(Pre)analytical considerations concerning the analysis of synovial calprotectin

Samples

In the University Medical Center Groningen (UMCG), synovial fluid samples were aspirated by using a 21 gauge, 1 in. needle that is attached to a 20 mL syringe, then transported to the laboratory in less than 30 min. The samples were used for routine calprotectin measurements on the same day. The residual synovial samples were stored at −80 °C in the biobank and used for this study. Samples were selected based on calprotectin concentration, synovial fluid volume, and, for the viscosity experiments, viscous aspect.

For the blood contamination and lipid contamination experiments, residual whole blood samples and lipaemic blood samples were used. For the blood contamination experiments EDTA anticoagulated tubes manufactured by Becton Dickonson (BD, Vianen, The Netherlands) are used. Highly lipaemic blood samples were left to stand for 3 h at room temperature and the lipaemic layer, merely containing chylomicrons, was subsequently harvested. In addition, anonymized residual subcutaneous fat tissue obtained after bariatric surgery was homogenized and stored at −20 °C prior to the lipid contamination experiment to release adipocyte intracellular content. The Medical Ethics Review Board (University Medical Centre Groningen (UMCG)) waived the need for ethical approval (METc 2021/168 and M22.302539).

Synovial calprotectin measurements

The levels of calprotectin were determined using the Lyfstone rapid LF test (Lyfstone, Tromsø, Norway), according to the manufacturer’s instructions. Within run and between run coefficients of variation (CV) of the assay are 20 and 25 %, respectively. To determine the CV’s freshly frozen and aliquoted samples were analyzed according to the CSLI-EP5-A protocol (Simple Precision).

Stability experiment (effect of time, temperature, EDTA and freeze-thaw cycles)

Synovial fluid samples of different synovial calprotectin levels; low (20–40 mg/L), medium (40–100 mg/L) and high (100–200 mg/L), were exposed to different storage conditions.

To investigate the effect of ethylene diamine tetra acetic acid (EDTA), time and temperature on sample stability, native fresh samples were split and either aspirated into an EDTA-containing vacutainer tube or a neutral tube and were either stored at room temperature (RT) or at 4 °C. For samples kept at RT, they were measured at six different time points; directly (t=0), after 3 h, 6 h, 24 h, three days and seven days. For samples kept at 4 °C, they were measured at five different time points; directly (t=0), 6 h, 24 h, three days and seven days. Each level and condition (with or without EDTA, temperature) contained three samples, hence a total of 396 samples were analyzed for the EDTA and temperature experiment.

To investigate the effect of freeze-thaw cycles on sample stability samples were exposed to 1, 2, 3 or 5 freeze-thaw cycles. For this experiment the same levels were used, each level and condition (with or without EDTA) contained three samples, hence a total of 72 samples were analyzed.

To compare the results between conditions and time points independent of the three different calprotectin levels, synovial calprotectin levels were expressed as percentage recovery based on the synovial calprotectin level at t=0.

Effect of blood and lipid contamination

To investigate the extent of blood and lipid contamination of routinely derived clinical synovial samples, our routine clinical database/biobank, containing 83 synovial samples and patient data from 74 individual patients collected since July 2015, was surveyed for the haemolytic index (H-index) and the lipemic index (L-index).

The effect of blood contamination on synovial calprotectin measurement was investigated at three different synovial calprotectin levels as described above, by the addition of A) hemolyzed whole blood and B) non-hemolyzed whole blood. Hemolyzed blood was obtained by three freeze-thaw cycles of an EDTA whole blood sample. The effect of lipid contamination on synovial calprotectin measurement was investigated also at the three different synovial calprotectin levels, by the addition of A) homogenized adipose tissue, B) intralipid (Fresenius Kabi Netherlands B.V., Zeist, the Netherlands) C) chylomicrons.

Synovial samples were separately mixed with either hemolyzed whole blood, non-hemolyzed whole blood, adipocyte homogenate, intralipid or chylomicrons in different ratios, which resulted in the following percentages of contamination: 100 , 50, 25, 12.5, 6.25, 3.12, 1.56, 0.76, 0.38 and 0 % blood or lipid. Each blood contamination condition was analyzed in duplo, whereas the adipocyte contamination experiment was performed in triplicate. In total, for the blood contamination experiment 40 samples and for lipid contamination experiment 108 samples were analyzed. In these experiments synovial samples were used with a mean calprotectin concentration of 87 or 103 mg/L (the adipocyte contamination experiment). This lower value was chosen to investigate the effect of contamination on samples with calprotectin values around the clinical cutoff value of 50 mg/L [3, 8]. The H and L indices were subsequently measured on a Cobas 6000 chemistry analyzer (Roche Diagnostics, Rotkreuz, Switzerland). To compare the results between conditions independent of the three different calprotectin levels, synovial calprotectin levels were expressed as percentage recovery based on the synovial calprotectin level in the sample without blood or lipid contamination.

