Influence of Vitamin D Binding Protein on Accuracy of 25-Hydroxyvitamin D Measurement Using the ADVIA Centaur Vitamin D Total Assay

Vitamin D status in different populations relies on accurate measurement of total serum 25-hydroxyvitamin D [25(OH)D] concentrations [i.e., 25(OH)D3 and 25(OH)D2]. This study evaluated agreement between the ADVIA Centaur Vitamin D Total assay for 25(OH)D testing (traceable to the NIST-Ghent reference method procedure) and a liquid chromatography tandem mass spectrometry (LC-MS/MS) method for various populations with different levels of vitamin D binding protein (DBP). Total serum 25(OH)D concentrations were measured for 36 pregnant women, 40 hemodialysis patients, and 30 samples (DBP-spiked or not) from healthy subjects. ELISA measured DBP levels. The mean serum DBP concentrations were higher for pregnancy (415 μg/mL) and lower for hemodialysis subjects (198 μg/mL) than for healthy subjects and were highest for spiked serum (545 μg/mL). The average bias between the ADVIA Centaur assay and the LC-MS/MS method was −1.4% (healthy), −6.1% (pregnancy), and 4.4% (hemodialysis). The slightly greater bias for samples from some pregnancy and hemodialysis subjects with serum DBP levels outside of the normal healthy range fell within a clinically acceptable range—reflected by analysis of their low-range (≤136 μg/mL), medium-range (137–559 μg/mL), and high-range (≥560 μg/mL) DBP groups. Thus, the ADVIA Centaur Vitamin D Total assay demonstrates acceptable performance compared with an LC-MS/MS method for populations containing different amounts of DBP.


Introduction
Increasing awareness of the important role of vitamin D for bone and other diseases has led to increased 25hydroxyvitamin D [25(OH)D] testing (D represents D 3 and D 2 forms). However, variability within and between methods and laboratories has often compromised correct diagnosis and the ability to compare results from different studies and national surveys [1][2][3][4][5]. Automated antibodybased, radioimmunoassays, high performance liquid chromatography (HPLC), and mass spectrometry methods for 25(OH)D testing are subject to variability issues that can arise from a variety of sources, such as differential detection of the D 3 and D 2 forms, interference by detection polyclonal antibodies, and nonspecific detection of other vitamin D metabolites such as the 3-epimer form of 25(OH)D [3epi-25(OH)D] and 24,25(OH) 2 D 3 . In addition, incomplete release of 25(OH)D from the vitamin D binding protein (DBP) has been identified as a potential source of variability for both manual and automated immunoassays [6].
Establishing an immunoassay for 25(OH)D is challenging because the majority of the highly hydrophobic 25(OH)D is tightly bound (dissociation constant, Kd, 5 × 10 −8 M) to a vast excess of DBP from which it must be separated; almost no 25(OH)D is found "free" (non-protein bound) in the circulation, and less than 5% of the available DBP binding sites are occupied by vitamin D compounds [7]. In addition, DBP binds vitamin D 3 along with other metabolites and vitamin D 2 , whose similar structures may be easier to release from DBP and difficult to differentiate; DBP has a higher affinity for vitamin D 3 than other metabolites and vitamin D 2 [8]; and generating specific antibodies against small antigenic molecules, such as 25(OH)D, is difficult, but it is mandatory because the Vitamin D Standardization Program (VDSP) states that 25(OH)D assays should measure 2 International Journal of Endocrinology equimolar amounts of 25(OH)D 3 and 25(OH)D 2 (total vitamin D) [9]. Measuring total vitamin D is required because some supplements contain the D 2 form, and not measuring D 2 would lead to lower 25(OH)D values. In methods such as radioimmunoassay, HPLC, and mass spectrometry, an initial extraction step with organic solvents releases all bound 25(OH)D from DBP [10][11][12][13]. However, organic solvents are not compatible with most automated immunoassays, and alternative releasing agents, which are proprietary, are used instead. Recent studies performed in populations with different levels of DBP have questioned the effectiveness of these proprietary releasing agents to completely free 25(OH)D from DBP [6].
The goal of this study was to examine the ability of the ADVIA Centaur Vitamin D Total assay by comparison with an LC-MS/MS method to accurately measure 25(OH)D levels in serum samples from healthy adults (endogenous) and healthy adults with exogenous DBP (endogenous + spiked) and from pregnant women and chronic kidney disease (CKD) patients receiving dialysis, who have higher and lower than normal serum levels of DBP, respectively [7,14,15]. The ADVIA Centaur Vitamin D Total assay is traceable to the NIST-Ghent reference measurement procedure (RMP) for vitamin D testing. (This version of the ADVIA Centaur Vitamin D Total assay is not currently available commercially in all regions, including the USA.)

