The 2018 Definition of Periprosthetic Hip and Knee Infection: An Evidence-Based and Validated Criteria

Published:February 26, 2018DOI:https://doi.org/10.1016/j.arth.2018.02.078

      Abstract

      Background

      The introduction of the Musculoskeletal Infection Society (MSIS) criteria for periprosthetic joint infection (PJI) in 2011 resulted in improvements in diagnostic confidence and research collaboration. The emergence of new diagnostic tests and the lessons we have learned from the past 7 years using the MSIS definition, prompted us to develop an evidence-based and validated updated version of the criteria.

      Methods

      This multi-institutional study of patients undergoing revision total joint arthroplasty was conducted at 3 academic centers. For the development of the new diagnostic criteria, PJI and aseptic patient cohorts were stringently defined: PJI cases were defined using only major criteria from the MSIS definition (n = 684) and aseptic cases underwent one-stage revision for a noninfective indication and did not fail within 2 years (n = 820). Serum C-reactive protein (CRP), D-dimer, erythrocyte sedimentation rate were investigated, as well as synovial white blood cell count, polymorphonuclear percentage, leukocyte esterase, alpha-defensin, and synovial CRP. Intraoperative findings included frozen section, presence of purulence, and isolation of a pathogen by culture. A stepwise approach using random forest analysis and multivariate regression was used to generate relative weights for each diagnostic marker. Preoperative and intraoperative definitions were created based on beta coefficients. The new definition was then validated on an external cohort of 222 patients with PJI who subsequently failed with reinfection and 200 aseptic patients. The performance of the new criteria was compared to the established MSIS and the prior International Consensus Meeting definitions.

      Results

      Two positive cultures or the presence of a sinus tract were considered as major criteria and diagnostic of PJI. The calculated weights of an elevated serum CRP (>1 mg/dL), D-dimer (>860 ng/mL), and erythrocyte sedimentation rate (>30 mm/h) were 2, 2, and 1 points, respectively. Furthermore, elevated synovial fluid white blood cell count (>3000 cells/μL), alpha-defensin (signal-to-cutoff ratio >1), leukocyte esterase (++), polymorphonuclear percentage (>80%), and synovial CRP (>6.9 mg/L) received 3, 3, 3, 2, and 1 points, respectively. Patients with an aggregate score of greater than or equal to 6 were considered infected, while a score between 2 and 5 required the inclusion of intraoperative findings for confirming or refuting the diagnosis. Intraoperative findings of positive histology, purulence, and single positive culture were assigned 3, 3, and 2 points, respectively. Combined with the preoperative score, a total of greater than or equal to 6 was considered infected, a score between 4 and 5 was inconclusive, and a score of 3 or less was not infected. The new criteria demonstrated a higher sensitivity of 97.7% compared to the MSIS (79.3%) and International Consensus Meeting definition (86.9%), with a similar specificity of 99.5%.

      Conclusion

      This study offers an evidence-based definition for diagnosing hip and knee PJI, which has shown excellent performance on formal external validation.