Effect of synovial viscosity

To examine the effect of viscosity reduction on synovial calprotectin analysis, three pretreatment modalities to liquify vicious samples were investigated: A) five subsequent freeze-thaw cycles, B) enzymatic pretreatment and C) sonification. For these approaches, fresh samples were included that were sufficiently viscous by visual inspection and semi-quantitatively scored (1–5) by two different technicians, independently. Viscosity was scored and calprotectin levels were measured before and after each pretreatment. Synovial calprotectin levels were expressed as percentage recovery based on the synovial calprotectin level in the sample without pretreatment.

Approach A consisted of five subsequent freeze-thaw cycles. For this experiment, the same samples were used as described in the stability experiments.

Approach B consisted of a) no treatment, b) hyaluronidase (Sigma Aldrich, Amsterdam, The Netherlands) treatment, c) DNAse (Sigma Aldrich, Amsterdam, The Netherlands) treatment or d) combined hyaluronidase and DNAse treatment.

Three different levels (68, 85 and 136 mg/L) of calprotectin concentrations were used. For each level, analysis was performed in triplicate. Each synovial fluid sample was further divided and transferred into four different tubes (one for each pretreatment modality). A hyaluronidase solution of 146 U/mL was prepared by dissolving 10 mg of bovine testicular hyaluronidase solution (439 U/mg) in 30 mL of saline buffer (0.9 g/L). A DNAse solution of 150 U/mL was prepared by dissolving one DNAse-containing tablet (10,000 U/tablet) in 66.6 mL of distilled water.

The untreated synovial sample was mixed in a 2:1 (sample:enzyme) ratio with the hyaluronidase (150 U/mL) solution and/or the DNAse solution. After incubation at 37 °C for 30 min with hyaluronidase, DNAse or both, synovial calprotectin was measured.

In total 36 samples were analyzed. Approach C consisted of sonification of the samples, for 10 s using a Vibra cell, model VC60 sonificator (Sonics & Materials Inc, Newtown, USA). Twenty-five subsequent fresh routine patient samples were used. Before and after sonification, viscosity was scored as described above, and synovial calprotectin was analyzed.

Statistics and data analysis

The mean and confidence interval of the H-index and L-index were calculated. Repeated measurements of calprotectin levels over time were analyzed using mixed model analysis, where group (i.e., preanalytical condition) and measurement (calprotectin recovery) were both added as fixed variables. The interaction term group × measurement was added to analyze differences between groups over time. A Wilcoxon matched-pairs signed rank test was used to compare results before and after sonification. For all comparisons, a p-value<0.05 was regarded as significant. Data were analyzed using the software program GraphPad Prism (GraphPad Software, LLC, version 8.4.2, Boston, USA) and SPSS (version 29.0; SPSS, Chicago, USA).

Stability experiment (effect of time, temperature, EDTA and freeze-thaw cycles)

No significant decrease or increase of measured calprotectin was observed in synovial samples kept at room temperature compared to samples kept at 4 °C for up to seven days of storage. After seven days of storage, the addition of EDTA, in samples kept at 4 °C resulted in a slight, non-significant reduction of values (p-value=0.07) (Figure 1).

Figure 1:

Percentage recovery of synovial calprotectin (y-axis) in relation to storage duration (x-axis), effect of EDTA and temperature. Duration is depicted in hours (h). Results expressed as mean and standard deviations (SD, errors bars). (A) Effect of EDTA on percentage calprotectin recovery at room temperature (RT). (B) Effect of EDTA on percentage calprotectin recovery at 4 Celsius (4 °C). (C) Effect of the two storage temperatures on percentage calprotectin recovery in EDTA free synovial fluid samples. (D) Effect of the two storage temperatures on percentage calprotectin recovery of calprotectin in synovial fluid samples with EDTA.

