Journal of Steroid Biochemistry & Molecular Biology 148 (2015) 41–46
Contents lists available at ScienceDirect
Journal of Steroid Biochemistry & Molecular Biology journal homepage: www.elsevier.com/locate/jsbmb
Review
Effect of 25(OH) vitamin D reference method procedure (RMP) alignment on clinical measurements obtained with the IDS-iSYS chemiluminescent-based automated analyzer Christine A. Simpson * , Anna Maria Cusano, Jessica Bihuniak, Joanne Walker, Karl L. Insogna Department of Internal Medicine, Section of Endocrinology, Yale School of Medicine, 330 Cedar Street, New Haven, CT 06520-8020, United States
A R T I C L E I N F O
A B S T R A C T
Article history: Received 5 August 2014 Accepted 12 September 2014 Available online 18 September 2014
The Vitamin D Standardization Program (VDSP) has identified ID-LC/MS/MS as the reference method procedure (RMP) for 25(OH) vitamin D and NIST Standard SRM2972 as the standard reference material (SRM). As manufacturers align their products to the RMP and NIST standard, a concern is that results obtained in aligned assays will be divergent from those obtained with pre-alignment assays. The Immunodiagnostic Systems Ltd., chemiluminescent, 25(OH) vitamin D iSYS platform assay, was recently harmonized to the RMP. To determine the impact of standardization on results obtained with iSYS reagents, 119 single donor serum samples from eight different disease categories were analyzed in four non-standardized and two standardized iSYS assays. There were strong correlations between the four non-standardized and two standardized assays with Spearman's rank r values between 0.975 and 0.961 and four of the eight r values were >0.97. R2 values for the eight best-fit linear regression equations ranging between 0.947 and 0.916. None of the slopes were found to be significantly different from one another. Bland–Altman plots showed that the bias was comparable when each of the four nonstandardized assays was compared to either of the standardized assays. When the data were segregated in values between 6 and 49 ng/mL (15–122 nmol/L) or between 50 and 100 ng/mL (125–250 nmol/L) significant associations remained between results obtained with non-standardized and standardized calibrators regardless of the absolute value. When five recent DEQAS unknowns were analyzed in one non-standardized and one standardized assay the mean percent difference from the NIST target in values obtained using standardized vs. non-standardized calibrators were not significantly different. Finally, strong and statistically significant associations between the results were obtained using nonstandardized and standardized assays for six of eight clinical conditions. The only exceptions were hypocalcemia and breast cancer, which likely reflect the small sample sizes for each of these diseases. These initial data provide confidence that the move to a NIST standardized assay will have little impact on results obtained with the iSYS platform. This article is part of a Special Issue entitled ‘17th Vitamin D Workshop’. ã 2014 Elsevier Ltd. All rights reserved.
Keywords: 25-hydroxy vitamin D Harmonization Standardization
Contents 1. 2.
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 Materials and methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 2.1. Specimen samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
* Corresponding author at: Department of Internal Medicine, Section of Endocrinology, Yale School of Medicine, P.O. Box 208020, New Haven, CT 065208020, United States. Tel.: +1 203 785 5452; fax: +1 203 785 6462. E-mail addresses: [email protected] (C.A. Simpson), [email protected] (A.M. Cusano), [email protected] (J. Bihuniak), [email protected] (J. Walker), [email protected] (K.L. Insogna). http://dx.doi.org/10.1016/j.jsbmb.2014.09.013 0960-0760/ ã 2014 Elsevier Ltd. All rights reserved.
42
3. 4. 5.
C.A. Simpson et al. / Journal of Steroid Biochemistry & Molecular Biology 148 (2015) 41–46
2.2. Methodology . . . . . 2.3. Reference materials Data analysis . . . . . . . . . . Results . . . . . . . . . . . . . . . Discussion . . . . . . . . . . . . . Acknowledgements . . . . . References . . . . . . . . . . . .
.. . .. .. .. .. ..
