two reviewers with regard to the final selection of studies was 94.7%, with a kappa statistic of 0.88 (standard error, 0.06; 95% confidence interval [CI], 0.77 to 0.99). We attempted to contact the authors of each included study to verify the data and obtain missing information; the authors of three studies13,27,28 could not be reached after multiple attempts. It was obvious from the article and/or confirmed by the corresponding author that three of the studies contained patient data from other studies that had already been included, and these three studies were excluded. Four studies analyzed the diagnostic test performance of frozen sections with use of two different histopathology criteria; thirty sets of results from twenty-six studies were therefore eligible for meta-analysis12-16,27-47.
Baseline Characteristics
The twenty-six studies involved 3269 patients undergoing revision arthroplasty, and 796 (24.3%) of these patients were confirmed to have a joint infection (see Appendix). Four studies enrolled only patients undergoing total hip arthroplasty, one included only patients undergoing total knee arthroplasty, and twenty-one included both types of patients. Fifteen sets of results used a minimum of five PMNs per high-power field as the criterion for acute inflammation, six utilized a threshold of ten PMNs per high-power field, and nine utilized a variety of other threshold values or were not specific. The reference standard for joint infection was growth of microorganisms in culture from at least two specimens as the sole criterion in eighteen studies, and it was positive microbiology plus a variety of other clinical or laboratory criteria in eight studies.
Quality Assessment
The two reviewers who performed the quality evaluation of the included studies were in agreement on 97.8% of the questions evaluated, with an overall kappa statistic of 0.83 (range, 0.71 to 0.95 on the individual questions). Ten (43%) of the twenty-three authors who could be contacted replied to our queries regarding verification of study information (see Appendix). All twenty-six studies satisfied ten of the fourteen quality items on the QUADAS checklist (items 3, 4, 5, 6, 8, 10, 11, 12, 13, and 14; see Appendix). The studies differed on four of the items: avoidance of spectrum bias, full description of the selection criteria, independence of the reference standard from the index test, and full description of the reference standard (see Appendix).
Meta-Analyses
Summary Estimates of Pooled Data
All twenty-six studies were included in the initial analysis to calculate pooled estimates of the diagnostic odds ratio and positive and negative likelihood ratios of frozen sections (based on any pathologist-defined acute inflammation) compared with a reference standard of culture-positive joint infection (based on microbiologic culture with or without other additional criteria for joint infection).
The pooled diagnostic odds ratio was 54.7 (95% CI, 31.2 to 95.7), with an I2 of 67.6% (p < 0.001) (Fig. 2). The pooled positive likelihood ratio was 12.0 (95% CI, 8.4 to 17.2), with an I2 of 69.3% (p < 0.001), and the pooled negative likelihood ratio was 0.23 (95% CI, 0.15 to 0.35), with an I2 of 89.1% (p < 0.001) (Figs. 3 and 4). The SROC curve that included all of the studies had a skewed distribution of data points (with no studies in right and upper quadrants; see Appendix), so no further analyses using such plots were conducted, as the Q value and area under the curve need to be interpreted with caution in such a setting.
The Spearman rank correlation coefficient (for the twenty-one sets of results that explicitly used a diagnostic threshold of five or ten PMNs per high-power field as the index test) was −0.06 (p = 0.80). This confirmed that the high variability across these studies was not explained by differences in the diagnostic threshold.
Threshold Value for Frozen Sections
The diagnostic odds ratio for frozen sections was 52.6 (95% CI, 23.7 to 116.2) for a threshold of five PMNs per high-power field and 69.8 (95% CI, 33.6 to 145.0) for ten PMNs (Table I). Interaction testing indicated that the diagnostic odds ratio estimates for these thresholds did not differ significantly (p = 0.61). The corresponding positive likelihood ratio was 10.3 (95% CI, 6.3 to 16.6) for five PMNs per high-power field and 16.9 (95% CI, 10.4 to 27.4) for ten PMNs (p = 0.16). The negative likelihood ratio estimate was 0.24 (95% CI, 0.15 to 0.39) for five PMNs and 0.27 (95% CI, 0.20 to 0.38) for ten PMNs (p = 0.69). The corresponding forest plots are illustrated in Figures 5 and 6 and the Appendix. There was insufficient evidence to detect a difference in the accuracy of frozen sections for the diagnosis of joint infection on the basis of a tissue density of five or ten PMN inflammatory cells per high-power field.
