Verification of the T50 Calciprotein Crystallization test: bias estimation and interferences

Patient samples

Approval from the local Ethical Research Committee was obtained to collect random and anonymized leftover serum samples from routine diagnostics at the Amsterdam University Medical Centre, the Netherlands. Left-over serum samples from a broad range of patients within Amsterdam UMC were stored at 4 °C for three days before collection from the routine laboratory. Blinded serum samples were either pooled or used as single samples depending on the analysis, and stored at 4 °C before measurement of the T50 test.

T50 test setup and measurement

The T50 test setup at the laboratory site of the manufacturer (Calciscon AG) was used as the benchmark (laboratory site 1). Based on the methodology developed by the manufacturer (Figure 1), the T50 test was set up at the laboratory of the Amsterdam UMC (laboratory site 2). The T50 values were measured using the T50 testing kit provided by Calciscon AG. Measurements were performed according to the manufacturer’s instructions. The test protocol has been described in detail by Pasch, et al. [12] The T50 test setup consisted of an automated pipetting machine (Freedom EVOware LiHa, TECAN, Männedorf, Switzerland), an orbital shaker (BioShake 3,000 elm, QINSTRUMENTS, Jena, Germany), and a nephelometer (NEPHELOstar plus, BMG Labtech, Ortenberg, Germany). The T50 value was determined with the T50 Analysis Software using a four-parameter logistic regression according to the manufacturer’s instructions.

Figure 1:

Testing principle for the T50 Calciprotein Crystallization test. Samples and reagents are prepared before pipetting them using an automatic liquid handling machine. Afterwards, the light scattering of the samples is measured. This measuring step has a duration of at least 600 min. Once the data is collected, it will be analyzed to determine the T50. Created with BioRender.com.

Laboratory site comparison (agreement)

The T50 test setup at the clinical laboratory of the Amsterdam UMC was compared to that of the manufacturer to estimate the bias between both sites. The T50 measurements were compared using 21 samples which were split into two vials – one vial for each laboratory site. The samples were shipped to Calciscon AG with ice packs. Measurements at both sites were conducted on the same day in triplicate by different operators. All samples were processed according to the manufacturer’s instructions.

Precision analysis

The precision of the T50 test was determined at the Amsterdam UMC according to the CLSI EP15-A3 protocol. The protocol specifies the use of two samples at different levels which are measured over five days, measuring once or twice a day based on the test duration. Due to the 10-h duration of the T50 test, the measurement was only performed once per day.

In this study, a third sample was measured in addition to the two samples instructed by the CLSI protocol. Pooled serum samples were used with three different ranges: low T50 (<270 min), mid T50 (∼350 min), and high T50 (>470 min). These ranges were determined based on the reference range for the T50 test (270–470 min). To obtain sufficient material for multiple testing days, the T50 values from individual serum samples were measured and pooled accordingly to form pooled samples at the three ranges. Pooled serum samples were aliquoted (500 µL), stored at 4 °C, and used within 7 days.

The three pooled serum samples were measured in four-fold over five consecutive days, alternating between two operators. The first day measurement was immediately on the same day when the serum samples were pooled. The coefficient of variation (CV%) and standard deviation (SD) were determined for each pooled sample for each day, and tested for the 95 % verification value based on the precision claims by the manufacturer for samples with similar T50 ranges (Table 1). The manufacturer provided their validated precision for within-run (S_within) and between-run (S_between) measurements. The S_wihtin and total precision (S_total) were used for verification of precision. The reason for using the total error is based on the assumption within the CLSI protocol in which it states that if the bias and estimated precision are acceptable, then the total error will also be acceptable [25]. The S_between was converted to S_total using the following formula: S t o t a l = S w i t h i n 2 + S b e t w e e n 2 . The test passes when the S_wihtin and S_total do not exceed the verification values.