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
1. Introduction Interest in the measurement of 25-hydroxy vitamin D (25(OH) vitamin D) continues to increase as does the controversy over how to interpret these data and their relevance to health. For example, the Institute of Medicine, after an extensive review of available data, recently recommended that circulating levels of 25(OH) vitamin D above 20 ng/ml (50 nmol/L) be considered sufficient for the general population despite a gradual acceptance over the past two decades that the lower limit of sufficiency should be 30 ng/mL (75 nmol/L) [1]. In addition, there remains considerable interest and disagreement about the impact of 25(OH) vitamin D concentration on non-skeletal conditions such as cardiovascular disease, diabetes, obesity, neurologic syndromes, and cancers [2]. Consequently, issues of method variability and the lack of an accepted reference standard have assumed greater urgency [3]. In a collaborative effort to address this problem, the Vitamin D Standardization Program (VDSP) was established in 2010 by the National Institutes of Health (NIH) Office of Dietary Supplements, the U.S. Centers for Disease Control and Prevention, the U.S. National Institute of Standards and Technology (NIST) and the Belgium Laboratory for Analytical Chemistry, Ghent, Belgium [4,5]. Their objective was to promote the use of a reference method procedure (RMP) by assay manufacturers as well as clinical research laboratories [3]. The program also provides assistance to investigators seeking to re-calibrate historical 25(OH) vitamin D research data to the NIST reference material SRM 2972 [6]. Isotope dilution liquid chromatography tandem mass spectrometry (ID-LC MS/MS) has been defined as the accepted RMP. In this procedure the stable isotope, hexadeuterated 25(OH) vitamin D, is added to serum as an internal standard. The serum is then extracted, fractionated on Sephadex and finally subjected to two-dimensional LC–MS/MS [7]. At present, the best way to assess the concordance of different methods for measuring 25(OH) vitamin D is the Vitamin D External Quality Assessment Scheme program known as DEQAS [8]. In DEQAS
[(Fig._1)TD$IG]
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
42 42 43 43 45 46 46
target values are provided by NIST using the validated RMP. Participants submit their results quarterly to DEQAS where in all platforms can be compared against the NIST validated target values. It is hoped that with harmonization of clinical assays to the NIST standard the degree of inter-method variability will be lessened [7]. Immunodiagnostic Systems Ltd., offers a chemiluminescent assay for 25(OH) vitamin D on its iSYS platform, which has recently been harmonized to the RMP. Using single donor human serum samples with RMP target values, supplied by the VDSP, IDS calibrated their master human serum samples library. This library was used to value-assign the internal calibrators to define the assay's master curve and kit lot-specific calibrators. The predefined master curve is adapted to the analyzer using kit lot-relevant calibrators. In the current study the effect of harmonization on the performance of the IDS iSYS 25(OH) vitamin D assay was assessed by comparing results obtained with non-harmonized (or nonstandardized) and harmonized (or standardized) assay calibrators. 2. Materials and methods 2.1. Specimen samples One hundred nineteen previously analyzed single donor serum samples, with concentrations of 25(OH) vitamin D ranging from 6 to 100 ng/mL (15–250 nmol/L) were chosen for study from the clinical material available in the Yale University Mineral Metabolism Laboratory. Samples were de-identified and aliquoted. The deidentified samples were given a separate numerical identifier in accordance with Yale University Human Investigation Committee regulations. All samples were stored at 70 C until analyzed using nonstandardized (NSC) and standardized calibrators (SC) and the IDSiSYS auto analyzer. Of the 119 single donor serum samples eightynine of these samples were selected based on disease category. These categories were selected to represent disorders where vitamin D metabolites are frequently measured, including, osteoporosis/osteopenia, chronic kidney disease stages III–IV, type 2 diabetes, obesity, multiple sclerosis, hypercalcemia, hypocalcemia and breast cancer. 2.2. Methodology Measurements of 25(OH) vitamin D were performed using a chemiluminescent-based assay in the IDS-iSYS auto analyzer, Boldon, England, U.K. Non-standardized reagent cartridges and calibrators (IS-1600) and calibration verifiers (IS-2735) were purchased from Immunodiagnostic Systems Ltd., Gaithersburg, MD, USA. Standardized reagent cartridges and calibrators (S25OHD) and calibration verifiers (IS-2735S) were provided by the manufacturer. 2.3. Reference materials
Fig. 1. Comparison of 25(OH) vitamin D results (ng/mL) obtained in a nonstandardized assay (NSC 3) and a standardized assay (SC 1) n = 119. Spearman's coefficient r = 0.975 (0.963–0.983, 95% CI).