      Keywords

      Diagnosing periprosthetic joint infection (PJI) of the hip and knee remains a major challenge as there is no test with absolute accuracy [
      • Fernández-Sampedro M.
      • Fariñas-Alvarez C.
      • Garces-Zarzalejo C.
      • Alonso-Aguirre M.A.
      • Salas-Venero C.
      • Martínez-Martínez L.
      • et al.
      Accuracy of different diagnostic tests for early, delayed and late prosthetic joint infection.
      ,
      • Ahmad S.S.
      • Shaker A.
      • Saffarini M.
      • Chen A.F.
      • Hirschmann M.T.
      • Kohl S.
      Accuracy of diagnostic tests for prosthetic joint infection: a systematic review.
      ]. The diagnosis of PJI is based on a combination of clinical findings, laboratory results from peripheral blood and synovial fluid, microbiological culture, histological evaluation of periprosthetic tissue, and intraoperative findings.
      The Musculoskeletal Infection Society (MSIS) and the Infectious Diseases Society (IDSA) have previously developed criteria to standardize the definition of PJI [
      • Osmon D.R.
      • Berbari E.F.
      • Berendt A.R.
      • Lew D.
      • Zimmerli W.
      • Steckelberg J.M.
      • et al.
      Executive summary: diagnosis and management of prosthetic joint infection: clinical practice guidelines by the Infectious Diseases Society of America.
      ,
      • Parvizi J.
      • Zmistowski B.
      • Berbari E.F.
      • Bauer T.W.
      • Springer B.D.
      • Della Valle C.J.
      • et al.
      New definition for periprosthetic joint infection: from the Workgroup of the musculoskeletal infection society.
      ]. A further consensus meeting in 2013 endorsed the MSIS definition of PJI and modified it slightly [
      • Parvizi J.
      • Gehrke T.
      • Chen A.F.
      Proceedings of the International consensus on periprosthetic joint infection.
      ]. These definitions have become widely accepted among surgeons worldwide and have dramatically improved diagnostic confidence and facilitated treatment. Moreover, their use in research allowed for consistency in definition between studies and enhanced the potential for collaboration and the overall quality of literature. However, existing guidelines were largely generated by expert opinions and have not been validated. Furthermore, while relatively specific, there is concern about the sensitivity of the current definitions [
      • Koh I.J.
      • Cho W.-S.
      • Choi N.Y.
      • Parvizi J.
      • Kim T.K.
      How accurate are orthopedic surgeons in diagnosing periprosthetic joint infection after total knee arthroplasty?: a multicenter study.
      ].
      Although definite evidence or major criteria for infection are identical between the different definitions, the supportive evidence or minor criteria differ and are less agreed upon. In the recent years, numerous markers have been evaluated and become available [
      • Patel R.
      • Alijanipour P.
      • Parvizi J.
      Advancements in diagnosing periprosthetic joint infections after total hip and knee arthroplasty.
      ,
      • Deirmengian C.
      • Kardos K.
      • Kilmartin P.
      • Cameron A.
      • Schiller K.
      • Parvizi J.
      Diagnosing periprosthetic joint infection: has the era of the biomarker arrived?.
      ,
      • Lee Y.S.
      • Koo K.-H.
      • Kim H.J.
      • Tian S.
      • Kim T.Y.
      • Maltenfort M.G.
      • et al.
      Synovial fluid biomarkers for the diagnosis of periprosthetic joint infection: a systematic review and meta-analysis.
      ], including serum D-dimer [
      • Shahi A.
      • Kheir M.M.
      • Tarabichi M.
      • Hosseinzadeh H.R.S.
      • Tan T.L.
      • Parvizi J.
      Serum D-dimer test is promising for the diagnosis of periprosthetic joint infection and timing of reimplantation.
      ], synovial leukocyte esterase (LE) [
      • Wyatt M.C.
      • Beswick A.D.
      • Kunutsor S.K.
      • Wilson M.J.
      • Whitehouse M.R.
      • Blom A.W.
      The alpha-defensin immunoassay and leukocyte esterase colorimetric strip test for the diagnosis of periprosthetic infection: a systematic review and meta-analysis.
      ,
      • Tischler E.H.
      • Cavanaugh P.K.
      • Parvizi J.
      Leukocyte esterase strip test: matched for musculoskeletal infection society criteria.
      ,
      Diagnosis of periprosthetic joint infection: the utility of a simple yet unappreciated enzyme.
      ], synovial alpha-defensin [
      • Deirmengian C.
      • Kardos K.
      • Kilmartin P.
      • Cameron A.
      • Schiller K.
      • Parvizi J.
      Combined measurement of synovial fluid α-Defensin and C-reactive protein levels: highly accurate for diagnosing periprosthetic joint infection.
      ,
      • Sigmund I.K.
      • Holinka J.
      • Gamper J.
      • Staats K.
      • Böhler C.
      • Kubista B.
      • et al.
      Qualitative α-defensin test (Synovasure) for the diagnosis of periprosthetic infection in revision total joint arthroplasty.
      ], synovial C-reactive protein (CRP) [
      • Tetreault M.W.
      • Wetters N.G.
      • Moric M.
      • Gross C.E.
      • Della Valle C.J.
      Is synovial C-reactive protein a useful marker for periprosthetic joint infection?.
      ,
      • Omar M.
      • Ettinger M.
      • Reichling M.
      • Petri M.
      • Guenther D.
      • Gehrke T.
      • et al.
      Synovial C-reactive protein as a marker for chronic periprosthetic infection in total hip arthroplasty.
      ], and molecular techniques such as next-generation sequencing [
      • Tarabichi M.
      • Shohat N.
      • Goswami K.
      • Alvand A.
      • Silibovsky R.
      • Belden K.
      • et al.
      Diagnosis of periprosthetic joint infection: the potential of next-generation sequencing.
      ]. Moreover, publications in the recent years have shown different weights (sensitivity and specificity) for the various tests used [
      • Lee Y.S.
      • Koo K.-H.
      • Kim H.J.
      • Tian S.
      • Kim T.Y.
      • Maltenfort M.G.
      • et al.
      Synovial fluid biomarkers for the diagnosis of periprosthetic joint infection: a systematic review and meta-analysis.
      ,
      • Shahi A.
      • Tan T.L.
      • Kheir M.M.
      • Tan D.D.
      • Parvizi J.
      Diagnosing periprosthetic joint infection: and the winner is?.
      ] and highlight the value of a high pretest probability in the overall diagnosis [
      • Deirmengian C.
      • Kardos K.
      • Kilmartin P.
      • Cameron A.
      • Schiller K.
      • Parvizi J.
      Combined measurement of synovial fluid α-Defensin and C-reactive protein levels: highly accurate for diagnosing periprosthetic joint infection.
      ,
      • Sousa R.
      • Serrano P.
      • Gomes Dias J.
      • Oliveira J.C.
      • Oliveira A.
      Improving the accuracy of synovial fluid analysis in the diagnosis of prosthetic joint infection with simple and inexpensive biomarkers: C-reactive protein and adenosine deaminase.
      ,
      • Tarabichi M.
      • Fleischman A.N.
      • Shahi A.
      • Tian S.
      • Parvizi J.
      Interpretation of leukocyte esterase for the detection of periprosthetic joint infection based on serologic markers.
      ].
      These advancements in the field of PJI diagnosis call for the modification of current diagnostic criteria to an evidence-based one that is inclusive of the recent tests and considers the relative weights of the different tests. The purpose of this multi-institutional study was, thus, to: (1) generate an evidence-based, weight-adjusted scoring system for the definition of PJI of hip and knee, (2) validate it on an external cohort, and (3) compare its performance against currently available definitions.

      Materials and Methods

      After the institutional review board approval, we conducted a retrospective review of the medical records of all patients undergoing revision total hip arthroplasty (THA) and total knee arthroplasty (TKA) arthroplasty from 3 academic centers between January 2001 and July 2016. We excluded patients without serum erythrocyte sedimentation rate (ESR) or serum CRP, as well as those without a joint aspiration or an attempt at aspiration. Patients in whom the aspiration or serum testing was performed more than 8 weeks before surgery were also excluded.

       Patient Population

       Developmental Model

      Patients were classified as having a PJI if they met major diagnostic criteria of MSIS and International Consensus Meeting (ICM) [
      • Osmon D.R.
      • Berbari E.F.
      • Berendt A.R.
      • Lew D.
      • Zimmerli W.
      • Steckelberg J.M.
      • et al.
      Executive summary: diagnosis and management of prosthetic joint infection: clinical practice guidelines by the Infectious Diseases Society of America.
      ,
      • Parvizi J.
      • Zmistowski B.
      • Berbari E.F.
      • Bauer T.W.
      • Springer B.D.
      • Della Valle C.J.
      • et al.
      New definition for periprosthetic joint infection: from the Workgroup of the musculoskeletal infection society.
      ,
      • Parvizi J.
      • Gehrke T.
      • Chen A.F.
      Proceedings of the International consensus on periprosthetic joint infection.
      ], namely the presence of a sinus tract (with evidence of communication to the joint or visualization of the prosthesis) or 2 positive cultures isolating the same pathogen from the periprosthetic tissue or synovial fluid samples. Patients coded as infected who did not meet these definitions and those with megaprosthesis or missing surgical data were excluded from the study. In addition, we excluded acute PJI cases, defined as occurring less than 3 months from the index arthroplasty, and acute hematogenous PJI, defined as acute symptoms occurring for less than 6 weeks but more than 3 months from index surgery. Aseptic revisions were defined as cases undergoing single-stage revision for a diagnosis other than infection (loosening, wear, instability, malalignment, adverse local tissue reactions, or other aseptic causes) who did not fail with infection, nor had any further reoperation on the same joint.