Reference Standard for Joint Infection Diagnosis
The diagnostic odds ratio for frozen sections was 48.1 (95% CI, 23.6 to 98.7) for joint infection defined by microbiologic culture only and 76.5 (95% CI, 32.8 to 178.6) when the reference standard was culture plus any additional criteria. The relative diagnostic odds ratio was 1.59 (95% CI, 0.52 to 4.82) (p = 0.41). The corresponding positive likelihood ratio estimates were 11.3 (95% CI, 7.2 to 17.6) and 13.5 (95% CI, 8.9 to 20.6) (p = 0.56). The corresponding negative likelihood ratios were 0.24 (95% CI, 0.14 to 0.42) and 0.21 (95% CI, 0.12 to 0.36) (p = 0.74).
Other Subgroup Analyses
Subgroup analyses of the diagnostic accuracy outcomes were performed with the prevalence of joint infection in patients undergoing revision surgery (≤20%, >20% to 40%, or >40%), independence of the reference standard from the index test (studies that included or excluded frozen sections as part of the reference standard), and study design (prospective, historical cohort, or cross-sectional) as independent variables (see Appendix). The diagnostic accuracy of frozen sections, as determined on the basis of the diagnostic odds ratio and the positive and negative likelihood ratios, did not differ significantly among the subgroups. There were insufficient primary data to permit subgroup analysis of the effects of underlying joint disease, prosthesis age, or prosthesis location on the diagnostic accuracy of frozen sections.
This systematic literature review evaluated the utility of intraoperative frozen sections, in comparison with a reference standard based on microbiologic culture, for predicting the presence of joint infection. We determined the extent to which a positive (or negative) frozen section increased (or decreased) the likelihood of infection in patients undergoing revision of a failed hip or knee arthroplasty.
Our survey has several substantial limitations. It included numerous small studies with a variety of study designs, a wide spectrum of included patients, different criteria for the histologic diagnosis of infection (different tissue concentrations of inflammatory cells), and subtle differences in the gold standard for joint infection diagnosis. Some of the studies included in our survey had an unusually high prevalence of joint infection, suggesting that these patients may not have been enrolled consecutively or were not necessarily selected on the basis of diagnostic uncertainty at the time of surgery. The pooled results represent the best estimates of the accuracy of frozen sections for the diagnosis of joint infection in the context of these limitations.
Regardless of the criteria defined by the histopathologist, frozen section analysis of periprosthetic tissue was very good at predicting culture-positive joint infection (positive likelihood ratio, typically >10), and it was moderately accurate in ruling out this diagnosis (negative likelihood ratio, 0.2 to 0.3). Figure 7 illustrates the clinical implications of these estimates.
We explored the possibility that differences in the criteria for diagnosing acute inflammation on frozen sections and/or differences in the reference standard for joint infection could introduce bias in our estimates. Although there was a trend toward a greater likelihood of infection when a threshold of ten PMNs per high-power field was chosen compared with five PMNs, this difference was not significant. There is no uniformly endorsed reference standard in the orthopaedic or infectious-disease communities for the diagnosis of joint infection. Most experts agree on isolation of the same microorganism in culture from two or more intraoperatively sampled tissue specimens48,49. All of the included studies were designed with such culture positivity from intraoperative specimens as a necessary criterion for establishing a diagnosis of joint infection. Other acceptable alternatives are positive cultures from preoperatively aspirated joint fluid or from sonication of the prosthesis after surgery, but neither of these methods was utilized in the included studies50. Observation of a sinus tract or gross purulence about the prosthesis at the time of surgery has also been widely recognized as a “stand-alone” criterion for the diagnosis of joint infection48-51. For the purposes of our review, these latter criteria will have introduced selection bias by including patients with “diagnostic certainty.” Most of the studies (eighteen of twenty-six) did not include a sinus tract or periprosthetic purulence as a requirement for a diagnosis of joint infection, and some explicitly excluded patients with these characteristics. Other debatable sets of criteria for the diagnosis of joint infection involve the combination of microbiologic culture (especially single positive cultures or growth in broth only) with other preoperative and/or intraoperative tests (including histopathology). Our subgroup analysis showed no significant difference in the accuracy of frozen sections when the frozen sections were included in the reference standard or were independent of the reference standard.
