This prospective study included consecutive patients undergoing knee joint aspiration for suspected PJI of a knee arthroplasty or suspected septic arthritis of the native knee joint. Both infected and non-infected native joints and knee arthroplasties were included to ensure a broad range of analyte concentrations and synovial fluid viscosities. All joint aspirations were performed under sterile conditions in the outpatient department. The knee joint was selected as the study focus because sufficient synovial fluid volume was required for multiple analyses, and knee effusions typically provide adequate sample volumes. Patients from whom less than 5 mL of synovial fluid could be obtained were excluded (n= 10). A total of 35 patients were included in the study, comprising 10 native knee joints and 25 knee arthroplasties.

Ethical approval was waived by the Ethics Committee of the Technical University of Munich (reference number 714/20S, 20 November 2020). Written informed consent was obtained from all participants prior to inclusion. This study was conducted and reported in accordance with the STARD 2015 guidelines for diagnostic accuracy studies.

All synovial fluid samples were processed immediately after aspiration. Aspirated synovial fluid was transferred to Blood Gas Monovette^® and S-Monovette^® Fluoride/EDTA FE (both Sarstedt, Nümbrecht, Germany) collection tubes for BGA analysis at the POC and analysis of lactate or glucose in the central laboratory, respectively. Native synovial fluid was used for pH measurements in the central laboratory. In the laboratory setting, synovial pH was measured using a calibrated SevenEasy^™ pH analyzer (Mettler-Toledo, Columbus, OH, USA) at room temperature (23 °C), while synovial lactate and synovial glucose concentrations were determined using the central laboratories Cobas^® c702 clinical chemistry analyzer (Roche Diagnostics, Basel, Switzerland). POC BGA measurements were performed using a RAPIDPoint^® 500e analyzer (Siemens Healthineers, Erlangen, Germany). Each prepared sample was measured three times, avoiding relevant time delays between the applied methods. In addition, one synovial pH measurement per sample was performed, using the BGA set to pleural fluid mode at room temperature (23 °C). The workflow of synovial fluid analysis is illustrated in Fig. 1.

For statistical analyses, the mean value derived from three repeated measurements per sample was used, which reduced random analytical variability. Agreement between BGA and central laboratory measurements for synovial pH, lactate, and glucose was assessed using Passing–Bablok and Bland–Altman analysis. Results outside the analytical range (Table 3) were excluded from statistical agreement analyses, as these values could not be reliably quantified. The mean difference (bias) and 95 % limits of agreement (mean ± 1.96 standard deviations) were calculated and graphically displayed. In addition, linear Deming regression analysis was performed to evaluate the relationship between BGA and central laboratory results, and the 95 % confidence interval (CI) for regression parameters (i.e., slope and intercept) as well as results from the cumulative sum (CUSUM) test of residuals were reported. The assumption of normally distributed residuals was evaluated through graphical inspection of the distribution of differences between methods, including histograms and normal P–P plots of regression-standardized residuals, as well as by applying the Shapiro–Wilk test. Homoscedasticity of residuals was assessed by testing for heteroscedasticity using the Breusch–Pagan test and by visually examining Bland–Altman plots for potential trends in variability of differences across the measurement range. In order to complete the description of the association strength between both methods, Pearson's correlation coefficients (r) were calculated.

Precision was evaluated from triplicate measurements for each sample and method. For synovial lactate and glucose, precision was expressed as coefficients of variation (CVs), whereas precision of synovial pH was expressed as standard deviation (SD) in pH units.

All analyses were performed using SPSS Statistics (version 29; IBM Corp., Armonk, NY, USA) and OriginPro (version 2025; OriginLab Corporation, Northampton, MA, USA). CUSUM testing was performed with the R “strucchange” package. A two-sided p value ≤ 0.05 was considered statistically significant where applicable.

In total, synovial fluid samples from 35 patients were collected and analyzed. Four samples yielded synovial pH values outside the quantifiable range of the BGA (Table 3), and an additional three samples did not provide reliable pH measurements using the central laboratory pH meter due to insufficient remaining sample volumes. For synovial lactate, one sample exceeded the analytical range of the Cobas^® c702 analyzer at the central laboratory and was not further diluted for analysis, while another sample could not be analyzed with the Cobas^® c702 analyzer because of excessive viscosity, despite hyaluronidase treatment. Furthermore, five and four samples were below the lower limit of quantification for glucose using the BGA and central laboratory device, respectively. In all cases where BGA analyses failed to yield definitive results due to values falling outside the device's measurement ranges, the findings were consistent with those obtained from central laboratory measurements. Consequently, in-depth comparative analyses between BGA and central laboratory measurements were performed on 29 samples for pH, on 33 for lactate, and on 30 for glucose, as detailed in the following sections. All analyses indicated normally distributed residuals. Furthermore, the results of the Breusch–Pagan tests, supported by a visual inspection of Bland–Altman plots, revealed no evidence of systematic heteroscedasticity across the measurement range for the analyzed biomarkers.

