The Quality of Diagnostic Studies in Periprosthetic Joint Infections: Can We Do Better?

Probability Score for the Diagnosis of Periprosthetic Joint Infection: Development and Validation of a Practical Multi-analyte Machine Learning Model

Background and objective The diagnosis of periprosthetic joint infection (PJI) relies on established criteria-based systems requiring interpretation and combination of multiple laboratory tests into scoring systems. In routine clinical care, clinicians implement these algorithms to diagnose PJI. Existing literature indicates suboptimal adoption and implementation of these criteria in clinical practice, even among experts. Recognizing the need for accurate PJI diagnosis through proper synthesis of multiple laboratory parameters, this study aimed to develop and validate a machine learning (ML) model that generates a preoperative PJI probability score based solely on synovial fluid (SF) biomarkers within 24 hours. Materials and methods A two-stage ML model was constructed using 104,090 SF samples from 2,923 institutions (2018-2024). First, unsupervised learning identified natural clusters in the data to label samples as "infected" or "not infected." Then, these labels trained a supervised logistic regression model that generated PJI scores (0-100), categorizing cases as PJI positive (> 80), PJI negative (< 20), or equivocal (20-80). The model incorporated 10 SF biomarkers: specimen integrity markers (absorbance at 280 nm, red blood cell count), inflammatory markers (white blood cell count, percentage of neutrophils, SF C-reactive protein), a PJI-specific biomarker (alpha-defensin), and microbial antigen markers (Staphylococcus, Enterococcus, Candida, and Cutibacterium acnes). Notably, culture results were excluded to allow for a 24-hour diagnosis. After splitting data into training (n = 83,272) and validation (n = 20,818) cohorts, performance was assessed against modified 2018 International Consensus Meeting criteria, including evaluation with probabilistically reclassified "inconclusive" cases. Results The ML model and resulting PJI score showed high diagnostic accuracy in the validation cohort. The PJI score achieved 99.3% sensitivity and 99.5% specificity versus the clinical reference before reclassification of inconclusive cases and 98.1% sensitivity and 97.6% specificity after probabilistic reclassification. With a disease prevalence of 20.7%, the positive predictive value reached 91.5% and the negative predictive value 99.5%. The model resolved 95% (1,363/1,442) of samples deemed inconclusive by the clinical standard. The analysis identified alpha defensin, percentage of neutrophils, and white blood cell count as the most influential model features. The model performed well in culture-negative infections. Conclusions The ML model and resulting PJI score demonstrated exceptional diagnostic accuracy by leveraging unsupervised SF biomarker pattern clustering. The model substantially reduced diagnostic uncertainty by definitively classifying most inconclusive cases, revealing their natural alignment with infected or non-infected patterns. This performance was achieved without SF culture results, enabling definitive diagnostic information within 24 hours based solely on biomarkers. The clinical significance demonstrates that an ML algorithm can match the diagnostic accuracy of complex clinical standards while transferring analytical complexity from clinicians to laboratories, minimizing the implementation gap that hinders current criteria-based approaches.

The Albumin to Globulin Ratio Performs Well for Diagnosing Periprosthetic Joint Infection: A Single-Center Retrospective Study

BACKGROUND

Accurate diagnosis of the periprosthetic joint infection (PJI) remains a challenge for surgeons. The purpose of this study was to assess the value of albumin to globulin ratio (AGR) and globulin (GLB) for diagnosing PJI.

METHODS

A total of 182 patients undergoing revision after arthroplasty were included and divided into 2 groups, 61 in knee group (PJI: 38; non-PJI: 23) and 121 in hip group (PJI: 26; non-PJI: 95). We used receiver operating characteristic curves to determine the diagnostic value of AGR, GLB, inflammatory markers (erythrocyte sedimentation rate [ESR] and C-reactive protein [CRP]).

RESULTS

The receiver operating characteristic curves showed the areas under the curve of AGR, GLB, ESR, and CRP in the knee group were 0.940, 0.928, 0.867, and 0.848, respectively, and they were 0.855, 0.831, 0.886, and 0.912 in the hip group. The optimal predictive cut-off values for AGR in knee and hip groups were 1.375 and 1.295, respectively. The sensitivity and specificity of AGR, respectively, were 94.7% and 87.0% (knee group) and 84.6% and 75.8% (hip group) for diagnosing PJI. The sensitivity of "AGR or ESR" and specificity of "AGR and GLB" in the knee group were 99.6% and 98.9%, respectively.