       Validation of the New Criteria

      We performed external validation on separate patients from the same 3 institutions, who were not included in the initial developmental model. This validation was performed on a group of PJI and aseptic patients.

       PJI Cases

      As there is no “gold standard” for PJI, we chose a representative sample of infected cases that was independent from any intrinsic bias from the commonly used definitions for infection. This group composed of patients who were treated as PJI cases (undergoing 2-stage revision THA and TKA) and failed with a reinfection within 1 year. All these patients were coded as infected (based on the International Classification of Diseases, Ninth Revision codes for PJI: 996.6, 996.66, 996.67, 998.5, 100, and 998.59). Data from the first infection were documented.

       Aseptic Cases

      A randomly selected holdout sample of 200 aseptic revisions was excluded from the developmental model for validation purposes. These patients met the same aforementioned criteria for aseptic revision and did not fail with infection within 1 year after surgery.

       Data Collection

      Patient characteristics, comorbidities, laboratory results (serum, synovial, and microbial), and intraoperative findings (purulence and histopathology) were documented (see Appendix A). Laboratory values and histopathology results were dichotomized as elevated or not based on the ICM cutoffs.5 For markers not present in the ICM definition (serum D-dimer, synovial CRP), a cutoff was determined using the Gini index [
      • Lerman R.I.
      • Yitzhaki S.
      A note on the calculation and interpretation of the Gini index.
      ].

       Statistical Analysis

      To maximize applicability to clinical practice, diagnostic markers were evaluated in a stepwise classification model based on the American Academy of Orthopaedic Surgeons guidelines accounting for simplicity and invasiveness [
      • Parvizi J.
      • Della Valle C.J.
      AAOS Clinical Practice Guideline: diagnosis and treatment of periprosthetic joint infections of the hip and knee.
      ]. This stepwise approach allowed us to take into account the pretest probability for infection and minimize missing data and the use of imputed data [
      • Fox-Wasylyshyn S.M.
      • El-Masri M.M.
      Handling missing data in self-report measures.
      ]. Missing data were filled by 10 imputations using multiple imputation by chained equations [
      • Buuren S van
      • Groothuis-Oudshoorn C.G.M.
      Mice: multivariate imputation by chained equations in R.
      ]. The probability for infection was evaluated independently for each step; progression from one step to another was determined to minimize false positive and negatives. The discriminatory capability of each step was then assessed by receiver-operating characteristic curve analysis. Area under the curve (AUC) scores were typically considered acceptable if the AUC exceeded 0.7, with an AUC of 0.5 representing a poor test (toss of a coin) and an AUC of 1.0 signifying a perfect test. The first step included serum markers (CRP, D-dimer, and ESR). Subjects with an extremely low probability for PJI would not proceed to step 2. Step 2 included testing of synovial markers which requires more invasive testing (synovial fluid white blood cell [WBC] count, polymorphonuclear percentage, CRP, LE, and alpha-defensin). Subjects with an extremely low and high probability for infection would not proceed to step 3. The evaluation of intraoperative findings (histology, purulence, and single positive culture) was performed in step 3. Single culture was evaluated in step 3 (and not in step 2) as patients reaching this point already have a high index of suspicion (pretest probability) for infection, thus minimizing the chance for false positive cultures and increasing its overall performance.
      For each step, a random forest analysis was performed to evaluate the relative weight and importance of each examined variable. Random forest is a robust method for ranking the prediction ability of the different variables [
      • Liaw A.
      • Wiener M.
      ]. The variables within each step were then evaluated in a multivariate logistic regression model. An integer point scoring system was created on the basis of the final beta coefficients; to make the scoring more practical and user friendly, beta coefficients were rounded to the nearest integer. Variables showing a high degree of colinearity (r > 0.7) were grouped together within a single criterion. Variables in the regression model which were not statistically significant were kept and scored, as we aimed to provide a scoring system that could be used when all markers are not available and not to guide which markers should be used.
      A preoperative diagnostic score incorporating stages 1 and 2 and an intraoperative diagnostic score incorporating patients with an ambiguous preoperative diagnostic score and intraoperative findings (stage 3) were then created. Relative scores were assigned based on the previously mentioned beta coefficients. Diagnostic cut-points were determined and validated using the holdout sample of 200 aseptic revisions and 222 PJI cases as predefined. No imputations were used for missing markers/tests. The diagnostic performance (false positive, negative, true positives and negatives, sensitivity and specificity) of the final scoring system was evaluated and compared to the traditional MSIS and ICM definitions. All statistical analyses were performed using R version 3.4.3.

      Results

      Overall 1504 patients (684 PJI and 820 aseptic revisions) were included in the developmental model. Of these, 663 (44.1%) underwent THA and 841 (55.9%) underwent TKA. The average age was 65.4 years (standard deviation 10.9), and 718 (47.7%) were male. Patients in the infected group were more likely to be male (P < .001), had most recently undergone a revision (as opposed to a primary procedure, P < .001), had undergone knee as opposed to hip arthroplasty (P = .009), and had a higher Charlson Comorbidity Index with a history of rheumatoid arthritis, diabetes, and malignancy (P < .001, Table 1).
      Table 1Characteristics of Patients Who Were Included in the Developmental Model (n = 1504).
      VariableOverall (n = 1504)PJI Cohort (n = 684)Aseptic Cohort (n = 820)P Value
      Age (y)65.4 (10.9)65.9 (11.0)64.9 (10.8).079
      Gender (male)718 (47.7%)366 (53.5%)352 (42.9%)<.001
      Statistically significant.
      Race (white)1270 (84.4%)569 (83.2%)70 (85.5%).05
      Joint (knee)841 (55.9%)409 (59.8%)432 (52.7%).009
      Statistically significant.
      Time from the most recent surgery (yr)6.0 (8.7)4.3 (9.3)7.4 (7.9)<.001
      Statistically significant.
      Most recent surgery–revision procedure416 (27.7%)284 (41.5%)132 (16.1%)<.001
      Statistically significant.
      Body mass index (kg/m2)31.1 (6.8)31.4 (7.5)30.9 (6.1).192
      Charlson Comorbidity Index (mean)1.80 (1.8)2.2 (1.7)1.3 (1.8)<.001
      Statistically significant.
       History of rheumatoid arthritis99 (6.6%)62 (9.1%)37 (4.5%)<.001
      Statistically significant.
       History of malignancy70 (4.7)57 (8.3%)s13 (1.6%)<.001
      Statistically significant.
       History of diabetes261 (17.4%)152 (22.2%)109 (13.3%)<.001
      Statistically significant.
      Data are presented as mean (standard deviation) or number (%); kilogram (kg); meter (m); year (yr).
      PJI, periprosthetic joint infection.
      a Statistically significant.