Five successive cycles of freeze-thawing did not influence the levels of measured calprotectin in EDTA-containing samples. In samples without EDTA, calprotectin levels increased significantly after the first and second freeze-thaw cycles (p-value=0.011 and 0.003, respectively), this difference obliviated after three or more freeze-thaw cycles (Figure 2).

Figure 2:

The percentage recovery of synovial calprotectin (y-axis) in relation to the number of freeze-thaw cycles (x-axis), in synovial samples with (dark gray) and without (light gray) EDTA.

Effect of blood and lipid contamination

In the routine clinical data and biobank, the median H-index was 34 with a range of 0–6,834, while the median L-index was 40 with a range of 4–900 (Figure 3).

Figure 3:

Frequency analysis of (A) H-index and (B) L-index in 83 synovial fluid samples.

Synovial calprotectin LF assay appears to report lower calprotectin values in synovial samples contaminated with hemolyzed blood. However, no statistical relation could be found between the level of H index, contamination percentage and the recovered values of calprotectin (Figure 4A).

Figure 4:

Percentage recovery of synovial calprotectin (y-axis) in relation to the H-index and the percentage of contamination (x-axis) (A) hemolyzed blood and (B) non hemolyzed blood.

The use of non-hemolyzed whole blood in these experiments resulted in higher H indices than the use of hemolyzed blood, 8,180 and 980, respectively. Synovial calprotectin LF assay appears to report lower values in the presence of whole blood, yet this difference was not significant. No significant correlation could be found between the level of H-index and the recovered values of calprotectin (Figure 4B).

Adipose tissue or chylomicrons contaminated synovial fluid resulted in higher L indices than in the intralipid experiment, a maximum of 7,260, 6,690 and 4,330, respectively. In the lipid contamination experiment, the synovial calprotectin LF assay shows no significant increase or decrease in calprotectin levels. Furthermore, there is no clear or significant relation between the level of adipocyte, intralipid or chylomicrons contamination and the recovered values of calprotectin (Figure 5A–C, respectively).

Figure 5:

Percentage recovery of synovial calprotectin (y-axis) in relation to the L-index and the percentage of contamination (x-axis) with (A) homogenized adipose tissue (B) intralipid (C) chylomicrons.

Effect of synovial viscosity

After five subsequent freeze-thaw cycles the viscosity of the samples did not decrease, as judged visually using the 5-point scoring system (data not shown).

After hyaluronidase/DNAse treatment, the viscosity of the samples reduced from a score of 4–1 (Figure 6). DNAse alone has a smaller effect in reducing the viscosity of the synovial fluid as measured using the viscosity score. Calprotectin levels did not differ significantly between the different pretreatment modalities.

Figure 6:

Calprotectin value in mg/L and viscosity score of synovial samples in a control group and three enzymatically pretreated groups. The results are depicted in three different calprotectin levels; low, medium and high (x-axis).

Twenty-five subsequent fresh synovial samples were analyzed before and after sonification. Sonification had a significant effect on sample viscosity; mean viscosity scores decreased from 3.2 to 1. In addition, median calprotectin concentration increased significantly, from 59 mg/L before sonification to 100.5 mg/L after sonification (p<0.01) (Figure 7).

Figure 7:

Calprotectin value in mg/L (y-axis) and viscosity degree of synovial samples before and after sonification (x-axis).

Limitations and considerations

Apart from the limitation of some of the experiments being limited to a small sample size which could affect the generalization of the results, this study is the first of its kind to investigate and demonstrate an extensive overview of the potential interfering factors that could affect the analysis synovial calprotectin.

Although most of the interfering factors did not have a statistically significant effect on the calprotectin levels in synovial fluid, quite frequently, recoveries of 80 and 120 % were observed in our results. Together with analytical variation, we propose caution when interpreting synovial calprotectin values around the cut-off value without taking into account other clinical parameters.

Since sonification results in higher calprotectin values than untreated samples, probably due to the release of calprotectin from NETs and/or neutrophils present in the sample, one may argue whether this improves the sensitivity of the test. The presence of these intact neutrophils is overlooked when no or other sample pretreatments, such as DNAse/hyaluronidase pretreatment, is used. The presence of these neutrophils, beside the free calprotectin levels in the synovial fluid, indicate that an active inflammatory process is ongoing. When choosing this sample pretreatment as the standard pretreatment, new clinical cut-off values should however, be established.