CONCLUSION

For knee or hip groups, the AGR exhibits good value for the diagnosis of PJI comparable with ESR and CRP. The AGR and GLB, together with CRP and ESR, should be used as the preferred indicators for diagnosing PJI. The "AGR or ESR" and "AGR and GLB" in the knee group have an excellent diagnostic value in sensitivity and specificity, respectively.

Diagnostic value of albumin/fibrinogen ratio and C-reactive protein/albumin/globulin ratio for periprosthetic joint infection: a retrospective study

Background

The study aims to explore diagnostic value of albumin/fibrinogen ratio (AFR) and C-reactive protein (CRP)/albumin (ALB)/globulin (GLO) ratio (CAGR) for periprosthetic joint infection (PJI).

Methods

A retrospective analysis was conducted on clinical data collected from 190 patients who had joint replacement surgery in Qilu Hospital of Shandong University (Qingdao), from January 2017 to December 2022. Based on the occurrence of PJI after surgery, patients were divided as an infection group (10 cases) and non-infection group (180 cases). Diagnostic indicators were analyzed, univariate and multivariate logistic regression analyses were further performed to identify factors related to PJI. Sensitivity and specificity of AFR and CAGR, both individually and in combination, were calculated using ROC curves, and their diagnostic performance was compared based on the area under the curve (AUC).

Results

Levels of CRP, ESR, FIB, GLO, and CAGR were significantly higher in the infection group than in non-infection group (P < 0.05). Levels of ALB and AFR were significantly lower in infection group (P < 0.05). Multivariate logistic regression analysis reviewed that CRP (OR = 3.324), ESR (OR = 2.118), FIB (OR = 3.142), ALB (OR = 0.449), GLO (OR = 1.985), AFR (OR = 0.587), and CAGR (OR = 2.469) were factors influencing PJI (P < 0.05). The AUC for AFR and CAGR in diagnosing PJI were 0.739 and 0.780, while AUC for their combined detection was 0.858.

Conclusion

Abnormal levels of AFR and CAGR are associated with PJI, and their combined use has certain diagnostic value for PJI.

The optimal diagnostic cut-off of WBC and PMN counts for joint aspiration in periprosthetic joint infection is 2479/µL and 67%, respectively: ICM criteria thresholds are too high

BACKGROUND

Various organizations have published definitions for periprosthetic joint infection (PJI) with significant differences in the cut-offs of white blood cell (WBC) count and polymorphonuclear (PMN) leukocyte cells. Herein, we aim to analyze optimal cut-offs in patients which are planned to undergo a prosthesis revision and compare them with the actual published thresholds of the International Consensus Meeting (ICM) and European Bone and Joint Infection Society (EBJIS).

METHODS

A test kit was compiled in a monocentric prospective study, according to the ICM criteria (2018) and 2021 EBJIS criteria. The kit was implemented using: blood samples (including leukocyte count and C-reactive protein); samples for examining the synovial fluid (WBC count, PMN cell differentiation, microbiological culture for incubation over 14 days, alpha-defensin ELISA laboratory test, and leukocyte-esterase test). The cut-offs for WBC and PMN counts were investigated using ROC analyses and Youden index. The ICM 2018 criteria were applied, using alpha-defensin in all cases. Patients which have to undergo a prosthesis revision were included, a pre-operative joint aspiration had been performed, and the patients had been followed up prospectively.

RESULTS

405 patients were examined with the compiled test kit; 100% had a complete dataset with respect to alpha-defensin; 383 patients, according to WBC count; and 256, according to PMN cell differentiation The cut-off of 2478.89 cells/µl in the WBC count (sensitivity: 87.70%; specificity: 88.10%) and the cut-off of 66.99% in PMN differentiation showed the best accuracy (sensitivity: 86.00%; specificity: 88.80%). Other published cut-offs for WBC were tested in this cohort and showed the following accuracy: 3000/µl (EBJIS/ICM; sensitivity: 82.10%; specificity: 91.00%), 2000/µl (sensitivity: 89.60%; specificity: 83.40%), and 1500/µl (sensitivity: 91.50%; specificity: 75.00%). The published cut-offs for PMN had the following accuracy in this cohort: 80% (ICM; sensitivity: 66.3%; specificity: 96.50%), 70% (sensitivity: 82.6%; specificity: 90%), and 65% (EBJIS, sensitivity: 86%; specificity: 88.8%).

CONCLUSIONS

This study aims to improve current cut-offs for PMN- and WB-Count, even though PJI diagnosis is based on the combination of all defined tests. The optimal diagnostic cut-off of WBC and PMN counts was found to be 2479/µL and 67%, respectively, whereas ICM cut-offs in this cohort seem too high, as they provide high specificity but very low sensitivity. On the other hand, a cut-off for WBC count of 1500/µl alone would be very low, leading to low specificity and very high suspicion of PJI. The current consensus guidelines could be actualized considering these results to significantly improve the diagnostic quality.