       Development of the Scoring Model

      In order of importance, the random forest analysis determined that elevated serum CRP (>1 mg/dL), D-dimer (>860 ng/mL), and ESR (>30 mm/h) as the most important variables associated with PJI in that order (Table 2); the beta coefficients were rounded to 2, 2, and 1, respectively. Owing to high collinearity between serum CRP and D-dimer (r = 0.72), they were grouped together within a single criterion. Further results from the random forest analysis of synovial markers pointed out elevated synovial fluid WBC count (>3000), alpha-defensin (signal-to-cutoff ratio >1), LE (++), polymorphonuclear percentage (>80%), and synovial CRP (6.9 mg/L) as the most important variables associated with PJI in that order (Table 2); the beta coefficients were rounded to 3, 3, 3, 2, and 1, respectively. Owing to high collinearity between synovial WBC and LE (r = 0.75), they were grouped together within a single criterion. The overall performance showed an AUC of 0.99. The relative scores from both steps were incorporated to a preoperative diagnostic score with patients having a score of greater than or equal to 6 considered infected, while patients with a score of 0 or 1 classified as not infected (Fig. 1).
      Table 2Simple Importance Based on Random Forest and Beta Coefficients Derived From a Multivariate Regression Analysis of Each Step.
      StepRandom ForestBetaStandard ErrorP ValueScore
      Step 1
       Serum CRP >1 mg/dL
      The following demonstrated a high collinearity (r > 0.7) and thus were grouped into a single criterion in the final model.
      1982.480.28<.0012
       Serum D-dimer > 860 ng/mL
      The following demonstrated a high collinearity (r > 0.7) and thus were grouped into a single criterion in the final model.
      1342.410.62<.0012
       Serum ESR >30 mm/h1121.390.29<.0011
      Step 2
       Synovial WBC count >3000 (cells/μL)
      The following demonstrated a high collinearity (r > 0.7) and thus were grouped into a single criterion in the final model.
      1092.650.80.0013
       Synovial alpha-defensin792.641.24.0413
       Synovial LE (++)
      The following demonstrated a high collinearity (r > 0.7) and thus were grouped into a single criterion in the final model.
      632.561.02.0173
       Synovial PMN% >80%471.730.92.1212
       Synovial CRP >6.9 mg/L220.851.12.4491
      Step 3
       Histology
      Greater than 5 neutrophils per high-power field in 5 high-power fields observed from histologic analysis of periprosthetic tissue at 400× magnification.
      173.211.02.0023
       Purulence123.471.32.0073
       Single culture82.251.45.1222
      CRP, C-reactive protein; ESR, erythrocyte sedimentation rate; LE, leukocyte esterase; PMN%, polymorphonuclear %; WBC, white blood cell.
      a The following demonstrated a high collinearity (r > 0.7) and thus were grouped into a single criterion in the final model.
      b Greater than 5 neutrophils per high-power field in 5 high-power fields observed from histologic analysis of periprosthetic tissue at 400× magnification.
      Figure thumbnail gr1
      Fig. 1New scoring based definition for periprosthetic joint infection (PJI). Proceed with caution in: adverse local tissue reaction, crystal deposition disease, slow growing organisms. CRP, C-reactive protein; ESR, erythrocyte sedimentation rate; LE, leukocyte esterase; PMN, polymorphonuclear; WBC, white blood cell. aFor patients with inconclusive minor criteria, operative criteria can also be used to fulfill definition for PJI. bConsider further molecular diagnostics such as next-generation sequencing.
      In cases where the preoperative score was between 2 and 5, intraoperative variables were considered to attempt to classify the patient as infected or aseptic. The intraoperative findings listed in descending order of importance were positive histology (3 points), purulence (3 points), and a single positive culture (2 points) (Table 2). The ability to discriminate PJI from aseptic revisions at this third stage alone was lower, with an AUC of 0.92. However, taking the results of the preoperative scores into account resulted in an AUC of 0.97. Thus, the relative preoperative scores were incorporated into the intraoperative diagnostic score with patients having an aggregate score (combined preoperative and intraoperative) of greater than or equal to 6 classified as infected, scores of 4-5 as inconclusive, and scores of 3 or less not infected (Fig. 1).

       External Validation

      The validation cohort included 222 infected and 200 aseptic revisions. Patient characteristics are presented in Appendix B. Of the PJI cohort, 212/222 (95.5%) were correctly diagnosed (true positives) as infected, 5/222 (2.3%) were falsely diagnosed as uninfected (false negatives), and in 5/222 (2.3%), no definite diagnosis could be made. In the aseptic cohort, 195/200 (97.5%) were correctly diagnosed as not infected (true negatives), 1/200 (0.5%) were falsely diagnosed as infected (false positives), and in 4/200 (2.0%), no definite diagnosis could be made.