Materials and Methods
In order to determine the influence of DBP on a vitamin D immunoassay, a study examining DBP as an endogenous interference, similar to how hemoglobin, cholesterol, or total protein would be measured, following Clinical and Laboratory Standards Institute (CLSI) Document EP7-A2 [16] was performed at the Siemens R&D facility in Tarrytown, NY, USA. Human native DBP (>95% pure) was purchased from Athens Research & Technology, Inc.

LC-MS/MS.
The LC-MS/MS method used in this study is traceable to the Esoterix ID-LC-MS/MS method, which is traceable to NIST. The LC-MS/MS method performed at Siemens used the Waters Acquity H-class ultrahigh performance liquid chromatography (UPLC) and triple quadrupole (TQD) tandem mass spectroscopy (MS) with MassLynx and QuanLynx software (Waters Acquity TQD system, Waters Corporation, Manchester, UK). This method is able to separate, identify, and separately quantify the concentrations of 25(OH)D 2 , 25(OH)D 3 , and 3-epi-25(OH)D 3 in a serum sample. As reported by the manufacturer, the LC-MS/MS method demonstrated a dynamic assay range of 2.5-220 ng/mL (6.25-550 nmol/L) ( 2 > 0.997). Three levels of 25(OH)D 2 and 25(OH)D 3 concentrations tested over five consecutive days yielded intra-assay coefficients of variation (CVs) of ≤7.7% and interassay precision CVs of <12% for 25(OH)D 2 and 25(OH)D 3 .  [4,9,17]. Recently, Siemens received confirmation from the Centers for Disease Control and Prevention (CDC) that the ADVIA Centaur Vitamin D Total assay is now a certified procedure of the Vitamin D Standardization-Certification Program (VDSCP). Certification was achieved by demonstrating that the total vitamin D [25(OH)D] results for the 40 VDSCP samples (10 quarterly challenges) agreed with the results from the ID-LC-MS/MS RMP method. The ADVIA Centaur Vitamin D Total assay achieved a mean bias of 0.3% (acceptance criterion was ±5.0%) and a mean imprecision of 5.5% (acceptance criterion was <10.0%). The ADVIA Centaur Vitamin D Total assay is a one-pass, 18-minute antibody competitive chemiluminescent immunoassay.  [18]. Precision was determined by assaying six samples twice a day in replicates of 4, over 20 days ( = 160 replicates per sample) according to the Clinical and Laboratory Standards Institute (CLSI) protocol EP5-A2 [19]; the run-to-run CVs were in the range of 4.2% and 11.9% [18]. All samples were run in singlicate on both the LC-MS/MS and a single ADVIA Centaur system.

Sample Population.
Clinical serum samples from 18 healthy adults were purchased from a commercial vendor (ProMedDx, Norton, MA, USA). Serum samples from 36 pregnant women in their third trimester and 40 CKD hemodialysis patients were purchased from another commercial vendor (Research Sample Bank, Delray Beach, FL, USA).

Samples.
Peripheral venous blood samples were collected, placed at 4 ∘ C, and centrifuged; serum aliquots were prepared and stored for less than four months at -20 ∘ C until analysis. Generally, no difference was found in the serum concentrations of DBP for men and women [11,15].   ). Subsequently, each of the five serum pools was divided into six aliquots, and DBP (ranging from 50 to 250 ug/mL in 50 ug/mL increments) was spiked into 5 of the 6 aliquots from each pool ( Table 3) 2.6. Statistics. Difference plots and bias values were obtained using Microsoft Excel (2010); Analyze-It add-in program in Excel was used to compare the different sets of data in order to obtain the 95% confidence interval (CI) and standard deviations (SD) for 95% limits of agreement. Correlation plots and correlation and Deming regression analyses were generated using GraphPad Prism, version 6.

