A comparison between two different automated total 25-hydroxyvitamin D immunoassay methods using liquid chromatography-tandem mass spectrometry

Introduction Total 25-hydroxyvitamin D [25(OH)D] is the most reliable indicator of vitamin D status. In this study, we compared two automated immunoassay methods, the Abbott Architect 25-OH Vitamin D assay and the Roche Cobas Vitamin D total assay, with the liquid chromatography-tandem mass spectrometry (LC-MS/MS). Materials and methods One hundred venous blood samples were randomly selected from routine vitamin D tests. Two of the serum aliquots were analyzed at the Abbott Architect i2000 and the Roche Cobas 6000’s module e601 in our laboratory within the same day. The other serum aliquots were analyzed at the LC-MS/MS in different laboratory. Passing-Bablok regression analysis and Bland-Altman plot were used to compare methods. Inter-rater agreement was analyzed using kappa (κ) analysis. Results The Roche assay showed acceptable agreement with the LC-MS/MS based on Passing-Bablok analysis (intercept: -5.23 nmol/L, 95% CI: -8.73 to 0.19; slope: 0.97, 95% CI: 0.77 to 1.15). The Abbott assay showed proportional (slope: 0.77, 95% CI: 0.67 to 0.85) and constant differences (intercept: 17.08 nmol/L; 95% CI: 12.98 to 21.39). A mean bias of 15.1% was observed for the Abbott and a mean bias of -14.1% was observed for the Roche based on the Bland-Altman plots. We found strong to nearly perfect agreement in vitamin D status between the immunoassays and LC-MS/MS. (κ: 0.83 for Abbott, κ: 0.93 for Roche) using kappa analysis. Conclusion Both immunoassays demonstrated acceptable performance, but the Roche Cobas assay demonstrated better performance than the Abbott Architect in the studied samples.

1,25-dihydroxyvitamin D [1,25(OH) 2 D] (calcitriol) is not a good indicator of the vitamin D status, as it has a very short half-life of approximately 4 h, and its blood levels are closely regulated by the serum levels of parathyroid hormone, calcium, and phosphate. Calcitriol concentration also does not reflect the vitamin D reserves, as levels are frequently elevated in individuals with hypovitaminosis D because of secondary hyperparathyroidism. Therefore, total 25(OH)D is the most reliable indicator of vitamin D status (6,7).
Although the use of vitamin D testing has recently increased substantially, there is little consensus on which assay should be used to measure its concentration, and there are serious concerns regarding the reliability of its measurement (7). To correctly assess vitamin D status, a method for reliably measuring 25(OH)D is needed. Several specifications should be considered when selecting a vitamin D assay, including total 25(OH)D measurement (the sum of 25(OH)D 2 and 25(OH)D 3 ), accuracy, reproducibility, turn-around time, inter-assay comparability, and cost-effectiveness (8). Until recently, no generally accepted reference method for 25(OH)D measurement was available. In 2010, Tai et al. developed isotope dilution liquid chromatography-tandem mass spectrometry (LC-MS/MS) as a candidate reference method, which in 2011 was recognized as a reference method by the Joint Committee for Traceability in Laboratory Medicine (9). Isotope dilution LC-MS/MS is currently considered the reference method for 25(OH)D measurement, as it can simultaneously quantitate 25(OH) D 2 and 25(OH)D 3 ; these values are summed to determine total 25(OH)D (10). Recently, some manufacturers have developed automated immunoassays for 25(OH)D measurement, including Siemens, IDS, Abbott, Roche, and DiaSorin. Except for the Roche assay, all current automated tests use a similar method design. Sample pretreatment dissociates 25(OH)D from vitamin D binding protein (VDBP), and 25(OH)D competes with exogenous labeled 25(OH)D for binding to assay specific antibodies. The Roche assay is a competitive proteinbinding assay and uses exogenous recombinant human VDBP to capture 25(OH)D in the patient sample (11).
Because 25(OH)D is highly lipophilic and has strong binding affinity for VDBP, it is not easy or simple to measure its concentration in patient sera. There is a great deal of variety among 25(OH) D assays because of the different techniques employed for separating 25(OH)D from its binding protein, as well as its detection and measurement (12). Although LC-MS/MS is considered the reference method for measuring 25(OH)D concentrations, the instrument is very expensive, unavailable in most clinical laboratories, and its turnaround time is relatively longer than that of immunoassays. Therefore, automated and high-throughput immunoassays may be a good alternative for clinical laboratories. However, there is no strong agreement between the current immunoassay methods because of intermethod variability due to the different standardizations used (12,13). In cooperation with the National Institutes of Health's Office of Dietary Supplements, the National Institute of Standards and Technology (NIST) developed standard reference materials (SRMs) in order to improve accuracy and to enable worldwide standardization for 25(OH)D measurement. The SRM 2972 contains defined amounts of 25(OH)D 2 and 25(OH)D 3 dissolved in ethanol. This material is used for direct calibration of LC-MS/MS reference measurement procedures. While this material can be used with chromatographic methods as calibrators, the antigen-antibody reaction prevents direct calibration of immunological methods by using this calibration solution. Therefore, serumbased materials are needed for calibration of immunological methods. The NIST developed serum-based reference material, SRM 972, which consists of four pools of human serum with analyte values for 25(OH)D 2 , 25(OH)D 3 , and 3-epi-25(OH)D 3 . Despite these standardization efforts, unresolved discrepancies and unacceptable biases persist between immunoassays compared to the reference methods. Therefore, further standardization studies for 25(OH)D measurements should be conducted to obtain reliable results and to correctly define the vitamin D status in patients (14).
Vitamin D testing orders have increased by 20-fold in the past 3 years in our laboratory, with nearly 75% of results indicating vitamin D deficiency or Kocak FE. et al. Comparison of 25-hydroxyvitamin D assay methods insufficiency. Therefore, in this study we compared two automated immunoassay methods, the Abbott Architect 25-OH Vitamin D assay and the Roche Cobas Vitamin D total assay, while referencing the LC-MS/MS method to verify our routine 25(OH)D measurements. In this study, methods were compared by considering recommendations of the Clinical and Laboratory Standards Institute (CLSI) Evaluation Protocol 9 (EP09-A3) specifications (15). This document provides guidance for designing an experiment and selecting methods for quantifying systematic measurement error (bias or difference) between measurement procedures based on comparison of patient samples. Given the the increase in use of vitamin D testing day by day, findings of this study would provide substantial data about comparability of commonly used two immunoassay methods with LC-MS/ MS method for laboratory professionals.