Interference analysis

Possible interferences with the T50 test were studied at the Amsterdam UMC by spiking pooled serum samples with serial dilutions of bilirubin, free hemoglobin, triglycerides, and LMWH. The T50 value was then determined for each spiked sample. The interference was studied using three different serum pools. For the T50 test, a deviation of ±10 % is accepted, similar to the manufacturer’s report. Therefore, interference was established for each analyte when the measurement deviated by>10 % from the control sample.

Bilirubin solution was prepared by dissolving bilirubin, 98+% (Thermo Fisher Scientific, Waltham, MA, USA) in 0.1 M NaOH. The solution was added to the serum pool to prepare a highly icteric serum sample with a concentration of approximately 1,000 μmol/L which was further diluted to obtain other concentrations, with the lowest concentration being approximately 20 μmol/L. In total, six bilirubin concentrations were analyzed. The icterus index and total bilirubin concentration were measured using the Roche Cobas Pro (Roche Diagnostics, Basel, Switzerland).

The effect of hemolysis on the T50 measurement was determined by spiking the serum sample with hemolysate and serially diluting (1:1) the sample to obtain eight samples ranging from approximately 0.005 mmol/L to 0.5 mmol/L hemoglobin. Hemolysate was prepared from EDTA plasma samples with a hemoglobin concentration equal to or higher than 10 mmol/L. Plasma was centrifuged and washed with ddH2O. The washed cells were resuspended in ddH2O and frozen overnight at −20 °C to lyse the cells. The hemolysate was separated from the cell debris and stored at −80 °C until use. After spiking the serum samples, the hemoglobin concentration of the hemolysate and the highest hemolyzed serum sample were measured using Sysmex XN-9000 (Sysmex Corporation, Kobe, Japan). The hemoglobin concentrations of the other spiked samples were calculated based on the highest hemolyzed sample. Additionally, the hemolysis index of all spiked samples was measured using the Roche Cobas Pro.

Lipemia interference was studied using Intralipid^® 10 % (Fresenius Kabi, Bad Homburg, Germany) creating eight lipemic serum samples ranging from approximately 2 to 20 mmol/L triglycerides. Lipemia index and triglyceride concentration were measured using the Roche Cobas Pro.

Interference of heparin was assessed with a focus on LMWH as this anticoagulant is often used in dialysis care. Fraxiparine (Viatris Inc., Canonsburg, PA, USA) was serially diluted with the pooled serum to obtain eight different concentrations ranging from 0.06 IU/mL to approximately 2.0 IU/mL LMWH. The concentration was determined by measuring the anti-Xa on the Sysmex CS-2500 (Sysmex Corporation, Kobe, Japan).

Statistical analyses

All T50 measurements in this study were performed in triplicate unless otherwise described. The mean, SD, and CV% of the T50 measurements were calculated. The method comparison data were evaluated by determining the slope and y-intercept of the Passing Bablok regression analysis with a two-sided 95 % confidence interval (CI). Additionally, a Bland-Altman plot was used to determine the absolute (minutes) and percentage bias between the two laboratory sites. For interference analysis, the percentage deviation of the T50 value was determined using the non-spiked sample as the control. For each analyte, the data points were plotted in an XY graph based on the specific analyte concentration and the percentage deviation. A simple linear regression analysis was performed to determine the upper concentration limit before interference. All analyses were performed using EP Evaluator (Version 12.3.0.2, Data Innovations LLC), GraphPad Prism (Version 10.2.0), and R (Version 4.3.2).

Setting up T50 test

Initially, during the setup phase, a large difference in the T50 test results was noted between the two laboratory sites. In addition, large and multiple air bubbles were present in the wells after pipetting, which interfered with the nephelometric measurement. This led to a detailed search for differences in hardware and sample handling. Differences in the automated pipettors were noted. While the Freedom EVOware automatic pipetting machine was used at both sites, the liquid handling system within the machine was different: laboratory site one was air-based (Air LiHa), and laboratory site two was liquid-based (LiHa). In addition, different pipetting tips were used with wider pipette-tip orifices at site two.

