Analytical performance evaluation of different test systems on serum creatinine assay

Abstract Background Serum creatinine (SCr) is a useful diagnostic marker for the assessment of renal function. Accurate quantitation of SCr is clinically important in calculation of glomerular filtration rate (GFR). Method To confirm whether there are differences in SCr between enzymatic kits of different manufacturers, the analytical performance of the matched and open test system in the measurement of SCr was evaluated. The analytical performance evaluation was conducted according to the Clinical and Laboratory Standards Institute (CLSI) guidelines. Precision, accuracy, linearity, dilution, lower limit of measurement and analytical interference were studied between the two test systems. Results The performance of SCr from the open test system was in compliance with the matched test system with good precision, accuracy, and linearity. In presence of most common interferents, both test systems could lead to accurate creatinine results except for the existence of specified drugs. For dobutamine, the open test system showed better anti‐interference performance than the matched system. Conclusion This study provides referable opinions for clinical laboratory selection on the test system and a framework for future analogous studies based on different test systems.


| INTRODUC TI ON
Creatinine is a breakdown product of creatine phosphate from muscle and protein metabolism. 1 It is produced and released by the body at a constant rate into the blood, and it is then carried to the kidneys through the blood circulation. 2 In the body's daily metabolism, creatinine is generated and subsequently excreted in the urine because it can pass through the glomerular membrane but rarely be absorbed in the renal tubules. 3,4 In clinical practice, the level of serum creatinine (SCr) is generally used as a significant indicator for renal function evaluation. 3 SCr determination is a widely applied diagnostic test to evaluate glomerular filtration rate (GFR) that when renal function is impaired, the creatinine level rises. 5 Clinical methods for serum creatinine determination include chemical and enzymatic methods. Chemical methods are susceptible to the interference of certain creatinine derivatives or homologues, resulting in inaccurate SCr level. 2 Although the chemical method has been modified, compensation is still needed to ensure accurate SCr determination results. 6 In comparison, the application of the enzymatic method is more expensive in SCr quantification, but it avoids the poor specificity of the Jaffe method and has strong anti-interference ability as well as less reagent toxicity. 6 Therefore, the enzymatic method is a cost-effective approach for SCr determination, which is routinely employed by most clinical laboratories.
An accurate creatinine determination is pivotal; therefore, analytical performances should be concerned, as suggested by NKDEP. 7 Numerous diagnostic reagent kits with manufacturer-claimed performances can be chosen in clinical laboratory; however, results may have large variations between different test systems. In this study, we hope to compare the performance of the two test systems in order to select a system with better performance in clinical practice, so as to provide clinicians with accurate results. A matched test system means the analyzer, reagent kit, and calibrators are from the same manufacturer. 8 Similarly, an open test system refers to a system where reagent and calibrator of the same manufacturer can be adapted to distinct analyzers, so that users could make choice on what reagent kits and calibrators as they want on open test systems. 9 This study established a performance evaluation method conducted between an open test system and a clinically used matching test system (Roche SCr quantification kit on Roche Cobas 8000 C702 chemical analyzer). Performances (including precision, linear relationship, and accuracy), consistency (method comparison), and the influence of common interfering substances were compared and evaluated. Lot. #45752, 45753) from Bio-Rad Laboratories (CA, USA) were utilized throughout the study. Accuracy analyses were partially carried out on a Waters ACQUITY UPLC ® system in positive electrospray ionization mode, namely the reference system (isotope-dilution liquid chromatography/tandem mass spectrometry, ID-LC/MS method) in our study. Calibration and QC testing of the reference system had been qualified prior to sample analysis. For linearity evaluation on samples of high-level creatinine, standard reference material of creatinine (product # SRM 914a, National Institute of Standards and Technology) was applied. For evaluation of analytical specificity, interference substances were used by mixing with serum pool at three concentration levels and subsequently measured creatinine level of the mixture. In the study, the influences of hemolysis, hyperbilirubinemia, lipemia, and drugs were evaluated. Interference substances include purified biliru-  Lipemic and hemolyzed samples were excluded. Specimens for method comparison study were collected from left-over clinical patient samples (N = 34), with creatinine levels ranging from 45 to 1610 μmol/L. Serum samples were all collected from adults, and the gender is randomly distributed who had not taken medication mentioned in this study. Samples at different creatinine levels were collected as well in accuracy evaluation study and subsequently blended at different ratios to reprepare specimens at 45 levels to be assigned by the reference method. Specimens at each level was divided into three aliquots. In addition, samples with creatinine levels close to the medicine decision levels (MDLs) were collected and applied in linearity and interference evaluation.