LEVEL OF EVIDENCE

II.

Reliable Diagnostic Tests and Thresholds for Preoperative Diagnosis of Non-Inflammatory Arthritis Periprosthetic Joint Infection: A Meta-analysis and Systematic Review

Objective

The current diagnostic criteria for periprosthetic joint infection (PJI) are diverse and controversial, leading to delayed diagnosis. This study aimed to evaluate and unify their diagnostic accuracy and the threshold selection of serum and synovial routine tests for PJI at an early stage.

Methods

We searched the MEDLINE and Embase databases for retrospective or prospective studies which reported preoperative‐available assays (serum, synovial, or culture tests) for the diagnosis of chronic PJI among inflammatory arthritis (IA) or non‐IA populations from January 1, 2000 to June 30, 2022. Threshold effective analysis was performed on synovial polymorphonuclear neutrophils (PMN%), synovial white blood cell (WBC), serum C‐reactive protein (CRP), and erythrocyte sedimentation rate (ESR) to find the relevant cut‐offs.

Results

Two hundred and sixteen studies and information from 45,316 individuals were included in the final analysis. Synovial laboratory‐based α‐defensin and calprotectin had the best comprehensive sensitivity (0.91 [0.86–0.94], 0.95 [0.88–0.98]) and specificity (0.96 [0.94‐0.97], 0.95 [0.89–0.98]) values. According to the threshold effect analysis, the recommended cut‐offs are 70% (sensitivity 0.89 [0.85–0.92], specificity 0.90 [0.87–0.93]), 4100/μL (sensitivity 0.90 [0.87–0.93], specificity 0.97 [0.93–0.98]), 13.5 mg/L (sensitivity 0.84 [0.78–0.89], specificity 0.83 [0.73–0.89]), and 30 mm/h (sensitivity 0.79 [0.74–0.83], specificity 0.78 [0.72–0.83]) for synovial PMN%, synovial WBC, serum CRP, and ESR, respectively, and tests seem to be more reliable among non‐IA patients.

Conclusions

The laboratory‐based synovial α‐defensin and synovial calprotectin are the two best independent preoperative diagnostic tests for PJI. A cut off of 70% for synovial PMN% and tighter cut‐offs for synovial WBC and serum CRP could have a better diagnostic accuracy for non‐IA patients with chronic PJI.

The quality of diagnostic studies used for the diagnostic criteria of periprosthetic joint infections

BACKGROUND

Although periprosthetic joint infection (PJI) is a serious complication following a total joint arthroplasty procedure, there remains uncertainty regarding the diagnosis of PJI due to the lack of a globally accepted, standardized definition. The goal of this review is to critically analyze the quality of the evidence used for the novel 2018 MSIS PJI definition and identify gaps and limitations with using the Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) tool.

METHODS

References from the modified 2018 MSIS definition for PJI by Parvizi et al. were retrieved and manually reviewed. A total of 11 studies were assessed using a validated QUADAS-2 tool.

RESULTS

Many included studies had an unclear or high risk of bias for the Index Test domain due to a lack of blinding and lack of prespecified thresholds. A majority of studies utilized Youden's J statistic to optimize the thresholds which may diminish external validity. Likewise, several studies were assessed to have an unclear and high risk of bias for the Flow and Timing domain primarily due to a lack of reporting and a large number of exclusions. Overall, there was a low risk of bias for the choice of reference standard, its conduct and interpretation, as well as for the Patient Selection domain.

CONCLUSION

Although the literature used for the MSIS 2018 PJI definition is fraught with potential sources of bias, there may be a trend toward an improvement in the quality of evidence when compared to the earlier definition of PJI.

Sequencing of Circulating Microbial Cell-Free DNA Can Identify Pathogens in Periprosthetic Joint Infections

BACKGROUND

Over 1 million Americans undergo joint replacement each year, and approximately 1 in 75 will incur a periprosthetic joint infection. Effective treatment necessitates pathogen identification, yet standard-of-care cultures fail to detect organisms in 10% to 20% of cases and require invasive sampling. We hypothesized that cell-free DNA (cfDNA) fragments from microorganisms in a periprosthetic joint infection can be found in the bloodstream and utilized to accurately identify pathogens via next-generation sequencing.