Adjustments in liquid handling variables proved to be important for obtaining highly comparable outcomes at both sites. First, for handling serums samples with the LiHa system, the dispensing speed was increased by 20 % compared to the manufacturer’s protocol. Reagents were handled slower with a decrease of 40 % and 70 % for the aspirating and dispensing speed, respectively. The break-off speed was decreased by 62.5 %. Lastly, a system trailing air gap (STAG) was included with the LiHa system to prevent contamination with the system liquid.

The second hardware device required for the T50 test is a nephelometer. At both locations the NEPHELOstar was used, but at laboratory site 2, a newer version was used (NEPHLOstar plus) which has a more powerful laser. An increase in relative nephelometric unit (RNU) was observed at site 2 (Figure 2C and D); however, this increase in RNU did not influence the T50 values.

Figure 2:

Laboratory sites comparison. (A) Passing-Bablok regression demonstrating a strong agreement between two laboratory sites including the identity line and 95 % confidence interval for the regression line. (B) Bland-Altman plot between the two laboratory sited demonstrating a small mean bias of −4.0 min (−1.3 %). (C) The T50 curve of laboratory site 1 (manufacturer) in which a sigmoid curve can be observed. (D) In the T50 curve of laboratory site 2 a drop at the beginning of the graph can be seen. Additionally, higher RNU (y-axis) is measured compared to site one despite using the same sample. This initial drop and higher RNU did not affect the T50.

Laboratory comparison

After aligning the methodology as described above, a close agreement was observed between the laboratory sites (Figure 2). Passing-Bablok regression showed a high correlation coefficient of 0.993. The y-intercept indicated a non-significant bias (−4.5; 95 % CI = −24.0 – 14.2). The slope of the regression was not significantly different from 1 (1.002; 95 % CI=0.948–1.059), indicating no difference between the laboratory sites (Figure 2A). Similarly, the Bland-Altman plot also showed no significant bias between the laboratories (Figure 2B).

A difference in the T50 curves was observed between the laboratories. A higher RNU was measured at site two compared to site one when measuring the same sample, but this did not affect the T50 (Figure 2C and D). Furthermore, a drop at the beginning of the curve was observed in the majority of the T50 curves obtained at site 2. This curve deviates from the expected S-shaped curves as observed at site 1. Adjustments in the setting for liquid handling did not resolve this drop. Nonetheless, when comparing the results between the laboratory sites, the drop did not appear to affect the T50 result, as the outcomes at both sites were comparable.

Precision of T50

The performance of the T50 test was analyzed based on the S_within for each day and S_total. The CV% and errors for the three T50 levels are listed in Table 1. The results of the measurements at laboratory site two showed that the S_within and S_total for the low T50 sample were within the acceptance criteria. In contrast, the standard errors for the mid and high T50 samples did not meet the acceptance criteria.

Interference analysis

The concentrations until interference for free hemoglobin, triglycerides, and LMWH established in this study were similar to or higher than the concentrations claimed by the manufacturer (Table 2).

No clinically meaningful changes were found for potential interference by bilirubin, free hemoglobin, triglyceride, and LMWH within the normal ranges (Figure 3). For bilirubin and LMWH no critical changes were detected at the highest tested concentrations. Based on linear regression analysis, the calculated hypothetical interference concentrations were 2085 μmol/L for bilirubin and 4.35 IU/mL for LMWH. Bilirubin showed a trend with a negative coefficient which indicates that samples with extremely, unphysiologically high bilirubin concentrations will have a lower T50 compared to the true T50. As for LMWH, a positive coefficient was detected indicating that samples with extremely high LMWH concentrations will have a higher T50 compared to the true T50. Hemoglobin and triglyceride did demonstrate some interference starting from 0.18 mmol/L and 19.56 mmol/L, respectively. For both, a negative coefficient was observed.

Figure 3:

Interference of (A) bilirubin, (B) free hemoglobin, (C) triglyceride, and (D) LMWH on the T50 test presented as percentage difference from baseline (non-spiked sample). Dotted lines are showing the 10 % deviation. The colored areas show the normal expected values in clinical practice. A linear regression shows the trend of deviation of the T50 for each analyte.