This retrospective cohort study was conducted in patients with suspected PJI. From January 2018 to December 2023, 172 patients aged 18 years or older with suspected PJI after total joint arthroplasty and who underwent diagnostic laboratory testing were consecutively enrolled in the First Medical Center of Chinese PLA General Hospital. Exclusion criteria were cancer, liver cirrhosis, severe renal insufficiency and the absence of confirmed infection-related diagnoses or necessary data. The diagnosis of PJI was made following the Musculoskeletal Infection Society definition of PJI in 2014 (Parvizi and Gehrke, 2014) integrating all available data under the clinical scenario. This study was approved by the Ethics Committee of Chinese PLA General Hospital. We conducted the study following the principles of the Declaration of Helsinki and current ethics standards. All patients signed the written informed consent.

SF specimens from the included subjects were harvested via aseptic arthrocentesis, which were aliquoted and stored at -80 °C within 2 h of acquisition. Before storage, routine laboratory tests and pathogen cultures of SF were conducted. When conditions of “punctio sicca” were encountered, joint irrigations with sterile normal saline followed by aspirations of the lavage fluid were performed for microbiological analyses, and cytological studies were thus omitted due to dilution. At the same time, venous blood was collected into three tubes, containing 3.2 % sodium citrate, ethylene diamine tetra-acetic acid (EDTA) and clot activator (BD Biosciences, New Jersey, USA), respectively. Citrated whole blood was centrifuged at 1500 g at room temperature for 15 min to obtain the plasm. Clotted whole blood was centrifuged at 1500 g at room temperature for 15 min to obtain the serum.

WBC count, neutrophil percentage (NEUT %) and platelet (PLT) count in EDTA-anticoagulated whole blood, as well as SF-WBC and SF-PMN %, were determined using a Sysmex XN-20 analyzer (Sysmex, Kobe, Japan). The absolute counts of neutrophils (NEUT) were calculated by multiplying the WBC and NEUT %, and the SF polymorphonuclear cell count (SF-PMN) was calculated as the product of SF-WBC and SF-PMN %. The levels of ESR and CRP in EDTA-anticoagulated whole blood were measured using a Greiner Bio-One SRS 100/II analyzer (Greiner Bio-One, Frickenhausen, Germany) and a Lifotronic PA-990 analyzer (Lifotronic, Shenzhen, China), respectively. The D-dimer level in plasma was measured using a Stago STA R Max analyzer (Stago, Paris, France). The interleukin-6 (IL-6) level in serum was measured using an Immulite 1000 immunoassay system (Siemens, Forchheim, Germany). For pathogen culture, the SF specimens were injected into BACT/ALERT PF Pediatric FAN vials and BACT/ALERT PF FA FAN vials, respectively, and detected using a BioMérieux BACT/ALERT system (BioMérieux, Marcy l'Etoile, France). All specimens for microbiological tests were incubated for up to 14 d unless the positive growth of pathogens, and the results, were documented as “Culture”.

The concentrations of the three main constituents of NETs in SF, including cell-free dsDNA (SF-dsDNA), citrullinated histone H3 (SF-CitH3) and nucleosomes (SF-Nucleosome), were measured (Liu et al., 2021). The SF aliquots were diluted twice with phosphate buffer solution. The kits were read using a SpectraMax M2 microplate reader (Molecular Devices, California, USA).

The Quant-iT™ PicoGreen^® dsDNA Assay Kit (Invitrogen, California, USA) was applied to determine the concentration of SF-dsDNA. Two 96-well black plates (Corning, New York, USA) were utilized for fluorescence acquisition, with an excitation wavelength of 480 nm and an emission wavelength of 520 nm. The Citrullinate Histone H3 ELISA Kit (Cayman, Michigan, USA) was employed to measure SF-CitH3, with the absorption wavelength set at 450 nm. The SF-Nucleosome was measured using the Cell Death Detection ELISA^PLUS Kit (Roche, Basel, Switzerland), and the value of absorption at the wavelength of 405 nm subtracted from that of 490 nm was adopted. The absorbance value of the negative control well was subtracted from that of the test well and the positive control well, respectively. Subsequently, the ratio of the resulting value for SF-Nucleosome concentration was calculated.