DEQAS unknowns 451–455 (April 2014 distribution) were obtained from DEQAS, Charring Cross Hospital London, England, U. K. Target values for these unknowns are provided by NIST.
C.A. Simpson et al. / Journal of Steroid Biochemistry & Molecular Biology 148 (2015) 41–46
43
Table 1 Spearman's rank correlation of 25(OH) vitamin D measurements generated from four non-standardized and two standardized assays. Coefficients above 0.7 indicate a strong relationship. NSC = non-standardized assays, SC = standardized assays. n = 119. SC 1
NSC NSC NSC NSC
1 2 3 4
SC 2
Spearman's coefficient (r)
95% confidence interval
Spearman's coefficient (r)
95% confidence interval
0.969 0.961 0.975 0.972
0.956–0.979 0.944–0.973 0.963–0.982 0.960–0.981
0.963 0.966 0.973 0.972
0.946–0.974 0.950–0.976 0.960–0.981 0.959–0.981
3. Data analysis Statistical analyses were performed using GraphPad Prism version 6 software for Mac (GraphPad Software, La Jolla, California, USA). Shapiro–Wilk was used to test for normality. Because not all data were normally distributed Spearman's rank correlation analysis was performed to evaluate the relationship between non-standardized and standardized assays. For instances where data were simply categorized as using non-standardized or standardized calibrators, or bifurcated by 25(OH) vitamin D ranges, a square root transformation was applied and simple linear regression analyses were performed. All but one of the six data sets (one of the non-standardized data sets) were normally distributed after transformation. Bland–Altman plots were generated by subtracting the results obtained using standardized calibrators from the results obtained using non-standardized calibrators compared to the mean value obtained in both assays. A Wilcoxon's signed rank test was used to compare the results for the analysis of DEQAS unknowns using assays employing nonstandardized and standardized calibrators by examining the mean percent difference in the two assays for each of the five unknowns. A probability level 0.05 indicated statistical significance. 4. Results Fig. 1 illustrates a Spearman's rank correlation between one IDS-iSYS assay using non-standardized calibrators (NSC 3) and one in which standardized calibrators (SC 1) were used. The correlation coefficient was 0.975 (0.963–0.983, 95% CI). Table 1 summarizes the correlations between three additional non-standardized and two standardized assays. In all cases the correlations were extremely strong. Table 2 shows the best-fit slopes generated by regressing the two standardized assays against the four nonstandardized assays. For all eight analyses there were strong and statistically significant associations between the non-standardized and standardized assays as reflected by the R2 values and the 95% CIs. None of the slopes were found to be significantly different from one another. Fig. 2 shows Bland–Altman plots for results between the four non-standardized and two standardized calibrated assays. In general the bias was comparable when each of the NSC assays was compared to SC 1 vs. SC 2.