METHODS

In this prospective observational study performed at a musculoskeletal specialty hospital in the U.S., we enrolled 53 adults with validated hip or knee periprosthetic joint infections. Participants had peripheral blood drawn immediately prior to surgical treatment. Microbial cfDNA from plasma was sequenced and aligned to a genome database with >1,000 microbial species. Intraoperative tissue and synovial fluid cultures were performed per the standard of care. The primary outcome was accuracy in organism identification with use of blood cfDNA sequencing, as measured by agreement with tissue-culture results.

RESULTS

Intraoperative and preoperative joint cultures identified an organism in 46 (87%) of 53 patients. Microbial cfDNA sequencing identified the joint pathogen in 35 cases, including 4 of 7 culture-negative cases (57%). Thus, as an adjunct to cultures, cfDNA sequencing increased pathogen detection from 87% to 94%. The median time to species identification for cases with genus-only culture results was 3 days less than standard-of-care methods. Circulating cfDNA sequencing in 14 cases detected additional microorganisms not grown in cultures. At postoperative encounters, cfDNA sequencing demonstrated no detection or reduced levels of the infectious pathogen.

CONCLUSIONS

Microbial cfDNA from pathogens causing local periprosthetic joint infections can be detected in peripheral blood. These circulating biomarkers can be sequenced from noninvasive venipuncture, providing a novel source for joint pathogen identification. Further development as an adjunct to tissue cultures holds promise to increase the number of cases with accurate pathogen identification and improve time-to-speciation. This test may also offer a novel method to monitor infection clearance during the treatment period.

LEVEL OF EVIDENCE

Diagnostic Level II. See Instructions for Authors for a complete description of levels of evidence.

What Is the Optimal Timing for Reading the Leukocyte Esterase Strip for the Diagnosis of Periprosthetic Joint Infection?

BACKGROUND

The leucocyte esterase (LE) strip test often is used to diagnose periprosthetic joint infection (PJI). In accordance with the manufacturer's directions, the LE strip test result is read 3 minutes after exposing it to joint fluid, but this has not been supported by robust research. Moreover, we have noted that the results of the LE strip test might change over time, and our previous studies have found that centrifugation causes the results of the LE strip test to degrade. Still, there is no evidence-based recommendation as to when to read the LE strip test to maximize diagnostic accuracy, in general, and the best reading times for the LE strip test before and after centrifugation need to be determined separately, in particular.

QUESTIONS/PURPOSES

(1) What is the optimal timing for reading LE strip test results before centrifugation to diagnose PJI? (2) What is the optimal timing for reading LE strip test results after centrifugation to diagnose PJI?

METHODS

This study was a prospective diagnostic trial. In all, 120 patients who were scheduled for revision arthroplasty and had signs of infection underwent joint aspiration in the outpatient operating room between July 2018 and July 2019 and were enrolled in this single-center study. For inclusion, patients must have had a diagnosis of PJI or nonPJI, valid synovial fluid samples, and must not have received antibiotics within 2 weeks before arthrocentesis. As such, 36 patients were excluded; 84 patients were included for analysis, and all 84 patients agreed to participate. The 2018 International Consensus Meeting Criteria (ICM 2018) was used for the classification of 49 patients with PJI (score ≥ 6) and 35 without PJI (score ≤ 2). The classification was used as the standard against which the different timings for reading LE strips were compared. All patients without PJI were followed for more than 1 year, during which they did not report the occurrence of PJI. All patients were graded against the diagnostic criteria regardless of their LE strip test results. In 83 patients, one drop of synovial fluid (50 μL) was applied to LE strips before and after centrifugation, and in one patient (without PJI), the sample was not centrifuged because the sample volume was less than 1.5 mL. The results of the strip test were read on an automated colorimeter. Starting from 1 minute after centrifugation, these strips were automatically read once every minute, 15 times (over a period of 16 minutes), and the results were independently recorded by two observers. Results were rated as negative, ±, 1+, and 2+ upon the machine reading. Grade 2+ (dark purple) was used as the threshold for a positive result. An investigator who was blinded to the study performed the statistics. Optimal timing for reading the LE strip before and after centrifugation was determined by using receiver operative characteristic (ROC) analysis. The specificity, sensitivity, and positive predictive and negative predictive values were calculated for key timepoints.

RESULTS

Before centrifugation, the area under the curve was the highest when the results were read at 5 minutes (0.90 [95% CI 0.83 to 0.98]; sensitivity 0.88 [95% CI 0.75 to 0.95]; specificity 0.89 [95% CI 0.72 to 0.96]). After centrifugation, the area under the curve was the highest when the results were read at 10 minutes (0.92 [95% CI 0.86 to 0.98]; sensitivity 0.65 [95% CI 0.50 to 0.78]; specificity 0.97 [95% CI 0.83 to 1.00]).