An IBM SPSS 26.0.0 (New York, USA) and GraphPad Prism 8.0.2 (Massachusetts, USA) were employed for statistical analysis. All data had non-normal distributions; thus, they were presented as median (25th, 75th percentiles). A Mann–Whitney U test was used to identify variables with statistically significant differences between the PJI and non-PJI groups. Spearman's rank correlation coefficient ρ was used for nonparametric correlation analysis. Outliers twice the interquartile range and larger than the 75th percentile were excluded in the linear regression analysis. Binomial logistic regression analysis was used to generate odds ratios (ORs), with 95 % confidence intervals (CI). An ROC curve analysis was adopted to estimate the diagnostic efficacy of the respective indices. All box plots were presented with the horizontal bars representing medians and the vertical bars representing the 2.5th and the 97.5th percentiles. P < 0.050 was considered as statistically significant.

Of the total of 172 subjects, 19 fulfilled the exclusion criteria, and the rest were included in the study. Among the 153 subjects, 64 were diagnosed with PJI, all of which were classified as chronic PJI based on the available medical history and diagnostic findings. The demographic and clinical characteristics of the two groups are shown in Table 1. The median age of participants was 67.0 (60.0, 72.0) years, with a significant difference between the PJI and non-PJI groups (P=0.021). There were 41 hips and 112 knees involved in total, with no significant difference between groups (P=0.494).

The results for all parameters in the two groups are displayed in Table 2. We found that PJI patients had higher levels of SF-dsDNA and SF-CitH3 than non-PJI patients (P < 0.001 for both), but there is no significant difference in SF-Nucleosome (P=0.135).

Notably, the SF-WBC level in PJI group was almost 40 times higher than that in non-PJI group (P < 0.001). Compared to the patients without PJI, the levels of SF-PMN % and SF-PMN were higher in those with PJI (P < 0.001 for both). For Culture, the number of positive cases in the PJI group was significantly more than that of non-PJI group (P < 0.001); and the sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) were 54.10 %, 86.75 %, 75.00 % and 72.00 %, respectively. Among the 33 Culture-positive subjects in the PJI group, staphylococci were the predominant pathogens, with Staphylococcus aureus isolated in eight cases (24 %) and coagulase-negative staphylococci (CoNS) in nine cases (27 %; S. epidermidis, S. haemolyticus, S. hominis, S. warneri, S. caprae and S. lugdunensis). Other organisms included enterococci (four cases), Gram-negative bacilli (four cases), viridans group streptococci and S. agalactiae (two cases), Candida parapsilosis (two cases), Brucella spp. (two cases), and occasional Micrococcus and rapidly growing mycobacteria. Three polymicrobial cases were identified. Among the non-PJI cases, 11 had a single positive Culture result, including five CoNS, two Gram-negative bacilli, two Bacillus spp. and two Rhizopus spp.

There is no significant difference in the level of WBC between two groups (P=0.293). However, compared to non-PJI patients, the level of NEUT % was higher in PJI patients (P=0.002), and the level of NEUT tended to be higher in patients with PJI (P=0.053). Patients without PJI manifested lower levels of D-dimer than those with PJI (P=0.008). Compared to the non-PJI group, the levels of ESR and PLT were higher in patients with PJI (P < 0.001 and P=0.020, respectively). Finally, the levels of CRP and IL-6 demonstrated a significant increase in the PJI group compared to the non-PJI group (P < 0.001 for both).

A correlation analysis was conducted between NET constituents (SF-dsDNA, SF-Nucleosome and SF-CitH3) and SF-WBC, SF-PMN % and SF-PMN. As shown in Fig. 1, there were positive correlations between SF-dsDNA and SF-WBC (ρ=0.464, P < 0.001), SF-PMN % (ρ=0.455, P < 0.001) and SF-PMN (ρ=0.535, P < 0.001). Similarly, SF-CitH3 also showed significant positive correlations with SF-WBC (ρ=0.436, P < 0.001), SF-PMN % (ρ=0.446, P < 0.001) and SF-PMN (ρ=0.494, P < 0.001) (Fig. 1). SF-Nucleosome was not significantly correlated with local WBC-related indices (P>0.050).