Results
The mean serum concentrations of DBP in healthy subjects (endogenous and endogenous + spiked), pregnant women, and dialysis patients are presented in Table 1. For the five serum pools, the average endogenous serum DBP concentration (±SD) was 348 ± 106 g/mL (range 260.7 to 519.0 g/mL), which is consistent with the results of other studies [6,20,21]. For healthy serum samples spiked with DBP, the average DBP concentration was higher (545 ± 185 g/mL, range 261.2 to 980.6 g/mL) than endogenous levels. For pregnancy samples, the average DBP concentration was also greater (415 ± 245 g/mL, range 82.2 to 874.5 g/mL) than that for healthy serum samples. In contrast, for CKD patients receiving dialysis, the average DBP concentration was lower (198 ± 173 g/mL, range 63.4 to 1115.7 g/mL; median 142.1 g/mL) than levels in healthy serum and pregnancy samples. The mean total serum 25(OH)D concentrations and range as measured by the LC-MS/MS method and the ADVIA Centaur Vitamin D Total assay are presented in Table 2. The mean 25(OH)D levels (±SD) were 44.6 ± 19.8 and 44.8 ± 20.1 ng/mL for healthy serum samples (endogenous), 44.6 ± 18.0 and 43.5 ± 16.7 ng/mL for healthy (endogenous + spiked) serum samples, and 44.6 ± 18.0 and 43.7 ± 17.0 ng/mL for both endogenous and endogenous + spiked healthy serum samples, and they were lower for pregnancy serum samples, 27.3 ± 9.6 and 25.3 ± 8.7 ng/mL, and dialysis serum samples, 28.1 ± 14.8 and 29 ± 15.3 ng/mL. Consistent with previous reports, no correlation was found between the DBP and 25(OH)D concentrations for serum from dialysis patients (Pearson's correlation coefficient = 0.1) [11,14,15]. Pregnancy samples demonstrated a positive correlation ( = 0.35; = 0.013) between serum concentrations of DBP and 25(OH)D for LC-MS/MS, but no correlation was found for ADVIA Centaur ( = 0.15; = 0.37). There were too few non-spiked healthy samples for valid 25(OH)D and DBP correlation assessment.
The overall average bias of all samples from healthy individuals (endogenous and endogenous + spiked) for the ADVIA Centaur Vitamin Total assay to the LC-MS/MS method was -1.4%; for all third-trimester pregnancy samples, the average bias was -6.1%; and for all renal dialysis samples, the average bias was 4.4%. The results for bias, percent   bias, 95% CI, and SD (95% limits of agreement = 1.96 SD) as a function of DBP concentration for each populationseparate and combined-are presented in Figure 1. When all populations were combined, positive bias (versus LC-MS/MS) was observed at very low serum DBP concentrations and negative bias was observed at very high serum DBP concentrations ( Figure 1). With respect to dialysis samples with generally lower DBP concentrations, we do not know if uremic serum properties contributed to bias, and we question the validity of analyzing combined populations. Nevertheless, we examined how well the methods in subjects with serum DBP concentrations at extremes of the serum DBP concentration range (very low and very high)-for combined and separate populations (Figures 2, 3, and 4). Very low and very high serum DBP concentrations were defined as two SD below and above the mean for healthy subjects which is 348 ± 106 g/mL; hence, the very low DBP group comprised samples having concentrations of ≤136 g/mL, and high DBP group comprised samples having concentrations of ≥560 g/mL. The middle range group had samples with DBP concentrations ranging from 137 to 559 g/mL. The following populations were analyzed: (1) healthy, spiked, pregnancy, and dialysis ( Analysis of 25(OH)D values for ADVIA Centaur and LC-MS/MS as a function of low, medium, and high serum DBP concentrations demonstrated that correlations between the two methods were acceptable at low and high serum DBP levels for all populations analyzed (combined and separate) (Figures 2, 3, and 4), with pregnancy samples demonstrating the lowest correlation at very high serum DBP concentrations ( = 0.87, < 0.0002). Healthy samples (endogenous and endogenous + spiked) showed very good correlations and agreement between methods. The mean bias obtained for combined populations and each population separately for their low, medium, and high range DBP groups represented acceptable assay performance (Table 4).