The inconsistency of the results across studies was moderate to high, and this could not be explained by differences in the prevalence of joint infection in the study populations or by differences in study design. The significant variability among studies may represent differences in the surgical sampling of periprosthetic tissue, technique or experience of the examining pathologists, number and diameter of microscopic fields examined, underlying joint disease, age of the prostheses, acuity of presentation, and antecedent use of antimicrobial agents. However, these inconsistencies could not be explored further because of the paucity or lack of reported data on these variables in the included studies. The utility of frozen sections in the diagnosis of “culture-negative joint infection” is beyond the scope of this work, as we did not include (or indeed find) any studies that accepted a diagnosis of joint infection without positive results from culture of tissue.
In spite of the methodological differences among the included studies, our review has some major strengths. The very high QUADAS quality assessment scores (see Appendix) are attributable primarily to the concurrent sampling of tissue for culture and histology and to the independence in interpretation of the results from the microbiology and pathology laboratories. Data for this meta-analysis were collected in duplicate, with very high agreement between the investigators. Further verification of the data was obtained from the authors of the included studies. We have included a Fagan nomogram (Fig. 7), which is a practical tool for clinical interpretation of the likelihood ratios. Although we summarize the accuracy of frozen sections on the basis of “any pathologist-defined criteria for acute inflammation,” we also provide subgroup estimates (Table I) that can be tailored to the specific diagnostic criteria used at the reader's own institution. The remarkable finding from this review, however, is that the nuances of the case definitions for the diagnosis of joint infection on the basis of the frozen section or reference standard did not adversely affect the qualitative degree of accuracy (excellent for the positive likelihood ratio or moderate for the negative likelihood ratio) or the quantitative estimates in the Fagan nomogram. The reader can take home essentially the same message about the accuracy of frozen section histopathology for the diagnosis of joint infection, regardless of how strictly or inclusively they will define the test or the outcome.
There is currently no published practice guideline from the Infectious Disease Society of America or the American Academy of Orthopaedic Surgeons regarding a diagnostic algorithm for periprosthetic joint infection. A few published articles have sought to address this issue on the basis of narrative reviews, single-center observational studies, or expert opinion52,53. The inclusion of intraoperative periprosthetic histopathology in the diagnostic work-up for joint infection has been suggested by these authors. Nevertheless, consistency in requesting or performing the test is dependent on the preference of the orthopaedic surgeon or institution and on the availability of an experienced pathologist. Patient selection for frozen sections also varies and may be limited to those with true diagnostic uncertainty at the time of revision surgery.
Given that infection is the cause of 14.8% total hip arthroplasty failures and 25.2% of total knee arthroplasty failures, there is a strong case for more frequent use of periprosthetic frozen section analysis, especially when the potential for infection remains after a thorough preoperative evaluation5,6. In our review, the likelihood of infection was increased severalfold when the frozen section results were positive (Fig. 7). It must be emphasized that frozen section histopathology is one of several (preoperative, intraoperative, and postoperative) diagnostic variables that informs a surgeon's decision regarding the infection status of a joint arthroplasty. A negative frozen section result in an obviously purulent joint or in a joint with a sinus tract should not be used to exclude the diagnosis of infection. Similarly, if aseptic failure is suggested by preoperative tests such as serum inflammatory markers and the synovial fluid leukocyte count or by the intraoperative gross appearance of the joint, then the frozen section result should be interpreted in context and with the aid of the Fagan nomogram.