ROC analyses were performed for all three components of SF-NETs, as well as for all indices that differed significantly between the PJI and non-PJI groups, including SF-WBC, SF-PMN %, SF-PMN, D-dimer, ESR, NEUT %, PLT, CRP and IL-6. The ROC curves of the different indicators are depicted in Fig. 2. As shown in Table 3, the AUC values of SF-dsDNA (0.844) and SF-CitH3 (0.821) were both greater than 0.800, and the AUC values of SF-PMN % (0.847) and SF-PMN (0.842) were similar to the AUC value of SF-dsDNA. The AUC values of SF-Nucleosome, SF-WBC, D-dimer, ESR, NEUT %, PLT, CRP and IL-6 were 0.571, 0.808, 0.637, 0.781, 0.649, 0.614, 0.798 and 0.721, respectively. The cutoff values of these indicators were calculated by the Youden index as 1.038 ng mL⁻¹ (SF-dsDNA), 2.050 (SF-Nucleosome), 1171.694 ng mL⁻¹ (SF-CitH3), 2095.000 × 10⁶ L⁻¹ (SF-WBC), 68.500 % (SF-PMN %), 947.800 × 10⁶ L⁻¹ (SF-PMN), 1.195 mg L⁻¹ FEU (D-dimer), 18.500 mm h⁻¹ (ESR), 66.900 % (NEUT %), 267.000 × 10⁹ L⁻¹ (PLT), 1.235 mg dL⁻¹ (CRP) and 8.915 pg mL⁻¹ (IL-6). The sensitivity, specificity, PPV and NPV of the identified factors were calculated using their cutoff values, and SF-CitH3 displayed the highest specificity (95.5 %) (Table 3).

To further analyze the diagnostic value of SF-NETs for PJI, different combinations were identified. For instance, according to the ROC curve analysis, SF-NETs₂₊ was categorized when the test values of two out of the three components were equal to or greater than their respective cutoff values. Similarly, SF-NETs₁₊ and SF-NETs₃₊ were also documented (Table 4). As shown in Table 4, SF-NETs₁₊ displayed the greatest sensitivity (82.8 %), and the sensitivity of SF-NETs₂₊ and SF-NETs₃₊ were 59.4 % and 15.6 %, respectively. Meanwhile, SF-NETs₃₊ exhibited the largest specificity of 100.0 %, and the specificity of SF-NETs₁₊ and SF-NETs₂₊ were 78.7 % and 95.5 %, respectively. Additionally, among the NETs₁₊ false-negative subjects, four cases yielded positive cultures, comprising three cases of CoNS and one case of enterococci.

A binomial logistic regression analysis was conducted to identify the risk factors of PJI. Variables that were independently and significantly different between the PJI and non-PJI groups – including age, SF-NETs₁₊, SF-PMN %, SF-WBC, Culture, D-dimer, ESR, NEUT %, WBC, PLT, CRP, and IL-6, were included into the analysis. Three indices, namely SF-NETs₁₊, SF-PMN % and Culture – were independent risk factors for PJI (Table 5). A larger number of SF-NETs₁₊ (P=0.001) and Culture (P=0.010) were associated with the increased risk of PJI. Furthermore, high levels of SF-PMN % (P=0.003) increased the risk of PJI.

Consequently, a formula was developed to calculate a new indicator named NET related index (NETRI): NETRI = 10.529 × SF-NETs1++28.114 × SF-PMN % + 8.210 × Culture - 1.452. In this formula, the positive results of SF-NETs₁₊ and Culture were marked as “1” and their negative results were marked as “0”. As shown in Fig. 3A, compared to patients without PJI, patients with PJI exhibited significantly higher levels of NETRI [0.483 (-0.751, 1.376) vs. 4.138 (3.571, 6.009), P < 0.001]. The ROC curve analysis was conducted for NETRI (Fig. 3B), yielding an AUC of 0.922 (95 % CI: 0.862–0.981) with a cutoff value of 3.346. The sensitivity, specificity, PPV and NPV of NETRI were 82.2 %, 93.6 %, 92.5 % and 84.6 %, respectively.