Collection of samples and measurement of serum total 25(OH)D
After an overnight fasting, venous blood samples were collected into an evacuated serum separator clot activator tube (Vacuette® Z Serum Sep Clot Activator, GreinerBio-One, Kremsmunster, Austria) between 9 a.m. and 10 a.m. Blood samples were centrifuged at 1500 × g for 10 min within 1 h of collection. Hemolytic, lipemic, and icteric serum samples and samples collected in inadequate test tubes or samples with insufficient volume were excluded from the study. The serum samples were aliquoted into three separate polystyrene tubes. Two aliquots were immediately processed on the following platforms: Abbott Architect i2000 (Abbott Laboratories, Wiesbaden, Germany) by chemiluminescent microparticle immunoassay using Abbott Architect 25-OH Vitamin D assay reagent (Abbott Laboratories, Wiesbaden, Germany) and Roche Cobas 6000's module e601 (Roche Diagnostics GmbH, Mannheim, Germany) by electrochemiluminescence immunoassay using Roche Cobas Vitamin D total assay reagent (Roche Diagnostics GmbH, Mannheim, Germany). Samples were processed in a single batch in duplicate on each analyzer according to the manufacturer's instructions. Calibration curves were constructed using calibrators provided in the kits. The third aliquot was stored at -20 °C. After 1 week of storage, collected aliquots were transported to the Medical Faculty Hospital of Selcuk University on dry ice for 25(OH)D analysis by LC-MS/MS. Before analysis, frozen samples were thawed and single processing was conducted. The LC-MS/MS assay method was adapted from a method described by Sahillioglu et al. (16). MS detection was carried out with an AB Sciex API 3200 triple quadrupole liquid tandem mass spectrometer (AB Sciex Instruments, Biopolis, Singapore) using atmospheric pressure chemical ionisation as the ionization source.