       Relative Performance Compared With Current Definitions

      The overall sensitivity and specificity of the scoring system was 97.7% (95% confidence interval [CI] 94.7%-99.3%) and 99.5% (95% CI 97.3%-99.99%), respectively. Our scoring system demonstrated a higher sensitivity compared to the ICM (86.9%, 95% CI 81.8%-91.1%) and MSIS (79.3%, 95% CI 73.4%-84.4%) definitions, with similar specificity (Table 3). The sensitivity and specificity remained high in a subgroup analysis examining hips and knees separately (Appendix C), as well as for each of the 3 different intuitions individually (Appendix D).
      Table 3Performance of the New Definition Compare With the Traditionally Used Musculoskeletal Infection Society (MSIS) and International Consensus Meeting (ICM) Criteria.
      CriteriaPJI Cohort (n = 222)Aseptic Cohort (n = 200)Sensitivity (95% CI)Specificity (95% CI)
      True PositivesFalse NegativesInconclusiveTrue NegativeFalse PositivesInconclusive
      MSIS (2011)176 (79.3%)46 (20.7%)-199 (99.5%)1 (0.5%)-79.3% (73.4-84.4)99.5% (97.3-99.99)
      ICM (2013)193 (86.9%)29 (13.1%)-199 (99.5%)1 (0.5%)-86.9% (81.8-91.1)99.5% (97.3-99.99)
      New definition (2018)212 (95.5%)5 (2.3%)5 (2.3%)195 (97.5%)1 (0.5%)4 (2.0%)97.7% (94.7-99.3)99.5% (97.2-99.99)
      CI, confidence interval; PJI, periprosthetic joint infection.

      Discussion

      In the absence of a test with absolute accuracy, the diagnosis of a clinical condition needs to rely on a combination of criteria. This scenario is common in medicine, with the diagnosis of numerous conditions such as ankylosing spondylitis, rheumatoid arthritis, endocarditis, and sepsis all relying on diagnostic criteria. PJI is also one such clinical entity, the diagnosis of which has relied on the presence of criteria. Up until recently and before the introduction of the diagnostic criteria by MSIS and IDSA, the definition of PJI was fairly subjective and at the discretion of the treating clinician. Furthermore, research investigations also suffered from lack of standardized criteria. The efforts invested by the MSIS and the IDSA in introducing the diagnostic criteria for PJI brought standardization and uniformity into the field that has served many patients. Over the last few years, extensive research efforts were invested in diagnosis of PJI, and a few novel tests have been introduced [
      • Deirmengian C.
      • Kardos K.
      • Kilmartin P.
      • Cameron A.
      • Schiller K.
      • Parvizi J.
      Combined measurement of synovial fluid α-Defensin and C-reactive protein levels: highly accurate for diagnosing periprosthetic joint infection.
      ,
      • Tarabichi M.
      • Shohat N.
      • Goswami K.
      • Alvand A.
      • Silibovsky R.
      • Belden K.
      • et al.
      Diagnosis of periprosthetic joint infection: the potential of next-generation sequencing.
      ,
      • Parvizi J.
      • Jacovides C.
      • Antoci V.
      • Ghanem E.
      Diagnosis of periprosthetic joint infection: the utility of a simple yet unappreciated enzyme.
      ].
      The objective of this study was to conduct a comprehensive review of the literature to identify tests that may have a role in diagnosis of PJI and devise diagnostic criteria for PJI that may improve the accuracy of the prior diagnostic criteria further. A group of established biostatisticians with expertise in bioinformatics and statistics from the University of Tel Aviv were recruited to help us in this initiative. The intention was to validate the newly developed diagnostic criteria on a holdout cohort of patients from 3 academic institutions. In addition, by using appropriate statistical methods, we were able to assign a relative and quantitative weight for established and newly introduced tests. We were able to confirm that the newly introduced diagnostic criteria will have better performance, in terms of accuracy, compared to previous MSIS and ICM definitions.
      The new diagnostic criteria were introduced to address a few of the limitations of the prior definitions. First, the prior definitions do not consider chronicity or invasiveness of the diagnostic tests. For instance, 3 of the 6 minor criteria within the current MSIS definition are typically intraoperative findings (purulence, single culture growth, and positive histology). This can make the preoperative diagnosis of infection extremely difficult. Second, novel tests that allow us to reach a preoperative diagnosis more easily and with high diagnostic performance are not accounted for [
      • Lee Y.S.
      • Koo K.-H.
      • Kim H.J.
      • Tian S.
      • Kim T.Y.
      • Maltenfort M.G.
      • et al.
      Synovial fluid biomarkers for the diagnosis of periprosthetic joint infection: a systematic review and meta-analysis.
      ,
      • Wyatt M.C.
      • Beswick A.D.
      • Kunutsor S.K.
      • Wilson M.J.
      • Whitehouse M.R.
      • Blom A.W.
      The alpha-defensin immunoassay and leukocyte esterase colorimetric strip test for the diagnosis of periprosthetic infection: a systematic review and meta-analysis.
      ,
      • Tischler E.H.
      • Cavanaugh P.K.
      • Parvizi J.
      Leukocyte esterase strip test: matched for musculoskeletal infection society criteria.
      ]. Third, the prior definitions do not account for interplay between individual test results or their combined influence upon pretest probability. The additive effects of positive diagnostic markers has been demonstrated in multiple studies to substantially affect post-test probability [
      • Deirmengian C.
      • Kardos K.
      • Kilmartin P.
      • Cameron A.
      • Schiller K.
      • Parvizi J.
      Combined measurement of synovial fluid α-Defensin and C-reactive protein levels: highly accurate for diagnosing periprosthetic joint infection.
      ,
      • Sousa R.
      • Serrano P.
      • Gomes Dias J.
      • Oliveira J.C.
      • Oliveira A.