As reported recently by Cavalier et al. [9] there may be greater variability in 25(OH) vitamin D measurements in selected disease states, presumably because of matrix effects, the presence of various metabolites of vitamin D or the amount of vitamin D binding protein present. We therefore analyzed our samples based on diagnosis. Although the number of disease-specific samples analyzed in the current study was relatively small, we observed strong and statistically significant associations between the results obtained using standardized and non-standardized assays for eight different clinical conditions (Table 3). The only exceptions were hypocalcemia (NSC 1, NSC 3 and NSC 4). These data likely reflect the small sample sizes. Controversy continues over whether 20 or 30 ng/mL (50 or 75 nmol/L) represents the lower level of the 25(OH) vitamin D sufficient range and whether values above 50 ng/mL (125 nmol/L) are safe and efficacious [1,10]. Another debate has risen regarding the efficacy and safety of maintaining levels of 25(OH) vitamin D above 50 ng/mL (125 nmol/L) [11,12]. Cashman et al. [6] reported that high 25(OH) vitamin D levels showed greater variability in results obtained by LC–MS/MS pre and post harmonization. We therefore separated our results into two data sets representing vitamin D levels from 6 to 49 ng/mL (15–122 nmol/L) and 50 to 100 ng/mL (125–250 nmol/L). The correlations in Table 4 show strong and highly significant associations between results obtained with non-standardized and standardized calibrators regardless of the absolute value albeit, like Cashman, the correlations for the higher values were not quite as strong. When the data in Table 4 were normalized by square root transformation, linear regression analyses of results obtained with the four non-standardized calibrators and the two standardized calibrators, all regressions were highly significant. The slopes of the regression equations in the 6–49 ng/mL pool (15–122 nmol/L) ranged between 0.919 and 1.090 with a mean value of 1.03 0.06. The slopes in the 50–100 ng/mL pool (125–250 nmol/L) ranged between 0.624 and 0.933 with a mean value of 0.825 0.11 (data not shown). The data summarized in Fig. 3 represent results from the five DEQAS unknowns (451–455) provided to participants in April 2014 distribution. The unknowns were analyzed in one nonstandardized and one standardized assay. When the mean percent difference from the NIST target in values obtained using standardized vs. non-standardized calibrators was analyzed by
Table 2 Simple linear regression of 25(OH) vitamin D results generated from four non-standardized and two standardized assays. SC 1
NSC NSC NSC NSC
1 2 3 4
SC 2
Slope
95% CI slope
R2
P value
Slope
95% CI slope
R2
P value
0.899 0.024 0.951 0.025 0.985 0.022 0.924 0.022
0.850–0.946 0.900–1.001 0.943–1.028 0.881–0.967
0.925 0.923 0.947 0.939
Contents lists available at ScienceDirect
Journal of Steroid Biochemistry & Molecular Biology journal homepage: www.elsevier.com/locate/jsbmb
Review
Effect of 25(OH) vitamin D reference method procedure (RMP) alignment on clinical measurements obtained with the IDS-iSYS chemiluminescent-based automated analyzer Christine A. Simpson * , Anna Maria Cusano, Jessica Bihuniak, Joanne Walker, Karl L. Insogna Department of Internal Medicine, Section of Endocrinology, Yale School of Medicine, 330 Cedar Street, New Haven, CT 06520-8020, United States
A R T I C L E I N F O
A B S T R A C T
Article history: Received 5 August 2014 Accepted 12 September 2014 Available online 18 September 2014
The Vitamin D Standardization Program (VDSP) has identified ID-LC/MS/MS as the reference method procedure (RMP) for 25(OH) vitamin D and NIST Standard SRM2972 as the standard reference material (SRM). As manufacturers align their products to the RMP and NIST standard, a concern is that results obtained in aligned assays will be divergent from those obtained with pre-alignment assays. The Immunodiagnostic Systems Ltd., chemiluminescent, 25(OH) vitamin D iSYS platform assay, was recently harmonized to the RMP. To determine the impact of standardization on results obtained with iSYS reagents, 119 single donor serum samples from eight different disease categories were analyzed in four non-standardized and two standardized iSYS assays. There were strong correlations between the four non-standardized and two standardized assays with Spearman's rank r values between 0.975 and 0.961 and four of the eight r values were >0.97. R2 values for the eight best-fit linear regression equations ranging between 0.947 and 0.916. None of the slopes were found to be significantly different from one another. Bland–Altman plots showed that the bias was comparable when each of the four nonstandardized assays was compared to either of the standardized assays. When the data were segregated in values between 6 and 49 ng/mL (15–122 nmol/L) or between 50 and 100 ng/mL (125–250 nmol/L) significant associations remained between results obtained with non-standardized and standardized calibrators regardless of the absolute value. When five recent DEQAS unknowns were analyzed in one non-standardized and one standardized assay the mean percent difference from the NIST target in values obtained using standardized vs. non-standardized calibrators were not significantly different. Finally, strong and statistically significant associations between the results were obtained using nonstandardized and standardized assays for six of eight clinical conditions. The only exceptions were hypocalcemia and breast cancer, which likely reflect the small sample sizes for each of these diseases. These initial data provide confidence that the move to a NIST standardized assay will have little impact on results obtained with the iSYS platform. This article is part of a Special Issue entitled ‘17th Vitamin D Workshop’. ã 2014 Elsevier Ltd. All rights reserved.