We uncovered several research gaps that need addressing. From the diagnostic standpoint, there need to be uniform criteria for the histopathologic diagnosis of joint infection that lend reproducibility to the diagnosis. A simple “acute inflammation present” or “no acute inflammation” may suffice for the practicing surgeon or physician. However, a quantitative analysis that is more detailed and consistent will be more useful for research purposes. It will provide other pathologists with opportunities to validate their findings without having to choose from a host of preset criteria. Furthermore, future research studies could focus on defining the threshold number of PMNs per high-power field that is most predictive of joint infection. This could be achieved with SROC curves comparing various thresholds to optimize sensitivity and specificity, although one may argue that the “best test performance” may not necessarily be the one that maximizes the area under the curve since the consequences of failure to diagnose joint infection exceed those of “overtreating” an aseptic joint. There also needs to be consensus on the gold-standard set of criteria for joint infection diagnosis to enable easier and more accurate comparison of diagnostic modalities and treatment outcomes. In terms of reporting of diagnostic studies of joint infection, future publications should provide more detailed inclusion and exclusion criteria as well as baseline characteristics of participants including patient age, prosthesis age, underlying joint disease, and antecedent use of antibiotics. The definition of culture-negative joint infection should be well established to facilitate studies that will evaluate the role of frozen sections in this subgroup.
In summary, intraoperative frozen section histologic evaluation was very good at predicting a diagnosis of culture-positive joint infection and had moderate accuracy in ruling out this diagnosis. Frozen section histopathology should be considered a valuable part of the diagnostic work-up for patients undergoing revision arthroplasty, especially when the potential for infection remains after a thorough preoperative evaluation. The optimum diagnostic threshold for the number of PMNs per high-power field required to distinguish periprosthetic joint infection from aseptic failure could not be discerned from the studies reviewed. There was no significant difference between the diagnostic accuracy of frozen section histopathology utilizing the most common thresholds of five or ten PMNs per high-power field.
Tables showing the characteristics of the included studies, the search terms used, the QUADAS questionnaire items, the quality of the included studies, and the results of the additional subgroup analyses as well as figures showing the quality of the included studies, forest plots of the positive and negative likelihood ratios for a threshold of ten PMNs per high-power field, and the SROC curve for frozen sections compared with microbiologic culture are available with the online version of this article as a data supplement at jbjs.org.
Note: The authors extend their appreciation to the following authors of the primary studies included in this review, who responded to our inquiries for data verification: N.A. Athanasou (two articles), B. Fink, C.J. Della Valle, J.M. Hartford, L.V. Nuñez, P. Schäfer, L. Savarino, H. Amstutz, and D.S. Feldman.
Kurtz
S;
Mowat
F;
Ong
K;
Chan
N;
Lau
E;
Halpern
M. Prevalence of primary and revision total hip and knee arthroplasty in the United States from 1990 through 2002. J Bone Joint Surg Am.
2005 Jul;87(
7):1487-97.[CrossRef]
Kurtz
S;
Ong
K;
Lau
E;
Mowat
F;
Halpern
M. Projections of primary and revision hip and knee arthroplasty in the United States from 2005 to 2030. J Bone Joint Surg Am.
2007 Apr;89(
4):780-5.[CrossRef]
Sharkey
PF;
Hozack
WJ;
Rothman
RH;
Shastri
S;
Jacoby
SM. Insall Award paper. Why are total knee arthroplasties failing today?Clin Orthop Relat Res.
2002 Nov;(
404):7-13.
Fehring
TK;
Odum
S;
Griffin
WL;
Mason
JB;
Nadaud
M. Early failures in total knee arthroplasty. Clin Orthop Relat Res
2001(
392):315-8.
Bozic
KJ;
Kurtz
SM;
Lau
E;
Ong
K;
Chiu
V;
Vail
TP;
Rubash
HE;
Berry
DJ. The epidemiology of revision total knee arthroplasty in the United States. Clin Orthop Relat Res.
2010 Jan;468(
1):45-51. .[CrossRef]
Bozic
KJ;
Kurtz
SM;
Lau
E;
Ong
K;
Vail
TP;
Berry
DJ. The epidemiology of revision total hip arthroplasty in the United States. J Bone Joint Surg Am.
2009 Jan;91(
1):128-33.[CrossRef]
Trampuz
A;
Osmon
DR;
Hanssen
AD;
Steckelberg
JM;
Patel
R. Molecular and antibiofilm approaches to prosthetic joint infection. Clin Orthop Relat Res.
2003 Sep;(
414):69-88.