      Improving the accuracy of synovial fluid analysis in the diagnosis of prosthetic joint infection with simple and inexpensive biomarkers: C-reactive protein and adenosine deaminase.
      ,
      • Tarabichi M.
      • Fleischman A.N.
      • Shahi A.
      • Tian S.
      • Parvizi J.
      Interpretation of leukocyte esterase for the detection of periprosthetic joint infection based on serologic markers.
      ]. By using a stepwise approach in developing the current criteria, which was based on the current American Academy of Orthopaedic Surgeons guidelines, we were able to assess the relative weights of these tests and take into account their pretest probability. Our score can be used to diagnose patients within the preoperative period more easily. It is also more accurate for those ambiguous cases with a high pretest probability, for which we provide a tool for an intraoperative diagnosis. The stepwise approach also enabled us to avoid false positives, which may have become an issue with increased sensitivity. Finally, external validation confirmed higher sensitivity compared to the established MSIS and ICM definitions with no significant increase in false positives. The fact that many of these infected patients had only a limited number of tests/markers available highlights the applicability of this proposed definition.
      Another novel finding of the present definition is the introduction of patients in which a diagnosis is inconclusive. These patients are often encountered in clinical practice and represent a real diagnostic challenge. Pointing out this unique group or “gray area” of patients promotes awareness in both clinical practice and the need for further research focused at this cohort. Interestingly, all patients with an inconclusive diagnosis had negative cultures and may thus benefit from molecular diagnostic testing such as next-generation sequencing which has recently shown promising results [
      • Tarabichi M.
      • Shohat N.
      • Goswami K.
      • Alvand A.
      • Silibovsky R.
      • Belden K.
      • et al.
      Diagnosis of periprosthetic joint infection: the potential of next-generation sequencing.
      ].
      This study is not without limitations. The retrospective design and incorporation of data from multiple centers may have resulted in variation in both data collection and reporting. While we implemented a stepwise stratification approach, which allowed us to minimize missing data, some data were imputed which may have limited the power of certain variables. Nevertheless, the scoring system was validated on an external model and remained very sensitive and specific. Another point to make is that we deliberately included certain variables in the final model which were not statistically significant. The reason for this was to maximize usability for physicians by incorporating markers used in everyday practice. It is, therefore, important to emphasize that the present scoring system is not designed or intended to be used as a guide for which tests should be ordered; rather, it should be used as a tool to diagnose patients when a panel of tests are already available. Furthermore, it is important to stress that the major diagnostic criteria was used as the “gold standard” for PJI in the development of this model. This group may have a more overt infection that may have affected the value of each score in the final model. Nevertheless, we tried to minimize bias by using the standard ICM threshold values for each individual test (Table 4). Another limitation relates to the use of chronic infections for model development and validation. While we believe these criteria should apply also for acute infections, both the scoring system and proposed thresholds require further validation. Finally, while we show an excellent performance, clinical judgment should still prevail and guide physicians in management of patients. There may be situations when a patient is infected and does not meet the diagnostic criteria and vice versa (Table 5) [
      • Yi P.H.
      • Cross M.B.
      • Moric M.
      • Levine B.R.
      • Sporer S.M.
      • Paprosky W.G.
      • et al.
      Do serologic and synovial tests help diagnose infection in revision hip arthroplasty with metal-on-metal bearings or corrosion?.
      ].
      Table 4Proposed Thresholds Based on the 2013 ICM Combined With Current Findings.
      MarkerChronic (>90 d)Acute (<90 d)
      Serum CRP (mg/dL)1.010
      Serum D-dimer (ng/mL)860860
      Further studies are needed to validate a specific threshold.
      Serum ESR (mm/h)30-
      Synovial WBC count (cells/μL)300010,000
      Synovial PMN (%)8090
      Synovial CRP (mg/L)6.9
      Further studies are needed to validate a specific threshold.
      6.9
      Synovial alpha-defensin (signal-to-cutoff ratio)1.01.0
      CRP, C-reactive protein; ESR, erythrocyte sedimentation rate; ICM, International Consensus Meeting; PMN, polymorphonuclear; WBC, white blood cell.
      a Further studies are needed to validate a specific threshold.
      Table 5Patients in Whom the Proposed Criteria May Be Inaccurate.
      Red Flag Patients
      Adverse local tissue reaction (ALTR)
      Crystalline deposition arthropathy
      Inflammatory arthropathy flare
      Infection with slow growing organisms
      Such as Propionibacterium acnes, coagulase negative Staphylococcus, and others.
      a Such as Propionibacterium acnes, coagulase negative Staphylococcus, and others.
      In conclusion, we present the first validated, evidence-based criteria for diagnosing PJI after hip and knee arthroplasty. We hope this new definition will guide clinicians and further improve the quality of research surrounding this devastating condition. Despite extensive investigations, diagnosis of the problem in some patients remains uncertain. These patients may benefit from the use of novel techniques such as next-generation sequencing.