Keywords: 25-hydroxy vitamin D Harmonization Standardization
Contents 1. 2.
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 Materials and methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 2.1. Specimen samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
* Corresponding author at: Department of Internal Medicine, Section of Endocrinology, Yale School of Medicine, P.O. Box 208020, New Haven, CT 065208020, United States. Tel.: +1 203 785 5452; fax: +1 203 785 6462. E-mail addresses: [email protected] (C.A. Simpson), [email protected] (A.M. Cusano), [email protected] (J. Bihuniak), [email protected] (J. Walker), [email protected] (K.L. Insogna). http://dx.doi.org/10.1016/j.jsbmb.2014.09.013 0960-0760/ ã 2014 Elsevier Ltd. All rights reserved.
42
3. 4. 5.
C.A. Simpson et al. / Journal of Steroid Biochemistry & Molecular Biology 148 (2015) 41–46
2.2. Methodology . . . . . 2.3. Reference materials Data analysis . . . . . . . . . . Results . . . . . . . . . . . . . . . Discussion . . . . . . . . . . . . . Acknowledgements . . . . . References . . . . . . . . . . . .
.. . .. .. .. .. ..
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
1. Introduction Interest in the measurement of 25-hydroxy vitamin D (25(OH) vitamin D) continues to increase as does the controversy over how to interpret these data and their relevance to health. For example, the Institute of Medicine, after an extensive review of available data, recently recommended that circulating levels of 25(OH) vitamin D above 20 ng/ml (50 nmol/L) be considered sufficient for the general population despite a gradual acceptance over the past two decades that the lower limit of sufficiency should be 30 ng/mL (75 nmol/L) [1]. In addition, there remains considerable interest and disagreement about the impact of 25(OH) vitamin D concentration on non-skeletal conditions such as cardiovascular disease, diabetes, obesity, neurologic syndromes, and cancers [2]. Consequently, issues of method variability and the lack of an accepted reference standard have assumed greater urgency [3]. In a collaborative effort to address this problem, the Vitamin D Standardization Program (VDSP) was established in 2010 by the National Institutes of Health (NIH) Office of Dietary Supplements, the U.S. Centers for Disease Control and Prevention, the U.S. National Institute of Standards and Technology (NIST) and the Belgium Laboratory for Analytical Chemistry, Ghent, Belgium [4,5]. Their objective was to promote the use of a reference method procedure (RMP) by assay manufacturers as well as clinical research laboratories [3]. The program also provides assistance to investigators seeking to re-calibrate historical 25(OH) vitamin D research data to the NIST reference material SRM 2972 [6]. Isotope dilution liquid chromatography tandem mass spectrometry (ID-LC MS/MS) has been defined as the accepted RMP. In this procedure the stable isotope, hexadeuterated 25(OH) vitamin D, is added to serum as an internal standard. The serum is then extracted, fractionated on Sephadex and finally subjected to two-dimensional LC–MS/MS [7]. At present, the best way to assess the concordance of different methods for measuring 25(OH) vitamin D is the Vitamin D External Quality Assessment Scheme program known as DEQAS [8]. In DEQAS
[(Fig._1)TD$IG]
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
42 42 43 43 45 46 46
target values are provided by NIST using the validated RMP. Participants submit their results quarterly to DEQAS where in all platforms can be compared against the NIST validated target values. It is hoped that with harmonization of clinical assays to the NIST standard the degree of inter-method variability will be lessened [7]. Immunodiagnostic Systems Ltd., offers a chemiluminescent assay for 25(OH) vitamin D on its iSYS platform, which has recently been harmonized to the RMP. Using single donor human serum samples with RMP target values, supplied by the VDSP, IDS calibrated their master human serum samples library. This library was used to value-assign the internal calibrators to define the assay's master curve and kit lot-specific calibrators. The predefined master curve is adapted to the analyzer using kit lot-relevant calibrators. In the current study the effect of harmonization on the performance of the IDS iSYS 25(OH) vitamin D assay was assessed by comparing results obtained with non-harmonized (or nonstandardized) and harmonized (or standardized) assay calibrators. 