For synovial pH measurements in blood mode, Passing–Bablok and Deming regression analyses demonstrated a linear relationship between BGA and laboratory measurements. Pearson's correlation coefficient indicated a statistically significant association (r= 0.808; p<0.001). The relationship was described by the following Deming regression equation: BGA pH = 0.712 × laboratory pH + 2.015 (Fig. 2a, Table 1).

The CUSUM test showed no significant deviation from linearity (S = 0.796, p = 0.142). Bland–Altman analysis indicated good agreement, with a mean bias of -0.10 pH units and 95 % limits of agreement ranging from -0.51 to +0.31 pH units (Fig. 2b, Table 1).

With the BGA set to pleural fluid mode at 23 °C, Bland–Altman analysis indicated better agreement between BGA and laboratory pH measurements, yielding a mean bias of -0.04 pH units with 95 % limits of agreement ranging from -0.37 to +0.28 pH units. However, due to the narrower analytical range of the pleural fluid mode compared with the conventional BGA pH mode, only 22 (62 %) result pairs were eligible for statistical analysis.

For synovial lactate measurements, Passing–Bablok and Deming regression analyses demonstrated a linear relationship between BGA and central laboratory results. Pearson's correlation coefficient indicated a statistically significant association (r= 0.989; p<0.001). The relationship was described by the following Deming regression equation: BGA lactate = 0.865 × laboratory lactate + 0.809 (Fig. 3a, Table 1).

Linearity between methods was confirmed by the CUSUM test (S = 0.664, p = 0.302). Bland–Altman analysis indicated good overall agreement, with a mean bias of -0.04 mmol L⁻¹ and 95 % limits of agreement ranging from -1.77 to +1.69 mmol L⁻¹ (Fig. 3b, Table 1), with some dispersion observed at lower and higher concentration ranges.

For synovial glucose measurements, Passing–Bablok and Deming regression analyses demonstrated a linear relationship between BGA and central laboratory results. Pearson's correlation coefficient indicated the strongest statistically significant association among the three biomarkers (r= 0.997; p<0.001). The relationship was described by the following Deming regression equation: BGA glucose = 1.092 × laboratory glucose - 6.761 (Fig. 4a, Table 1).

A mild proportional bias was observed at higher glucose concentrations, without relevant clinical impact. Linearity between methods was confirmed by the CUSUM test (S = 0.786, p = 0.151). Bland–Altman analysis indicated excellent agreement, with a mean bias of +0.58 mg dL⁻¹ and 95 % limits of agreement ranging from -7.11 to +8.27 mg dL⁻¹ (Fig. 4b, Table 1).

In order to assess the precision of BGA analyses, within-run variability derived from triplicate measurements was evaluated. Precision analysis based on triplicate measurements demonstrated slightly lower variability in BGA measurements compared with central laboratory measurements for synovial pH. For this analyte, precision was expressed as the standard deviation (SD) in pH units, reflecting the logarithmic scale of pH. BGA measurements showed a mean SD of 0.024 ± 0.021 pH units, compared with 0.030 ± 0.028 pH units for central laboratory measurements. In contrast, central laboratory measurements exhibited lower variability for synovial lactate and glucose when expressed as coefficients of variation (CVs) (Table 2).

Previous studies have demonstrated favorable diagnostic performance of synovial pH in the diagnosis of PJI, with reported sensitivities ranging from 50 % to 89 % and specificities between 81 % and 96 % when cut-offs between 7.11 and 7.40 were used (Theil et al., 2022; Judl et al., 2024; Müller et al., 2025).