      Appendix ASupplementary Data

      Appendix

      Appendix AData Collection.
      VariableComment
      Baseline characteristics
       Septic/aseptic revisionBased on predefined criteria
       Time from last surgery on the joint
       Primary/revision (most recent surgery)Was the latest surgery primary or a revision
       Gender
       Age
       Body mass index (kg/m2)
       JointHip or knee
       RaceWhite, black, or others
       Charlson Comorbidity IndexOverall score
       Rheumatoid arthritis
       Malignancy
       Diabetes
      Major criteria
       Sinus tractSinus tract with evidence of communication to the joint or visualization of the prosthesis
       2 positive culturesSame pathogen is isolated from at least 2 separate tissues or fluid samples obtained from the affected prosthetic joint
      Cultures
       Single positive culture
       Type of organism and resistance
      Serum markers
       Serum ESR (mm/h)
       Serum CRP (mg/dL)
       Serum D-dimer (μg/mL)
      Synovial markers
       Synovial fluid CRP (mg/L)
       Synovial fluid WBC count (cells/μL)
       Synovial fluid PMN (%)
       Synovial fluid LENegative, trace, (+), or (++)
       Synovial fluid alpha-defensin
      Intraoperative
       PurulenceAs described in the operative note
       Positive histopathologyGreater than 5 neutrophils per high-power field in 5 high-power fields observed from histologic analysis of periprosthetic tissue at 400× magnification.
       Positive molecular findings
      CRP, C-reactive protein; ESR, erythrocyte sedimentation rate; LE, leukocyte esterase; PMN, polymorphonuclear; WBC, white blood cell.
      Appendix BDemographics of Patients Which Were Included in the Validation Model (n = 422).
      VariableOverall (n = 422)PJI Cohort (n = 222)Aseptic Cohort (n = 200)P Value
      Age64.6 (10.6)64.364.9.606
      Gender (male)205 (48.6%)116 (52.3%)89 (44.5%).143
      Race (white)364 (86.3%)191 (86.0%)173 (86.5%).782
      Joint (knee)283 (67.1%)152 (68.5%)131 (65.5%).535
      Time from last surgery (yr)5.6 (8.9)4.2 (10.7)7.1 (6.1).001
      Statistically significant.
      Most recent surgery a revision procedure153 (36.3%)95 (42.8%)58 (29.0%)<.001
      Statistically significant.
      Body mass index (kg/m2)32.5 (7.8)33.1 (9.1)31.9 (6.0).09
      Charlson Comorbidity Index (mean)1.8 (1.9)2.01.7.102
      History of rheumatoid arthritis31 (7.3%)18 (8.1%)13 (6.5%).216
      History of malignancy17 (4.0%)12 (5.4%)5 (2.5%).13
      History of diabetes100 (23.7%)61 (27.5%)39 (19.5%).02
      Statistically significant.
      Data are presented as mean (standard deviation) or number (%); kilogram (kg); meter (m); year (yr).
      PJI, periprosthetic joint infection.
      a Statistically significant.
      Appendix CPerformance of the New Definition Stratified by Hip and Knee Joints.
      JointPJI Cohort (n = 222)Aseptic Cohort (n = 200)Sensitivity (95% CI)Specificity (95% CI)
      True PositivesFalse NegativesInconclusiveTrue NegativeFalse PositivesInconclusive
      Knees148/155 (95.5%)4/155 (2.6%)3/155 (1.9%)128/131 (97.7%)1/131 (0.8%)2/131 (1.5%)97.4% (93.4-99.3)99.2% (95.8-100)
      Hips64/67 (95.5%)1/67 (1.5%)2/67 (3.0%)67/69 (97.1%)0/69 (0%)2/69 (2.9%)98.5% (91.7-100)100% (94.6-100)
      CI, confidence interval; PJI, periprosthetic joint infection.
      Appendix DPerformance of the New Definition for Each of the 3 Different Institutions Individually.
      InstitutionPJI Cohort (n = 222)Aseptic Cohort (n = 200)Sensitivity (95% CI)Specificity (95% CI)
      True PositivesFalse NegativesInconclusiveTrue NegativeFalse PositivesInconclusive
      Center A41/42 (97.6%)0/42 (0%)1/42 (2.4%)45/46 (97.8%)0/46 (0%)1/46 (2.2%)100% (91.4-100)100% (92.1-100)
      Center B56/60 (93.3%)2/60 (3.3%)2/60 (3.3%)57/59 (96.6%)1/59 (1.7%)1/59 (1.7%)96.6% (88.1-99.6)98.3% (90.8-100)
      Center C115/120 (95.8%)3/120 (2.5%)2/120 (1.7%)93/95 (97.9%)0/95 (0%)2/95 (2.1%)97.5% (92.8-99.5)100% (96.1-100)
      CI, confidence interval; PJI, periprosthetic joint infection.