2. Materials and methods 2.1. Specimen samples One hundred nineteen previously analyzed single donor serum samples, with concentrations of 25(OH) vitamin D ranging from 6 to 100 ng/mL (15–250 nmol/L) were chosen for study from the clinical material available in the Yale University Mineral Metabolism Laboratory. Samples were de-identified and aliquoted. The deidentified samples were given a separate numerical identifier in accordance with Yale University Human Investigation Committee regulations. All samples were stored at 70 C until analyzed using nonstandardized (NSC) and standardized calibrators (SC) and the IDSiSYS auto analyzer. Of the 119 single donor serum samples eightynine of these samples were selected based on disease category. These categories were selected to represent disorders where vitamin D metabolites are frequently measured, including, osteoporosis/osteopenia, chronic kidney disease stages III–IV, type 2 diabetes, obesity, multiple sclerosis, hypercalcemia, hypocalcemia and breast cancer. 2.2. Methodology Measurements of 25(OH) vitamin D were performed using a chemiluminescent-based assay in the IDS-iSYS auto analyzer, Boldon, England, U.K. Non-standardized reagent cartridges and calibrators (IS-1600) and calibration verifiers (IS-2735) were purchased from Immunodiagnostic Systems Ltd., Gaithersburg, MD, USA. Standardized reagent cartridges and calibrators (S25OHD) and calibration verifiers (IS-2735S) were provided by the manufacturer. 2.3. Reference materials
Fig. 1. Comparison of 25(OH) vitamin D results (ng/mL) obtained in a nonstandardized assay (NSC 3) and a standardized assay (SC 1) n = 119. Spearman's coefficient r = 0.975 (0.963–0.983, 95% CI).
DEQAS unknowns 451–455 (April 2014 distribution) were obtained from DEQAS, Charring Cross Hospital London, England, U. K. Target values for these unknowns are provided by NIST.
C.A. Simpson et al. / Journal of Steroid Biochemistry & Molecular Biology 148 (2015) 41–46
43
Table 1 Spearman's rank correlation of 25(OH) vitamin D measurements generated from four non-standardized and two standardized assays. Coefficients above 0.7 indicate a strong relationship. NSC = non-standardized assays, SC = standardized assays. n = 119. SC 1
NSC NSC NSC NSC
1 2 3 4
SC 2
Spearman's coefficient (r)
95% confidence interval
Spearman's coefficient (r)
95% confidence interval
0.969 0.961 0.975 0.972
0.956–0.979 0.944–0.973 0.963–0.982 0.960–0.981
0.963 0.966 0.973 0.972
0.946–0.974 0.950–0.976 0.960–0.981 0.959–0.981
3. Data analysis Statistical analyses were performed using GraphPad Prism version 6 software for Mac (GraphPad Software, La Jolla, California, USA). Shapiro–Wilk was used to test for normality. Because not all data were normally distributed Spearman's rank correlation analysis was performed to evaluate the relationship between non-standardized and standardized assays. For instances where data were simply categorized as using non-standardized or standardized calibrators, or bifurcated by 25(OH) vitamin D ranges, a square root transformation was applied and simple linear regression analyses were performed. All but one of the six data sets (one of the non-standardized data sets) were normally distributed after transformation. Bland–Altman plots were generated by subtracting the results obtained using standardized calibrators from the results obtained using non-standardized calibrators compared to the mean value obtained in both assays. A Wilcoxon's signed rank test was used to compare the results for the analysis of DEQAS unknowns using assays employing nonstandardized and standardized calibrators by examining the mean percent difference in the two assays for each of the five unknowns. A probability level 0.05 indicated statistical significance. 