Del Pozo
JL;
Patel
R. Clinical practice. Infection associated with prosthetic joints. N Engl J Med.
2009 Aug 20;361(
8):787-94.[CrossRef]
Morgan
PM;
Sharkey
P;
Ghanem
E;
Parvizi
J;
Clohisy
JC;
Burnett
RS;
Barrack
RL. The value of intraoperative Gram stain in revision total knee arthroplasty. J Bone Joint Surg Am.
2009 Sep;91(
9):2124-9.[CrossRef]
Parvizi
J;
Ghanem
E;
Sharkey
P;
Aggarwal
A;
Burnett
RS;
Barrack
RL. Diagnosis of infected total knee: findings of a multicenter database. Clin Orthop Relat Res.
2008 Nov;466(
11):2628-33. .[CrossRef]
Kataoka
M;
Torisu
T;
Tsumura
H;
Yoshida
S;
Takashita
M. An assessment of histopathological criteria for infection in joint arthroplasty in rheumatoid synovium. Clin Rheumatol.
2002 May;21(
2):159-63.[CrossRef]
Banit
DM;
Kaufer
H;
Hartford
JM. Intraoperative frozen section analysis in revision total joint arthroplasty. Clin Orthop Relat Res.
2002 Aug;(
401):230-8.
Francés Borrego
A;
Martínez
FM;
Cebrian Parra
JL;
Grañeda
DS;
Crespo
RG;
López-Durán Stern
L. Diagnosis of infection in hip and knee revision surgery: intraoperative frozen section analysis. Int Orthop.
2007 Feb;31(
1):33-7. .[CrossRef]
Feldman
DS;
Lonner
JH;
Desai
P;
Zuckerman
JD. The role of intraoperative frozen sections in revision total joint arthroplasty. J Bone Joint Surg Am.
1995 Dec;77(
12):1807-13.
Lonner
JH;
Desai
P;
Dicesare
PE;
Steiner
G;
Zuckerman
JD. The reliability of analysis of intraoperative frozen sections for identifying active infection during revision hip or knee arthroplasty. J Bone Joint Surg Am.
1996 Oct;78(
10):1553-8.
Pandey
R;
Drakoulakis
E;
Athanasou
NA. An assessment of the histological criteria used to diagnose infection in hip revision arthroplasty tissues. J Clin Pathol.
1999 Feb;52(
2):118-23.[CrossRef]
Devillé
WL;
Buntinx
F;
Bouter
LM;
Montori
VM;
de Vet
HC;
van der Windt
DA;
Bezemer
PD. Conducting systematic reviews of diagnostic studies: didactic guidelines. BMC Med Res Methodol.
2002 Jul 3;2:9.[CrossRef]
Liberati
A;
Altman
DG;
Tetzlaff
J;
Mulrow
C;
Gøtzsche
PC;
Ioannidis
JP;
Clarke
M;
Devereaux
PJ;
Kleijnen
J;
Moher
D. The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate health care interventions: explanation and elaboration. Ann Intern Med.
2009 Aug 18;151(
4):W65-94. .
Whiting
P;
Rutjes
AW;
Reitsma
JB;
Bossuyt
PM;
Kleijnen
J. The development of QUADAS: a tool for the quality assessment of studies of diagnostic accuracy included in systematic reviews. BMC Med Res Methodol.
2003 Nov 10;3:25.[CrossRef]
Glas
AS;
Lijmer
JG;
Prins
MH;
Bonsel
GJ;
Bossuyt
PM. The diagnostic odds ratio: a single indicator of test performance. J Clin Epidemiol.
2003 Nov;56(
11):1129-35.[CrossRef]
Bhandari
M;
Guyatt
GH. How to appraise a diagnostic test. World J Surg.
2005 May;29(
5):561-6.[CrossRef]
Kester
AD;
Buntinx
F. Meta-analysis of ROC curves. Med Decis Making.
2000 Oct-Dec;20(
4):430-9.[CrossRef]
DerSimonian
R;
Laird
N. Meta-analysis in clinical trials. Control Clin Trials.
1986;7(
3):177-88.[PubMed][CrossRef]
Higgins
JP;
Thompson
SG;
Deeks
JJ;
Altman
DG. Measuring inconsistency in meta-analyses. BMJ.