Among the parameters analyzed in this study, synovial pH proved to be the most challenging to measure. It showed the greatest variability between different measurement methods. pH is a highly sensitive parameter that is influenced by multiple physiological and pre-analytical factors, particularly temperature fluctuations (Cheng et al., 1998; Ashwood et al., 1983; Mishra and Rahman, 2009). In addition, delays between sample collection and analysis can lead to artificially increased pH values due to a reduction in pCO₂ caused by carbon dioxide diffusion into ambient air (Cheng et al., 1998; Venkatesh et al., 1999; Mishra and Rahman, 2009). For this reason, as with conventional blood gas analysis with blood samples, synovial fluid blood gas analysis should ideally be performed immediately after joint aspiration whenever feasible.

The determination of pleural fluid pH provides a useful analogy. Pleural pH is an established diagnostic parameter in the evaluation of pleural effusions, with pH values below 7.20 indicating pleural empyema requiring tube thoracostomy (Houston, 1981; Good et al., 1980). Historically, pleural fluid pH has been measured using indicator strips, pH meters, or BGAs. Comparative studies have consistently demonstrated the superiority of BGAs over pH meters, showing that only BGA-derived pH values achieve sufficient accuracy for clinical decision-making (Cheng et al., 1998; Venkatesh et al., 1999; Mishra and Rahman, 2009; Lesho and Roth, 1997). The superiority of BGA pH measurements over pH meter analyses was also observed in this study, by showing less variability in repeated measurements and more reliable results.

Despite this evidence, approximately 30 %–50 % of laboratories in the United States reportedly continue to use inaccurate methods for pleural fluid pH measurement (Putnam et al., 2013; Bowling et al., 2012). This observation underscores the importance of standardized measurement techniques and supports the use of BGAs for synovial pH assessment in the diagnostic work-up of PJI and SANJO.

Elevated synovial lactate concentrations are indicative of joint infection. Previous studies investigating PJI have reported sensitivities ranging from 50 % to 73 % and specificities between 67 % and 88 %, using cut-off values between 5.30 and 8.45 mmol L⁻¹ (Sharma et al., 2020; Lenski and Scherer, 2014a, 2015; Müller et al., 2025). In addition, the 2023 EBJIS guideline on SANJO recommends the analysis of synovial lactate as part of the diagnostic work-up (Ravn et al., 2023).

In the present study, synovial lactate concentrations measured using the BGA showed good agreement with central laboratory measurements, with negligible mean bias.

Comparable findings have been reported in other clinical contexts. A recent study evaluating cerebrospinal fluid lactate determination for the diagnosis of acute meningitis found a small and systematic difference between results obtained with a RAPIDPoint 500 BGA and conventional laboratory assays (Fernández Reina et al., 2024). Despite this bias, the authors concluded that BGA-based lactate quantification remains a useful diagnostic tool, supporting the applicability of this approach for synovial fluid analysis.

Low synovial glucose concentrations are associated with septic arthritis in native joints (Söderquist, 1998) as well as PJI (Lenski and Scherer, 2014a, 2015; De Vecchi et al., 2016; Haertlé et al., 2022; Sabater-Martos et al., 2025). The 2023 EBJIS guideline on SANJO suggests the analysis of synovial glucose as an additional adjunct investigation (Ravn et al., 2023).

Because synovial glucose concentrations may be influenced by fluctuations in blood glucose levels (Brennan and Hsu, 2012), Sabater-Martos et al. proposed the serum : synovial glucose ratio as a diagnostic parameter for acute PJI. Using a cut-off value of > 0.69, they reported a sensitivity of 72 % and a specificity of 94 % (Sabater-Martos et al., 2025).

In our comparative analysis, synovial glucose concentrations obtained by using BGAs showed excellent agreement with central-laboratory-based glucose measurements, revealing only minimal variability between methods.

Similar observations have been reported for pleural fluid glucose measurements. Based on the robustness and reproducibility of glucose analysis, Mishra et al. (2009) suggested that pleural glucose levels may be used to support clinical decision-making when accurate pleural pH measurements are not available (Mishra and Rahman, 2009). These findings further support the reliability of synovial glucose results obtained as an adjunct diagnostic parameter using BGAs.

Some BGAs, such as the Siemens RAPIDPoint 500e, offer a dedicated mode for pleural fluid measurement. In our study, measurements performed in pleural fluid mode, calibrated to 23 °C, showed closer agreement with pH meter values. However, the pleural fluid mode is limited by a narrower analytical range of pH 7.000–7.500 (Table 3), allowing statistical analysis of only a limited set of samples.

This limitation is of minor clinical relevance for the diagnosis of PJI, as synovial fluid pH values below 7.000 are strongly associated with infection (Müller et al., 2025; Theil et al., 2022; Judl et al., 2024). Consequently, values below the quantification range are, in practice, already highly suggestive of PJI.