CONCLUSION

The LE strip test results are affected by time and centrifugation. For samples without centrifugation, we found that 5 minutes after application was the best time to read LE strips. We cannot deny the use of centrifuges because this is an effective way to solve the sample-mingling problem at present. We recommend 10 minutes postapplication as the most appropriate time to read LE strips after centrifugation. Multicenter and large-sample size studies are warranted to further verify our conclusion.

LEVEL OF EVIDENCE

Level II, diagnostic study.

Next generation sequencing for pathogen detection in periprosthetic joint infections

Periprosthetic joint infections (PJI) represent one of the most catastrophic complications following total joint arthroplasty (TJA). The lack of standardized diagnostic tests and protocols for PJI is a challenge for arthroplasty surgeons.Next generation sequencing (NGS) is an innovative diagnostic tool that can sequence microbial deoxyribonucleic acids (DNA) from a synovial fluid sample: all DNA present in a specimen is sequenced in parallel, generating millions of reads. It has been shown to be extremely useful in a culture-negative PJI setting.Metagenomic NGS (mNGS) allows for universal pathogen detection, regardless of microbe type, in a 24-48-hour timeframe: in its nanopore-base variation, mNGS also allows for antimicrobial resistance characterization.Cell-free DNA (cfDNA) NGS, characterized by lack of the cell lysis step, has a fast run-time (hours) and, together with a high sensitivity and specificity in microorganism isolation, may provide information on the presence of antimicrobial resistance genes.Metagenomics and cfDNA testing have reduced the time needed to detect infecting bacteria and represent very promising technologies for fast PJI diagnosis.NGS technologies are revolutionary methods that could disrupt the diagnostic paradigm of PJI, but a comprehensive collection of clinical evidence is still needed before they become widely used diagnostic tools. Cite this article: EFORT Open Rev 2021;6:236-244. DOI: 10.1302/2058-5241.6.200099.

Application of leukocyte esterase strip test in the screening of periprosthetic joint infections and prospects of high-precision strips

Periprosthetic joint infection (PJI) represents one of the most challenging complications after total joint arthroplasty (TJA). Despite the availability of a variety of diagnostic techniques, the diagnosis of PJI remains a challenge due to the lack of well-established diagnostic criteria. The leucocyte esterase (LE) strips test has been proved to be a valuable diagnostic tool for PJI, and its weight in PJI diagnostic criteria has gradually increased. Characterized by its convenience, speed and immediacy, leucocyte esterase strips test has a prospect of broad application in PJI diagnosis. Admittedly, the leucocyte esterase strips test has some limitations, such as imprecision and liability to interference. Thanks to the application of new technologies, such as machine reading, quantitative detection and artificial intelligence, the LE strips test is expected to overcome the limitations and improve its accuracy.

The Utility of Serum d-Dimer for the Diagnosis of Periprosthetic Joint Infection in Revision Total Hip and Knee Arthroplasty

BACKGROUND

There is scarce and contradicting evidence supporting the use of serum d-dimer for the diagnosis of periprosthetic joint infection in revision total hip (THA) and knee (TKA) arthroplasty. Therefore, the purpose of this study is to test the accuracy of serum d-dimer against the 2013 International Consensus Meeting (ICM) criteria.

METHODS

A retrospective review was performed on a consecutive series of 172 revision THA/TKA surgeries performed by 3 fellowship-trained surgeons at a single institution (August 2017 to May 2019) and that had d-dimer performed during their preoperative workup. Of this cohort, 111 (42 THAs/69 TKAs) cases had complete 2013 ICM criteria tests and were included in the final analysis. Septic and aseptic revisions were categorized per 2013 ICM criteria ("gold standard") and compared against serum d-dimer using an established threshold (850 ng/mL). Sensitivity, specificity, likelihood ratios, and positive/negative predictive values were determined. Independent t-tests, Fisher's exact tests, chi-squared tests, and receiver operating characteristic curve analysis were performed.

RESULTS

There was no statistically significant difference in baseline demographics between septic and aseptic cases per 2013 ICM criteria. When compared to ICM criteria, d-dimer demonstrated high sensitivity (95.9%) and negative predictive value (90.9%) but low specificity (32.3%), positive predictive value (52.8%), and overall, poor accuracy (61%) to diagnose periprosthetic joint infection. Positive likelihood ratio was 1.42 while negative likelihood ratio was 0.13. The area under the curve (AUC) was 0.742.

CONCLUSION

Serum d-dimer has poor accuracy to discriminate between septic and aseptic cases using a described threshold in the setting of revision THA and TKA.