Statistical analysis
The 25(OH)D results obtained by LC-MS/MS were used as the reference for method comparison studies. Results reading below or above the lower or upper limit of measurement ranges of the immunoassay methods were omitted from statistical evaluation. Concentrations of 25(OH)D were given in nmol/L International System of Units (SI). All data sets were tested for normality using Kolmogorov-Smirnov test. Passing-Bablok regression analysis was used to assess constant and proportional biases between methods, including the Cusum test for linearity. A P value < 0.05 indicates a significant deviation from linearity. For significant agreement, the 95% confidence interval (CI) of the intercept should contain the zero, while the 95%CI of the slope should contain 1 (17). A Bland-Altman plot was used to assess differences and biases between methods. Bland and Altman recommend plotting differences against the average of the methods rather than against that of the reference method, while CLSI recommends plotting differences against the reference method (15,18). Therefore, differences between values from comparative immunoassays and the reference method against the reference method value were displayed in the difference plots according to CLSI recommendations. The differences expressed as a percentage of the reference method value were plotted to illustrate whether the difference between the measurements made using the two methods was related to the magnitude of the measurement. Inter-rater agreement in assessment of vitamin D status between assays was analyzed using kappa (κ) analysis (19).  Table 1.
Differences and biases between methods were evaluated using a Bland-Altman plot (Figure 1 and 2).

25(OH)D results were classified to define vitamin D status, as recommended by The Endocrine Society's Clinical Practice Guideline on Vitamin D (20).
We compared the proportion of samples fulfilling vitamin D deficiency using the 50 nmol/L cut-off by different assays using κ analysis (Table 2).     3 . In addition, differential cross-reactivity of 25(OH)D 2 in immunoassays is another potential problem; however, this was not an issue for the differences observed between the assays in our study because 25(OH)D 2 concentrations were undetectable in the studied samples. This situation may be explained by vitamin D 3 supplementation which is common in our region.

LC-MS/MS (variable x)
Matrix effects are known to occur in immunoassays and can lead to false high or low results. The most important type of matrix effect is any effect that occurs between the matrix in the calibrants and the patient samples (13). Abbott assay's calibrators are composed of phosphate-buffered saline containing heat-inactivated horse serum and Roche assay's calibrators contain human serum as a matrix. Another factor may be the ability of an assay to separate 25(OH)D from its binding protein. In LC/MS-MS methods, 25(OH)D is separated from its binding protein by solvent extraction. However, in immunoassay methods, solvent extraction and chromatographic separation have been replaced by various blocking agents that displace 25(OHD) from VDBP, which shows varying success. Although this simplified sample pre-treatment method enables the use of high samplethroughput and automation, in case of incomplete extraction, false low 25(OH)D concentrations may be obtained. Strong binding between the highly hydrophobic 25(OH)D and VDBP creates competition with the capturing antibodies. VDBP must be be inactivated or completely removed from the sample, as residual active VDBP at concentrations as low 4 nmol/L (0.5% of total VDBP) may interfere with the assay (10,11,26). A recent study, including 50 healthy individuals, 52 pregnant women, 50 hemodialysis patients, and 50 intensive care patients suggested that in automated assays, not all of the 25(OH)D was extracted from the VDBP (27). The authors observed an inverse relationship between VDBP concentrations and deviations of immunoassay results from LC-MS/MS results. The Abbott assay (Architect i2000) showed VDBP concentration-dependent differences, but the Roche assay (Modular Analytics E170) did not show these differences (27).  (20). After classification of the 25(OH)D results according to recommendations of The Endocrine Society, we analyzed inter-rater agreement using the κ statistic and found strong to nearly perfect agreement in vitamin D status between the immunoassays and LC-MS/MS. These results are consistent with the findings of previous studies (22,28).  (29). Another limitation may be the difference in treatment between the samples used in the immunoassays (fresh) and the LC-MS/MS (one freeze/thaw cycle), but a recent study reported that long-term frozen storage does not affect serum vitamin D levels and that 25(OH)D is stable for 7 days at -20 °C (30). Additionally, sample collection tube used in this study may be possible preanalytical source of differences on different assays. However, in a previous study, the difference between serum separator tube with clot activator and EDTA tube was investigated and there was no difference between serum and plasma vitamin D concentration; the authors suggested that the choice of collection tube was not to affect vitamin D concentration (30).

Conclusıon
In this study, we found that both immunoassays demonstrated acceptable performance, but had some limitations when their performance was challenged with samples containing low and high total 25(OH)D concentrations. We found that the deviation increased in a concentration-dependent manner using Abbott Architect measurements compared to using LC-MS/MS. The Roche Cobas assay demonstrated better performance than the Abbott Architect in the studied samples. Therefore, immunoassay methods can give variable results, which is most apparent when the immunoassays are used to evaluate with a range of samples that challenge their analytical performance. Laboratory professionals should be aware of these issues when changing methods in their routine work and comparing results obtained from different platforms.