| Precision
The evaluation of precision was carried out using four quality control materials according to the EP15-A2 evaluation protocol of the Clinical and Laboratory Standards Institute (CLSI). 11 QC materials were divided into five aliquots per level and frozen at −80 ºC. One aliquot of each level was thawed at room temperature 30 minutes before analysis and gently inversed to homogenize. Triplicate creatinine measurements of each level were then daily performed for a total of five nonconsecutive days (N = 15 per level). Results were presented in terms of coefficient of variation (CV%). The calculation is performed using the formulas as shown in the previous study. 12 A recommended minimum CV for serum creatinine analysis is less than 3.2%, which is three quarters the intraindividual biological variation. 13

| Accuracy
In the study of accuracy evaluation, five EQA materials were measured in triplicate by both test systems, and the test results were compared within a medically allowable bias to NCCL-given (NCCL, National Center for Clinical Laboratories) target values. The accuracy was accepted if they were within target values ±1/2 TEa% (allowable total error; 6%). Furthermore, forty-five fresh frozen/thawn patient samples evenly distributed over the measuring interval were determined in triplicate by the reference system and the two test systems. Results from the two test systems were compared to those from the reference system by performing a Passing-Bablok regression, and the estimated values at the MDLs were then calculated.
The relative biases between estimated values and MDLs were compared with the allowable specification. In addition, a Bland-Altman (BA) plot of percent differences of the two test systems and IDMS results were constructed.

| Linearity, dilution, and lower limit of measurement
The test for linearity was carried out in accordance with the CLSI protocol. We used standard reference material of creatinine to simulate high-level creatinine which claimed by manufacturers. The creatinine concentration of the median samples covered high concentration of samples that can be seen in clinical practice and the reference interval range. Samples with the level of creatinine close to the upper limit of reference interval were applied to evaluate the linearity of low-level creatinine. The linearity of high-level, medianlevel, and low-level creatinine were evaluated using eleven respective pools. In the linearity of high-level creatinine, besides the pool containing the highest-level concentration, the other ten pools were from mixtures of the mentioned serum sample with saline solution at ratios of 0:10, 1:9, 2:8, 3:7, 4:6, 5:5, 6:4, 7:3, 8:2, 9:1, and 10:0.
In the linearity of low-level creatinine, besides the pool containing creatinine concentration that close to the upper limit of reference interval, the other pools were prepared accordingly. Besides the pool containing creatinine concentration that close to either the lower limit of reference interval or clinical samples with high-level creatinine, the other nine pools were prepared correspondingly. Each pool was measured in triplicate. The linearity results were presented as the equation from the linear regression including slope, Y intercept, and the correlation coefficient (R 2 ). As acceptance criteria, the slope should be close to 1 and R 2 ≥ 0.99.
The dilution study was conducted by measuring clinical samples diluted by saline solution, and the allowable dilution ratio was acceptable when the bias was within 6%.
Isotonic saline solution (0.9%) was used to evaluate the lower limit of measurement. Each specimen was tested in 20 replicates, and mean values and standard deviation (SD) were then calculated.
Lower limit of measurement was defined as the mean value +3 SD according to the manufacturer.

| Method comparison
Comparison study was performed to compare the open test system to the matched test system, according to the CLSI EP09-A3 guidelines, 14 and subsequently, the correlation was evaluated. A total of 34 patient serum samples were aliquoted into two fractions, which were measured using the two systems within 2 hours. A Passing-Bablok regression was then conducted based on measured values, and relative bias at MDLs was calculated. A Bland-Altman (BA) plot was made accordingly.

| Analytical interferences
Interference study was carried out by adding different solutions of interferents to serum pools at three levels of creatinine. Each sample was determined in triplicate using the two test systems. The values of creatinine level of original pool were set as baseline values in which the measurement values were compared with in order to calculate the percentage creatinine recovery. Significant interference was defined when a recovery change exceeded 10% of the baseline values. [15][16][17] The relative bias of each specimen was calculated from the observed value and the baseline value. Table and plots were constructed to illustrate the influence of common interferents.

