Introduction 

• Periprosthetic Joint Infection (PJI) represents a severe clinical complication of arthroplasty, and a leading indication for surgical revision. 
• A comprehensive tissue sampling regime, collecting from distinct anatomical locations, is key to PJI diagnosis and organism detection —organisations such as EBJIS³ and MSIS⁴ recommend at least 3 samples be taken.
• It is unclear what role intraoperative prearthrotomy fluid samples play in PJI diagnosis.
• Significant heterogeneity exists in the literature regarding the relative performance of intraoperative fluid culture and tissue culture. 

 

 

Headline Findings / Conclusions

• Tissue culture uniquely detects more organisms than fluid culture
• Fluid and tissue cultures perform similarly well at diagnosing infection
• Swapping a single tissue sample for a fluid sample only works when very few samples are taken overall

Conclusion: Fluid sampling adds clinically meaningful value when used in addition to comprehensive tissue sampling, especially given the substantial impact of PJI 

 

 

Research Questions 

Question One: What is the incremental value of adding fluid sampling onto an existing comprehensive tissue sampling regime?

Question Two: Can a fluid sample substitute a tissue sample as part of a comprehensive sampling regime without compromising diagnostic yield?

 

 

Key Definitions

Microbiological Yield: The proportion of samples which identified a plausibly pathogenic organism.

Diagnostic Yield: The proportion of cases in which the microbiology results increased the EBJIS classification.

Fluid Sample: A sample of synovial fluid drawn intraoperatively prior to arthrotomy 

 

 

Methodology

• Retrospective observational study.
• Screened all consecutive revision hip and knee arthroplasties (DAIR & implant replacement) conducted at a single specialist orthopaedic centre in the UK between Nov 2024 & Nov 2025.
• 397 cases screened — 165 excluded (53 cancelled, 112 no paired fluid and tissue samples)
• Tissue samples collected according to surgeon discretion (number and location).
• Microbiology processed by a single on-site laboratory;  aerobic & anaerobic incubation up to 10 days.

 

Three levels of analysis: 

1. Diagnostic & Microbiological yield — overall concordance, single-modality detections, polymicrobial discordance. 
2. Diagnostic performance (sensitivity, specificity, PPV, and NPV) calculations. EBJIS classification binarised to enable calculation. Paired comparison using McNemar’s test (α = 0.05).
3. Resampling of cases with ≥4 tissue samples to enable comparison of tissue-only and tissue+fluid simulated sampling strategies at different overall sample counts. 5% non-inferiority margin used. 

 

 

Results

One: Diagnostic & Microbiological Yield

• 232 cases included: 68 “infection confirmed”, 6 “infection likely”, 154 “infection unlikely”.
• Among 74 “infection confirmed” & “infection likely”, 47 were monomicrobial (63.5%), 15 (20.3%) polymicrobial, 12 (16.2%) culture-negative. 
• Overall diagnostic concordance: 91.9%.
• Five monomicrobial cases were discordant; tissue uniquely identified the plausible pathogen in four cases, and fluid in 1 case.
• All uniquely grown organisms were CoNS.
• Five polymicrobial cases were discordant; tissue identified additional organisms in 4 cases, and fluid in 1 case (Table 2).

 

Two: Modality Sensitivity, Specificity, PPV, & NPV

• No statistically significant difference was found in diagnostic performance across all 4 performance metrics between both modalities (table 3).
• No difference when reference standard classified PJI-positive as “infection confirmed” + “infection likely” versus “infection confirmed” alone.
• Exclusion of microbiology from reference standard decreased PPV for both modalities (Fluid: -24.2%, Tissue: -21.7%), and decreased fluid specificity (-8.5%), no other statistically significant changes were found.

 

Three: Resampled Simulated Sampling Strategies

• Resampling showed that substituting one tissue sample with a fluid sample was only non inferior at an overall sample count of two (Figure 1A).
• At higher sample counts, tissue-only strategies consistently outperformed strategies substituting tissue for fluid .
• Addition of fluid sampling to tissue sampling regimes was found to be superior at low tissue sample counts (≤2), and non-inferior at higher tissue sample counts(Figure 1B).