      References

        • Fernández-Sampedro M.
        • Fariñas-Alvarez C.
        • Garces-Zarzalejo C.
        • Alonso-Aguirre M.A.
        • Salas-Venero C.
        • Martínez-Martínez L.
        • et al.
        Accuracy of different diagnostic tests for early, delayed and late prosthetic joint infection.
        BMC Infect Dis. 2017; 17: 592https://doi.org/10.1186/s12879-017-2693-1
        • Ahmad S.S.
        • Shaker A.
        • Saffarini M.
        • Chen A.F.
        • Hirschmann M.T.
        • Kohl S.
        Accuracy of diagnostic tests for prosthetic joint infection: a systematic review.
        Knee Surg Sports Traumatol Arthrosc Off J ESSKA. 2016; 24: 3064-3074https://doi.org/10.1007/s00167-016-4230-y
        • Osmon D.R.
        • Berbari E.F.
        • Berendt A.R.
        • Lew D.
        • Zimmerli W.
        • Steckelberg J.M.
        • et al.
        Executive summary: diagnosis and management of prosthetic joint infection: clinical practice guidelines by the Infectious Diseases Society of America.
        Clin Infect Dis Off Publ Infect Dis Soc Am. 2013; 56: 1-10https://doi.org/10.1093/cid/cis966
        • Parvizi J.
        • Zmistowski B.
        • Berbari E.F.
        • Bauer T.W.
        • Springer B.D.
        • Della Valle C.J.
        • et al.
        New definition for periprosthetic joint infection: from the Workgroup of the musculoskeletal infection society.
        Clin Orthop. 2011; 469: 2992-2994https://doi.org/10.1007/s11999-011-2102-9
        • Parvizi J.
        • Gehrke T.
        • Chen A.F.
        Proceedings of the International consensus on periprosthetic joint infection.
        Bone Joint J. 2013; 95-B: 1450-1452https://doi.org/10.1302/0301-620X.95B11.33135
        • Koh I.J.
        • Cho W.-S.
        • Choi N.Y.
        • Parvizi J.
        • Kim T.K.
        How accurate are orthopedic surgeons in diagnosing periprosthetic joint infection after total knee arthroplasty?: a multicenter study.
        Knee. 2015; 22: 180-185https://doi.org/10.1016/j.knee.2015.02.004
        • Patel R.
        • Alijanipour P.
        • Parvizi J.
        Advancements in diagnosing periprosthetic joint infections after total hip and knee arthroplasty.
        Open Orthop J. 2016; 10: 654-661https://doi.org/10.2174/1874325001610010654
        • Deirmengian C.
        • Kardos K.
        • Kilmartin P.
        • Cameron A.
        • Schiller K.
        • Parvizi J.
        Diagnosing periprosthetic joint infection: has the era of the biomarker arrived?.
        Clin Orthop. 2014; 472: 3254-3262https://doi.org/10.1007/s11999-014-3543-8
        • Lee Y.S.
        • Koo K.-H.
        • Kim H.J.
        • Tian S.
        • Kim T.Y.
        • Maltenfort M.G.
        • et al.
        Synovial fluid biomarkers for the diagnosis of periprosthetic joint infection: a systematic review and meta-analysis.
        J Bone Joint Surg Am. 2017; 99: 2077-2084https://doi.org/10.2106/JBJS.17.00123
        • Shahi A.
        • Kheir M.M.
        • Tarabichi M.
        • Hosseinzadeh H.R.S.
        • Tan T.L.
        • Parvizi J.
        Serum D-dimer test is promising for the diagnosis of periprosthetic joint infection and timing of reimplantation.
        J Bone Joint Surg Am. 2017; 99: 1419-1427https://doi.org/10.2106/JBJS.16.01395
        • Wyatt M.C.
        • Beswick A.D.
        • Kunutsor S.K.
        • Wilson M.J.
        • Whitehouse M.R.
        • Blom A.W.
        The alpha-defensin immunoassay and leukocyte esterase colorimetric strip test for the diagnosis of periprosthetic infection: a systematic review and meta-analysis.
        J Bone Joint Surg Am. 2016; 98: 992-1000https://doi.org/10.2106/JBJS.15.01142
        • Tischler E.H.
        • Cavanaugh P.K.
        • Parvizi J.
        Leukocyte esterase strip test: matched for musculoskeletal infection society criteria.
        J Bone Joint Surg Am. 2014; 96: 1917-1920https://doi.org/10.2106/JBJS.M.01591
      1. Diagnosis of periprosthetic joint infection: the utility of a simple yet unappreciated enzyme.
        (Available at:https://www.ncbi.nlm.nih.gov/pubmed/?term=Parvizi+J%2C+Jacovides+C%2C+Antoci+V%2C+Ghanem+E%2C+Diagnosis+of+periprosthetic+joint+infection%3A+The+utility+of+a+simple+yet+unappreciated+enzyme)
        Date accessed: September 27, 2017
        • Deirmengian C.
        • Kardos K.
        • Kilmartin P.
        • Cameron A.
        • Schiller K.
        • Parvizi J.
        Combined measurement of synovial fluid α-Defensin and C-reactive protein levels: highly accurate for diagnosing periprosthetic joint infection.
        J Bone Joint Surg Am. 2014; 96: 1439-1445https://doi.org/10.2106/JBJS.M.01316
        • Sigmund I.K.
        • Holinka J.
        • Gamper J.
        • Staats K.
        • Böhler C.
        • Kubista B.
        • et al.
        Qualitative α-defensin test (Synovasure) for the diagnosis of periprosthetic infection in revision total joint arthroplasty.
        Bone Joint J. 2017; 99-B: 66-72https://doi.org/10.1302/0301-620X.99B1.BJJ-2016-0295.R1
        • Tetreault M.W.
        • Wetters N.G.
        • Moric M.
        • Gross C.E.
        • Della Valle C.J.
        Is synovial C-reactive protein a useful marker for periprosthetic joint infection?.
        Clin Orthop. 2014; 472: 3997-4003https://doi.org/10.1007/s11999-014-3828-y
        • Omar M.
        • Ettinger M.
        • Reichling M.
        • Petri M.
        • Guenther D.
        • Gehrke T.
        • et al.
        Synovial C-reactive protein as a marker for chronic periprosthetic infection in total hip arthroplasty.
        Bone Joint J. 2015; 97-B: 173-176https://doi.org/10.1302/0301-620X.97B2.34550
        • Tarabichi M.
        • Shohat N.
        • Goswami K.
        • Alvand A.
        • Silibovsky R.
        • Belden K.
        • et al.
        Diagnosis of periprosthetic joint infection: the potential of next-generation sequencing.
        J Bone Joint Surg Am. 2018; 100: 147-154https://doi.org/10.2106/JBJS.17.00434
        • Shahi A.
        • Tan T.L.
        • Kheir M.M.
        • Tan D.D.
        • Parvizi J.
        Diagnosing periprosthetic joint infection: and the winner is?.
        J Arthroplasty. 2017; 32: S232-S235https://doi.org/10.1016/j.arth.2017.06.005
        • Sousa R.
        • Serrano P.
        • Gomes Dias J.
        • Oliveira J.C.
        • Oliveira A.
        Improving the accuracy of synovial fluid analysis in the diagnosis of prosthetic joint infection with simple and inexpensive biomarkers: C-reactive protein and adenosine deaminase.
        Bone Joint J. 2017; 99-B: 351-357https://doi.org/10.1302/0301-620X.99B3.BJJ-2016-0684.R1
        • Tarabichi M.
        • Fleischman A.N.
        • Shahi A.
        • Tian S.
        • Parvizi J.
        Interpretation of leukocyte esterase for the detection of periprosthetic joint infection based on serologic markers.
        J Arthroplasty. 2017; 32: S97-S100.e1https://doi.org/10.1016/j.arth.2017.03.045
        • Lerman R.I.
        • Yitzhaki S.
        A note on the calculation and interpretation of the Gini index.
        Econ Lett. 1984; 15: 363-368https://doi.org/10.1016/0165-1765(84)90126-5
        • Parvizi J.
        • Della Valle C.J.
        AAOS Clinical Practice Guideline: diagnosis and treatment of periprosthetic joint infections of the hip and knee.
        J Am Acad Orthop Surg. 2010; 18: 771-772
        • Fox-Wasylyshyn S.M.
        • El-Masri M.M.
        Handling missing data in self-report measures.
        Res Nurs Health. 2005; 28: 488-495https://doi.org/10.1002/nur.20100
        • Buuren S van
        • Groothuis-Oudshoorn C.G.M.
        Mice: multivariate imputation by chained equations in R.
        J Stat Softw. 2011; 45
        • Liaw A.
        • Wiener M.
        Classification and Regression by RandomForest. Vol 23. 2001
        • Parvizi J.
        • Jacovides C.
        • Antoci V.
        • Ghanem E.
        Diagnosis of periprosthetic joint infection: the utility of a simple yet unappreciated enzyme.
        J Bone Joint Surg Am. 2011; 93: 2242-2248https://doi.org/10.2106/JBJS.J.01413
        • Yi P.H.
        • Cross M.B.
        • Moric M.
        • Levine B.R.
        • Sporer S.M.
        • Paprosky W.G.
        • et al.
        Do serologic and synovial tests help diagnose infection in revision hip arthroplasty with metal-on-metal bearings or corrosion?.
        Clin Orthop. 2015; 473: 498-505https://doi.org/10.1007/s11999-014-3902-5

      Linked Article