| Interference of hemolysis, hyperbilirubinemia, and lipemia
To study the influence of hemolysis, hemolysate was added to each serum pool aliquot according to Fleming and Swaminathan. 18 The hemolysate was prepared from EDTA-anticoagulated whole blood.
The blood was centrifuged to separate plasma and cells, and the cells were then washed three times with saline solution (0.9% NaCl).
Later, the supernatant was removed and distilled water was added to the cells. After 15 min standing at 4°C, the mixture was centrifuged and separated. Subsequently, the concentration of hemoglobin for the hemolysate was determined at 60 g/L. A specified volume of hemolysate was then spiked with different serum pool aliquots at 1:9 ratio to obtain interference samples with hemoglobin concentration

| Statistical analysis
Statistical analysis was performed using the Microsoft Excel

| Precision
The open test system yielded within-run CVs ranging from 0.31% to 0.80% and total CVs ranging from 0.29% to 1.19%, while within-run CVs and total CVs of the matched test system were 0.56%-0.95% Measurand N

| Accuracy
Both test systems showed high accuracy. Firstly, the relative biases against NCCL target values of either test system were within ±6% limit, which were clinically acceptable (  (Table 3). According to the Bland-Altman analysis, the percent differences were small on creatinine determination, suggesting that the two observed test systems and the reference system reached good agreement ( Figure 1). These data indicated that the two observed test systems could lead to accurate determination results for patient samples on clinical practice.

| Linearity, dilution, and lower limit of measurement
The results of the linearity and dilution studies were shown in Table 4 and

| Method comparison
The correlation between systems was obtained by analyzing 34 samples with creatinine level in a dynamic range from 45 to 1610 μmol/L. The matched test system was used as a comparative method. The measured values from the two systems were analyzed and described with the Passing-Bablok regression fit: Y = 0.984 (95% CI: 0.977 to 0.996) X − 0.541 (95% CI: −1.967 to 0.580) (Figure 2A). Additionally, the relative biases at medical decision levels were −2.62%, −1.98%, and −1.70%, respectively, which were acceptable. Besides, the Bland-Altman plot in Figure 2B showed a mean percent difference of −1.842% and 95% limits of agreement ranging from −5.178% to 1.494%, and only 2 out of 34 data points fell out of the 95% limits of agreement depicted by the upper and lower line.

| Analytical interferences
The interference experiment was conducted by comparing determination results of serum samples with/without interfering substances, with an allowable bias of ±10%. As shown in Table 6 in the low-level creatinine group ( Figure 3A). Moreover, the same calcium dobesilate concentration produced significant interference in the matched test system in the medium-level creatinine group with the bias of −14.2% ( Figure 3A). The open test system exceeded the acceptable criteria of relative bias of ±10% in the medium-level creatinine group at calcium dobesilate concentration of 16 μg/ml ( Figure 3A). In the elevated-level creatinine group, calcium dobesilate concentration at 16 μg/ml caused significant interference (relative bias of −12.5%) in the matched test system, and 32 μg/ml calcium dobesilate negatively affected creatinine determination with relative bias of −14.9% in the open test system ( Figure 3A). As shown in Figure 3B, in the presence of 8 μg/ml

| DISCUSS ION
Serum creatinine is a biomarker for estimating glomerular filtration rate (eGFR) in patients and provides clinicians with an assessment of renal function. 20 Figure 1 and Table 3, results of each test system were in accordance with those assigned by the reference system, and relative biases at MDLs were both clinically acceptable, indicating either test system could lead to correct results for patient samples on clinical practice. Regarding the linearity of the test systems, the results showed that the correlation coefficients were >0.99 and the slopes were both close to 1, which indicated excellent linearity over ranges of high-, median-, and low-level creatinine. In TA B L E 4 Evaluation of the linearity of creatinine at different levels  The traditional picrate acid (Jaffe) method used to determine serum creatinine concentration is lack of analytical specificity as it is known to be subject to interference from certain substances. [23][24][25][26][27] The US National Kidney Disease Education Program Laboratory Working Group promoted the use of enzymatic assay for creatinine quantification, and it has been widely implemented for routine clinical laboratory use because it could offer more specific creatinine determination with improved accuracy. 7,23 However, creatinine determination by enzymatic assay is still reported that interfered by several substances. [28][29][30] In our study, we investigated the antiinterference performance of the test systems with the presence of common interferents or specific drugs. There was no significant interference observed in the presence of interferents listed in Table 6.
Furthermore, the interference of another two medications, calcium dobesilate and dobutamine were analyzed as well. Calcium dobesilate (calcium 2,5-dihydroxybenzenesulfonate) is a well-known vasoprotectant and also exerts protective effect on diabetic nephropathy 13 and gentamicin-induced acute kidney injury. 31

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.

R E FE R E N C E S F I G U R E 3
Effect of interference from calcium dobesilate and dobutamine on creatinine determination. Samples of creatinine at three different levels with interference substances or not were determined in triplicate using the two test systems, respectively. The acceptable bias against baseline values is set at 10%