Discussion
This study addressed the influence of DBP on the accuracy of the ADVIA Centaur Vitamin D Total assay by comparison with an LC-MS/MS method. The ADVIA Centaur Vitamin D Total assay results in this study are traceable to ID-LC-MS/MS 25(OH)D reference method procedure and the standard reference materials established by NIST and the University of Ghent [4,9,17,22].
In healthy individuals, endogenous serum DBP concentration (347.6 g/mL) was found to be within the range reported by others (300-600 g/mL) [20] and increased after DBP spiking (545 g/mL). Despite the increase in DBP concentrations after spiking, 25(OH)D measurements in individual samples were equivalent between the two methods. There were no healthy (endogenous or endogenous + spiked) samples in the low DBP range, as defined by two SD values below the mean of healthy samples (i.e., ≤136 g/mL) ( Figure 3). The high DBP range, as defined by two SD values above the mean of healthy samples (i.e., ≥560 g/mL), comprised twelve spiked samples (Figure 3), and 25(OH)D agreement was good between the methods for 25(OH)D values ( = 0.9927, < 0.0001; bias −3.0 ± 3.98%). Thus, there was not a significant bias observed for the ADVIA Centaur for healthy samples (endogenous + spiked). This demonstrates that DBP concentrations as high as 980 g/mL did not appear to interfere with the assay for this population. By comparison, endogenous serum DBP concentrations peaked at 519 g/mL in healthy subjects. Because the use of DBP-spiked samples may be suspect, these results will be confirmed in future studies which evaluate a greater number of samples from healthy subjects containing endogenous serum DBP concentrations in the higher range (although it is unlikely that normal healthy subjects exist with DBP concentrations that can be achieved at the high spiking concentrations). Depending on hormonal status or disease state serum matrix components may be different, and the levels of DBP may be higher or lower than those of healthy individuals [14,15]. In women who are receiving estrogen therapy and those who are pregnant, higher serum estrogen levels correlate with increases in circulating DBP and total 1,25(OH) 2 D. During pregnancy, increased 1,25(OH) 2 D 3 occurs in response to the growing calcium demands of the fetus [14,15]. Consistent with these reports, the mean DBP concentration was greater (415 g/mL) for samples from pregnant women than for those from healthy subjects (347.6 g/mL) and dialysis patients = 0.87, < 0.0002, bias -14.0 ± 9.9%, resp.). Although samples in the low and medium DBP range showed  less bias than those with very high DBP concentrations, the assay performance was acceptable for all groups. Four samples out of 36 contained measureable 25(OH)D 2 ; it is unlikely that 25(OH)D 2 influenced the assay bias because several samples lacking 25(OH)D 2 demonstrated similar levels of bias. In contrast to a previous study that found higher 25(OH)D levels in pregnant women compared to nonpregnant healthy women, this study found overall lower levels in pregnant women; this difference may relate to differences in vitamin D supplementation [6]. Nephrotic syndrome and CKD predialysis and dialysis patients demonstrate diminished serum levels of the bioactive 1,25(OH) 2 D, likely due, in part, to impaired renal synthesis, nutritional deficit, and lower 25(OH)D substrate levels [23][24][25][26]. Although some studies report no change in serum DBP levels in renal failure patients compared with healthy individuals, other studies demonstrate lower serum levels and increased DBP urinary excretion; lower serum concentrations of DBP likely reflect increased urinary loss due to proteinuria, which is a common finding in CKD patients [15,27,28]. In this study, the overall mean 25(OH)D level was equivalent between the ADVIA Centaur Vitamin D Total assay and the LC-MS/MS method and for the low and medium range DBP groups ( = 0.97 overall; = 0.98 low range DBP group; = 0.97 middle range DBP group, < 0.0001; bias was 4.35 ± 12.4% overall, 10.0 ± 10.6% for the low range DBP group and 0.0 ± 12.4% for the middle range DBP group, resp.), indicating acceptable performance of the ADVIA Centaur Vitamin D Total assay in the presence of DBP and uremic serum. It is not known whether unique components of uremic serum contributed to the bias observed. Nineteen samples from dialysis patients had detectable 25(OH)D 2 (range 1.6 to 35 ng/mL), eight of which had levels above 10 ng/mL. The 25(OH)D 2 containing samples appeared to contribute to the positive bias in this patient population. This result is consistent with the performance of the ADVIA Centaur Vitamin D Total assay which demonstrates a slight difference in recovery for 25(OH)D 2 and 25(OH)D 3 (104.5% versus 100.7%) as stated in the Instructions for Use Manual [18]. Only one uremic sample was found in the higher range [1115.7 g/mL DBP; 19.1 ng/mL 25(OH)D 3 by LC-MS/MS and 21.4 ng/mL 25(OH)D by ADVIA Centaur; 12% bias of ADVIA Centaur to LC-MS/MS]. Whether an error in DBP measurement was the cause for the unusually high DBP concentration is not known. Although the serum 25(OH)D concentrations in dialysis patients were lower than those found in healthy individuals, the values were approximately normal (according to the Endocrine Society Guidelines). This is likely due to patient adherence to vitamin D supplementation which is indicated for end-stage renal disease patients on dialysis. It is worth noting that lower levels of serum 25(OH)D concentrations in predialysis patients correlate with a greater risk of mortality [29]. This underscores the need to accurately evaluate and monitor serum 25(OH)D levels in the CKD patient population.
A recent study implicated ineffective 25(OH)D-DBP extraction procedures as the cause of variability in an evaluation of five automated assays compared to an RMP LC-MS/MS method [6]. The study, which included samples from healthy individuals, pregnant women, dialysis patients, and intensive care patients, found that the bias was, at least in part, dependent on DBP concentration. The ADVIA Centaur Vitamin D Total assay in the present study differed from the assay in the previous study in that it has a different standardization; this version is standardized with internal standards traceable to the NIST-Ghent VDSP RMP. This may have had some impact on why the results of this study differ from those previously reported.

Conclusions
The small positive bias found in renal dialysis patients with DBP concentrations below those found in normal healthy subjects and small negative bias found in pregnant subjects with DBP levels above those found in normal healthy subjects were within the acceptable range for the assay. Thus, for populations with different levels of DBP, the 25(OH)D results obtained by the ADVIA Centaur Vitamin D Total immunoassay were equivalent to the sum of 25(OH)D 2 and 25(OH)D 3 using the LC-MS/MS method-especially for individuals with serum DBP concentrations within the range for the healthy population (137 to 559 g/mL).