Our goal is to assess diagnostic accuracy of synovial fluid testing in diagnosing prosthetic joint infection (PJI) as defined by the European Bone and Joint Infection Society (EBJIS). In addition to differential leukocyte count, simples and inexpensive biomarkers such as synovial fluid C-reactive protein (CRP), adenosine deaminase (ADA) and alpha-2-macrogloblulin(A2M) were also investigated and its possible role in increasing accuracy assessed.
Between January/2013 and December/2019 total hip or knee arthroplasty revision cases (regardless of preoperative diagnosis) were prospectively included provided enough synovial fluid for biomarker analysis was collected and at least four tissue samples, as well as the implant for sonication, were gathered for microbiological study. Definitive diagnosis was classified according to the new EBJIS PJI definition. Using receiver operating characteristic curves, we determined cutoff values as well as diagnostic accuracy for each marker.
Out of 364 revision arthroplasties performed, 102 fully respected inclusion criteria. There were 58 unlikely, 8 likely and 36 confirmed infections. Synovial fluid total leukocyte count, proportion of polymorphonuclear neutrophils (PMN), CRP, ADA and A2M were significantly different between groups. Area under the curve was 0.94 for total leucocyte count, 0.91 for proportion of PMN, 0.90 for CRP, 0.82 for ADA and 0.76 for A2M. Sensitivity, specificity, and predictive values for statistically optimal but also selected rule-in and rule-out cutoffs values are shown in Table 1. Interpreting a raised level of CRP(>2.7mg/L) or ADA(>60U/L) together with high leukocyte count (>1470 cells/μL) or proportion of PMN (>62.5%) significantly increases specificity and positive predictive value for affirming PJI.
Differential leukocyte count cutoffs proposed by the EBJIS PJI definition are shown to perform well in ruling out (<1,500 cells/μL) and ruling in (>3,000 cells/μL) PJI. Adding simple and inexpensive biomarkers such synovial CRP or ADA is helpful in interpreting inconclusive results.
For any tables or figures, please contact the authors directly.