Similar considerations apply to the measurement ranges of synovial lactate and glucose. Extremely low synovial glucose concentrations (< 20 mg dL⁻¹) or markedly elevated synovial lactate levels (> 30 mmol L⁻¹) are strongly associated with infection (Müller et al., 2025). Therefore, even when values approach or exceed the upper or lower quantification limits of the BGA, their diagnostic interpretation remains clinically meaningful.

Synovial fluid exhibits a substantially higher viscosity than blood (Roškar et al., 2025; Fu et al., 2019), raising concerns that its application to BGAs could lead to clogging or damage. In the present study, encompassing 140 synovial fluid measurements, including those with highly viscous samples, no clotting events or BGA malfunctions were observed. Furthermore, consultation with one of the largest German manufacturers of BGAs confirmed that damage to the analyzer itself is highly unlikely. In a worst-case scenario, only the disposable measurement cartridge may be affected.

Comparable concerns were previously raised regarding pleural fluid pH analysis (Cheng et al., 1998); however, pleural fluid measurements using BGAs are now well established and manufacturer approved. Similarly, the development of a dedicated synovial fluid measurement mode by BGA manufacturers could further support standardized and optimized clinical applications.

An important advantage of BGA analysis over conventional laboratory testing is the minimal sample volume required, the feasibility of POC testing, and the rapid turnaround time. While laboratory-based analyses, covering pH electrode measurements and clinical chemical analyses, typically require several milliliters of synovial fluid, BGAs generally require approximately 100 µL. Moreover, due to the high viscosity of most synovial fluid samples, at least in central laboratories equipped with Cobas^® Pro c702 analyzers, pre-analytical hyaluronidase treatment of samples is necessary to avoid clogging of the analytical system. Besides significantly prolonging analysis times, this can falsify particularly synovial glucose and lactate concentrations in infected samples and explain some result variations between BGA and central laboratory measurements in the present study. Positioning BGAs at POC eliminates the need for sample transport and processing, thereby accelerating clinical decision-making.

In addition to pH, lactate, and glucose, BGAs provide multiple further diagnostic parameters, including sodium, potassium, partial pressure of carbon dioxide (pCO₂), ionized calcium, total hemoglobin, partial pressure of oxygen (pO₂), chloride, and CO-oximetry parameters. Notably, pCO₂ may be of particular interest, as it is already used in the diagnostic evaluation of pleural effusions (Light et al., 1973). Further studies are needed to determine the diagnostic relevance of these parameters in the diagnosis of PJI.

The present study has several limitations that are acceptable in the context of this feasibility study. First, only a single BGA model was evaluated; therefore, the findings may not be generalizable to BGAs from other manufacturers, which may yield different results. Although no device damage was observed during the study period, the long-term effects of synovial fluid analysis on BGAs remain unknown, and off-label use may void manufacturer warranties. In addition, BGAs have a limited analytical measurement range (Table 3). Although values exceeding this range are already strongly suggestive of joint infection, excluding measurements outside the analytical range may have introduced bias into the agreement estimates and represents a limitation of the analysis. Finally, agreement analyses were based on the mean of triplicate measurements per sample. While this approach reduces random analytical variability, it may slightly narrow the calculated limits of agreement compared with single measurements typically obtained in routine clinical practice. Treating triplicate measurements as independent observations in the agreement analyses yielded no substantial differences compared with analysis of triplicate means (data not shown), thereby supporting the robustness of our findings and their applicability to clinical routine. Triplicate measurements were also used to assess the precision of the BGA and central laboratory results with synovial fluid samples. Here, it needs to be noted that only within-run or intra-day variabilities were determined, but no inter-run or inter-day variabilities. The limitations outlined, which are beyond the scope of this feasibility study, should be addressed in future performance evaluation studies.

In conclusion, measurement of synovial pH, lactate, and glucose using BGAs demonstrates good analytical agreement with central laboratory methods and represents a reliable, rapid, and cost-effective approach for the assessment of these biomarkers that may support clinical decision-making in suspected septic arthritis and periprosthetic joint infection. These POC analyzers have the potential to facilitate efficient decision-making in routine orthopedic practice. Further studies are warranted to evaluate the performance of BGAs from different manufacturers and to establish brand-specific reference ranges. The development of a dedicated synovial fluid mode by BGA manufacturers would be highly desirable to facilitate standardized and optimized clinical use.