2003 Sep 6;327(
7414):557-60.[CrossRef]
Altman
DG;
Bland
JM. Interaction revisited: the difference between two estimates. BMJ.
2003 Jan 25;326(
7382):219.[CrossRef]
Fagan
TJ. Letter: Nomogram for Bayes theorem. N Engl J Med.
1975 Jul 31;293(
5):257.
Pons
M;
Anglés
F;
Sánchez
C;
Matamala
A;
Cuchi
E;
Salavert
M;
Forcada
P;
Ferrer
H. Infected total hip arthroplasty--the value of intraoperative histology. Int Orthop.
1999;23(
1):34-6.[PubMed][CrossRef]
Shokeir
MO;
Duncan
CP;
O'Connell
JX. Revision hip arthroplasty. A histologic review. Int J Surg Pathol.
1996 Apr;3(
4):247-56.
Abdul-Karim
FW;
McGinnis
MG;
Kraay
M;
Emancipator
SN;
Goldberg
V. Frozen section biopsy assessment for the presence of polymorphonuclear leukocytes in patients undergoing revision of arthroplasties. Mod Pathol.
1998 May;11(
5):427-31.
Athanasou
NA;
Pandey
R;
de Steiger
R;
Crook
D;
Smith
PM. Diagnosis of infection by frozen section during revision arthroplasty. J Bone Joint Surg Br.
1995 Jan;77(
1):28-33.
Bori
G;
Soriano
A;
García
S;
Gallart
X;
Casanova
L;
Mallofre
C;
Almela
M;
Martínez
JA;
Riba
J;
Mensa
J. Low sensitivity of histology to predict the presence of microorganisms in suspected aseptic loosening of a joint prosthesis. Mod Pathol.
2006 Jun;19(
6):874-7.
Della Valle
CJ;
Sporer
SM;
Jacobs
JJ;
Berger
RA;
Rosenberg
AG;
Paprosky
WG. Preoperative testing for sepsis before revision total knee arthroplasty. J Arthroplasty.
2007 Sep;22(
6 Suppl 2):90-3. .[CrossRef]
Fehring
TK;
Cohen
B. Aspiration as a guide to sepsis in revision total hip arthroplasty. J Arthroplasty.
1996 Aug;11(
5):543-7.[CrossRef]
Fehring
TK;
McAlister
JA
Jr. Frozen histologic section as a guide to sepsis in revision joint arthroplasty. Clin Orthop Relat Res.
1994 Jul;(
304):229-37.
Fink
B;
Makowiak
C;
Fuerst
M;
Berger
I;
Schäfer
P;
Frommelt
L. The value of synovial biopsy, joint aspiration and C-reactive protein in the diagnosis of late peri-prosthetic infection of total knee replacements. J Bone Joint Surg Br.
2008 Jul;90(
7):874-8.[CrossRef]
Kanner
WA;
Saleh
KJ;
Frierson
HF
Jr. Reassessment of the usefulness of frozen section analysis for hip and knee joint revisions. Am J Clin Pathol.
2008 Sep;130(
3):363-8.[CrossRef]
Ko
PS;
Ip
D;
Chow
KP;
Cheung
F;
Lee
OB;
Lam
JJ. The role of intraoperative frozen section in decision making in revision hip and knee arthroplasties in a local community hospital. J Arthroplasty.
2005 Feb;20(
2):189-95.[CrossRef]
Malhotra
R;
Morgan
DA. Role of core biopsy in diagnosing infection before revision hip arthroplasty. J Arthroplasty.
2004 Jan;19(
1):78-87.[CrossRef]
Mirra
JM;
Marder
RA;
Amstutz
HC. The pathology of failed total joint arthroplasty. Clin Orthop Relat Res.
1982 Oct;(
170):175-83.
Morawietz
L;
Classen
RA;
Schröder
JH;
Dynybil
C;
Perka
C;
Skwara
A;
Neidel
J;
Gehrke
T;
Frommelt
L;
Hansen
T;
Otto
M;
Barden
B;
Aigner
T;
Stiehl
P;
Schubert
T;
Meyer-Scholten
C;
König
A;
Ströbel
P;
Rader
CP;
Kirschner
S;
Lintner
F;
Rüther
W;
Bos
I;
Hendrich
C;
Kriegsmann
J;
Krenn
V. Proposal for a histopathological consensus classification of the periprosthetic interface membrane. J Clin Pathol.
2006 Jun;59(
6):591-7.[CrossRef]
Müller
M;
Morawietz
L;
Hasart
O;
Strube
P;
Perka
C;
Tohtz
S. [Histopathological diagnosis of periprosthetic joint infection following total hip arthroplasty: use of a standardized classification system of the periprosthetic interface membrane]. Orthopade.