4. Results Fig. 1 illustrates a Spearman's rank correlation between one IDS-iSYS assay using non-standardized calibrators (NSC 3) and one in which standardized calibrators (SC 1) were used. The correlation coefficient was 0.975 (0.963–0.983, 95% CI). Table 1 summarizes the correlations between three additional non-standardized and two standardized assays. In all cases the correlations were extremely strong. Table 2 shows the best-fit slopes generated by regressing the two standardized assays against the four nonstandardized assays. For all eight analyses there were strong and statistically significant associations between the non-standardized and standardized assays as reflected by the R2 values and the 95% CIs. None of the slopes were found to be significantly different from one another. Fig. 2 shows Bland–Altman plots for results between the four non-standardized and two standardized calibrated assays. In general the bias was comparable when each of the NSC assays was compared to SC 1 vs. SC 2.
As reported recently by Cavalier et al. [9] there may be greater variability in 25(OH) vitamin D measurements in selected disease states, presumably because of matrix effects, the presence of various metabolites of vitamin D or the amount of vitamin D binding protein present. We therefore analyzed our samples based on diagnosis. Although the number of disease-specific samples analyzed in the current study was relatively small, we observed strong and statistically significant associations between the results obtained using standardized and non-standardized assays for eight different clinical conditions (Table 3). The only exceptions were hypocalcemia (NSC 1, NSC 3 and NSC 4). These data likely reflect the small sample sizes. Controversy continues over whether 20 or 30 ng/mL (50 or 75 nmol/L) represents the lower level of the 25(OH) vitamin D sufficient range and whether values above 50 ng/mL (125 nmol/L) are safe and efficacious [1,10]. Another debate has risen regarding the efficacy and safety of maintaining levels of 25(OH) vitamin D above 50 ng/mL (125 nmol/L) [11,12]. Cashman et al. [6] reported that high 25(OH) vitamin D levels showed greater variability in results obtained by LC–MS/MS pre and post harmonization. We therefore separated our results into two data sets representing vitamin D levels from 6 to 49 ng/mL (15–122 nmol/L) and 50 to 100 ng/mL (125–250 nmol/L). The correlations in Table 4 show strong and highly significant associations between results obtained with non-standardized and standardized calibrators regardless of the absolute value albeit, like Cashman, the correlations for the higher values were not quite as strong. When the data in Table 4 were normalized by square root transformation, linear regression analyses of results obtained with the four non-standardized calibrators and the two standardized calibrators, all regressions were highly significant. The slopes of the regression equations in the 6–49 ng/mL pool (15–122 nmol/L) ranged between 0.919 and 1.090 with a mean value of 1.03 0.06. The slopes in the 50–100 ng/mL pool (125–250 nmol/L) ranged between 0.624 and 0.933 with a mean value of 0.825 0.11 (data not shown). The data summarized in Fig. 3 represent results from the five DEQAS unknowns (451–455) provided to participants in April 2014 distribution. The unknowns were analyzed in one nonstandardized and one standardized assay. When the mean percent difference from the NIST target in values obtained using standardized vs. non-standardized calibrators was analyzed by
Table 2 Simple linear regression of 25(OH) vitamin D results generated from four non-standardized and two standardized assays. SC 1
NSC NSC NSC NSC
1 2 3 4
SC 2
Slope
95% CI slope
R2
P value
Slope
95% CI slope
R2
P value
0.899 0.024 0.951 0.025 0.985 0.022 0.924 0.022
0.850–0.946 0.900–1.001 0.943–1.028 0.881–0.967
0.925 0.923 0.947 0.939
