Participating laboratories and experimental design
The present study was performed in 2019 and involved 23 clinical laboratories located in 23 hospitals from 15 cities in different regions of China. The biochemical analysers used in the 23 laboratories included the Abbott Architect (n = 1) (Abbott Diagnostics, Park City, IL, USA), Beckman AU (n = 15) (Beckman Coulter, Brea, CA, USA), Hitachi 7600 or 7180 (n = 3) (Tokyo, Japan), and Roche Cobas (n = 4) (Roche, Basel, Switzerland). The analytical performance of the Cr-R assays was evaluated on these 23 platforms. The Cr-R assays and calibrators were provided by Maccura Biotechnology (Chengdu, China). All reagents and calibrators used in this study were standardised across participating laboratories, with matching lot numbers. The detailed characteristics of the instruments and Cr-C assays used by the 23 participating laboratories are shown in Supplementary Table 1. Precision, method comparison and bias estimation, and anti-CaD ability of the Cr-R assay were evaluated based on the experimental workflow in Fig.1.
Preparation of serum samples
Pooled serum samples were prepared in a central laboratory (Department of Laboratory Medicine, Peking Union Medical College Hospital, PUMCH) and used to measure the precision of the different creatinine assays, as well as for method comparison, and in vitro interference evaluation. The creatinine concentrations of the low- and high-level serum pools used the in the precision study were approximately 80 and 300 μmol/L, respectively. A panel of 40 serum pools were prepared for method comparison to cover a wide range of creatinine concentrations, from 40–1000 μmol/L. The creatinine concentrations of the three base serum pools used for the in vitro interference assessment were approximately 80, 133, and 442 μmol/L, respectively. The serum pools were fully mixed with non-icteric and non-haemolysed serum samples obtained from patients receiving physical examinations or who were hospitalised at PUMCH in June 2019; all patients had appropriate creatinine concentrations and no CaD. All pools were prepared within one week and frozen at −80 °C until shipped on dry ice to the reference laboratory (ID-MS/MS) and 23 participating clinical laboratories. The pools were stored in the participating laboratories at −80 °C until analysis.
The precision evaluation was performed according to Clinical and Laboratory Standards Institute (CLSI) EP15-A guidelines . The high- and low-level serum pools were evaluated using the Cr-R assay four times daily for five consecutive days in each of the 23 participating clinical laboratories. An internal quality control was included to ensure test quality. The standard deviation (SD) and coefficient of variation (%CV) were calculated to determine the repeatability (within-run precision) and total precision (within-lab precision) for each assay. The inter-laboratory variation in the 23 participating laboratories was also calculated.
The within subject (CVI) (4.4%) and between-subject (CVg) (14.3%) variation of creatinine from the European Federation of Clinical Chemistry and Laboratory Medicine (EFLM) Biological Variation Database  was used for the assessment of the performance characteristics of the creatinine assays. The optimal, desirable, and minimal analytical precision for creatinine were 1.1, 2.2, and 3.3%, respectively, calculated using the following formula :
Optimal Imprecision as CV (%): CVoptimal = 0.25 CVI; desirable imprecision as CV (%): CVdesirable = 0.5 CVI; minimal imprecision as CV (%): CVminimal = 0.75 CVI.
Method comparison and bias estimation
The creatinine assays were compared according to CLSI EP9-A3 guidelines . Forty samples were measured in the 23 laboratories using the Cr-R assayson their automated analysers in 5 days, with 8 samples per day. Each sample was tested twice and within 3 h of thawing. The target values of serum pools were assigned by a reference measurement based on isotope dilution liquid chromatography-tandem mass spectrometry (LC-IDMS/MS) in two joint committees on traceability in laboratory medicine-listed reference laboratories: the NCCL, China, and the Reference Laboratory of Maccura Biotechnology (Chengdu, China). The average of these two laboratory measurements was used as the target value for each sample.
Correlation studies for each creatinine assay and the reference method were performed using Passing-Bablok regression, and bias at the medical decision levels and the 95% confidence interval (CI) were calculated. We set 88.4 μmol/L, 133 μmol/L, 265 μmol/L, and 442 μmol/L as the levels for medical decisions involving creatinine. The analytical performance specification applied in bias evaluation was based on the EFLM BV data . The optimal, desirable, and minimal bias specification for creatinine were defined as ±1.9%, ±3.7%, and ±5.6%, respectively, calculated using the following formula:
Optimal bias%: Biasoptimal = 0.25 (CVI2 + CVg2)1/2; desirable bias%: Biasdesirable = 0.5 (CVI2 + CVg2)1/2; minimal bias%: Biasminimal = 0.75(CVI2 + CVg2)1/2.
In vitro interference study
Three drug-free serum pools (Cr = 80, 150, 442 μmol/L) were spiked with a CaD standard to prepare a dose-response series according to CLSI EP7-A2 guidelines . The final CaD concentrations in each sample were 0, 8, 15, 30, 60, and 100 μg/mL. The CaD standard (100573, 94.8%) used in in vitro experiments was purchased from the National Institutes for Food and Drug Control of China (Beijing, China). In each participating hospital, the CaD-containing serum series was analysed using the Cr-C and Cr-R assay in triplicate within one analytical run to obtain an average value. The percent bias at each concentration of CaD was calculated relative to that of the drug-free specimen. Based on the biological variation, the acceptable limit of deviation for the in vitro experiments was defined as ±5.6%.
Interference study based on real-world clinical data
Twenty-three participating laboratories used the Cr-C and Cr-R assays to simultaneously detect creatinine in the remaining serum of the clinical samples for 3–5 consecutive days. Each laboratory evaluated approximately 3000 samples, which is clinically required to detect creatinine. If the creatinine measurement according to the Cr-C assay was 10% lower than that of the Cr-R measurement, the remaining serum was collected and stored in aliquots at -80 °C. CaD levels were accurately measured in these serum samples at PUMCH as described below. Fifty samples were randomly selected from those samples containing CaD to measure the creatinine levels using the above reference method, and deviations between the results of the two assays and the reference method were calculated.
Serum CaD concentration measurement by ultra-performance liquid chromatography (UPLC)
CaD (μg/mL) levels were measured using UPLC at the Department of Laboratory Medicine, PUMCH. Chromatographic separation was performed on an ACQUITY UPLC® BEH C18 column (100 µm × 2.1 mm, 1.7 μm beads, Waters, Milford, MA, USA). The mobile phase was HPLC-grade water (A) (0.3% trimethylamine and pH regulated to 3.5 with glacial acetic acid) and acetonitrile (B). Isocratic elution was performed at a flow rate of 0.6 mL/min. The detection wavelength was 305 nm, and the retention time was approximately 4 min. The low and high limits of quantitation were 0.977 and 250 μg/mL, respectively. The analytical recovery was 100 ± 5%. Within-laboratory CVs ranged from 2.46% to 3.31%.
MedCalc statistical software (ver. 18.11.6; MedCalc Software, Ostend Belgium) and GraphPad Prism software (ver. 5.01; La Jolla, CA, USA) were used for statistical analysis. Continuous data are presented as the means and standard deviations or medians (range, interquartile range [IQR], 25–75%], as appropriate. Paired Student’s t tests were used for all comparisons, with P < 0.05 indicating a statistically significant difference.