2009 Nov;38(
11):1087-96. .[CrossRef]
Nilsdotter-Augustinsson
A;
Briheim
G;
Herder
A;
Ljunghusen
O;
Wahlström
O;
Ohman
L. Inflammatory response in 85 patients with loosened hip prostheses: a prospective study comparing inflammatory markers in patients with aseptic and septic prosthetic loosening. Acta Orthop.
2007 Oct;78(
5):629-39.[CrossRef]
Nuñez
LV;
Buttaro
MA;
Morandi
A;
Pusso
R;
Piccaluga
F. Frozen sections of samples taken intraoperatively for diagnosis of infection in revision hip surgery. Acta Orthop.
2007 Apr;78(
2):226-30.[CrossRef]
Savarino
L;
Tigani
D;
Baldini
N;
Bochicchio
V;
Giunti
A. Pre-operative diagnosis of infection in total knee arthroplasty: an algorithm. Knee Surg Sports Traumatol Arthrosc.
2009 Jun;17(
6):667-75. .[CrossRef]
Schäfer
P;
Fink
B;
Sandow
D;
Margull
A;
Berger
I;
Frommelt
L. Prolonged bacterial culture to identify late periprosthetic joint infection: a promising strategy. Clin Infect Dis.
2008 Dec 1;47(
11):1403-9.[CrossRef]
Schinsky
MF;
Della Valle
CJ;
Sporer
SM;
Paprosky
WG. Perioperative testing for joint infection in patients undergoing revision total hip arthroplasty. J Bone Joint Surg Am.
2008 Sep;90(
9):1869-75.[CrossRef]
Tohtz
SW;
Müller
M;
Morawietz
L;
Winkler
T;
Perka
C. Validity of frozen sections for analysis of periprosthetic loosening membranes. Clin Orthop Relat Res.
2010 Mar;468(
3):762-8. .[CrossRef]
Berbari
EF;
Hanssen
AD;
Duffy
MC;
Steckelberg
JM;
Ilstrup
DM;
Harmsen
WS;
Osmon
DR. Risk factors for prosthetic joint infection: case-control study. Clin Infect Dis.
1998 Nov;27(
5):1247-54.[CrossRef]
Betsch
BY;
Eggli
S;
Siebenrock
KA;
Täuber
MG;
Mühlemann
K. Treatment of joint prosthesis infection in accordance with current recommendations improves outcome. Clin Infect Dis.
2008 Apr 15;46(
8):1221-6.[CrossRef]
Trampuz
A;
Piper
KE;
Jacobson
MJ;
Hanssen
AD;
Unni
KK;
Osmon
DR;
Mandrekar
JN;
Cockerill
FR;
Steckelberg
JM;
Greenleaf
JF;
Patel
R. Sonication of removed hip and knee prostheses for diagnosis of infection. N Engl J Med.
2007 Aug 16;357(
7):654-63.[CrossRef]
Marculescu
CE;
Berbari
EF;
Hanssen
AD;
Steckelberg
JM;
Harmsen
SW;
Mandrekar
JN;
Osmon
DR. Outcome of prosthetic joint infections treated with debridement and retention of components. Clin Infect Dis.
2006 Feb 15;42(
4):471-8. .[CrossRef]
Parvizi
J;
Ghanem
E;
Menashe
S;
Barrack
RL;
Bauer
TW. Periprosthetic infection: what are the diagnostic challenges?J Bone Joint Surg Am.
2006 Dec;88
Suppl 4:138-47.[CrossRef]
Bauer
TW;
Brooks
PJ;
Sakai
H;
Krebs
V;
Borden
L. A diagnostic algorithm for detecting an infected hip arthroplasty. Orthopedics.
2003 Sep;26(
9):929-30.