Clinical Chemistry 53: 1104-1108, 2007.
First published April 13, 2007; 10.1373/clinchem.2006.077560
(Clinical Chemistry. 2007;53:1104-1108.)
© 2007 American Association for Clinical Chemistry, Inc.
|
Endocrinology and Metabolism |
Evaluation of a 1,25-Dihydroxyvitamin D Enzyme Immunoassay
Isolde Seiden-Long1 and
Reinhold Vieth1,2,a
1 Department of Laboratory Medicine and Pathobiology, University of Toronto Pathology and Laboratory Medicine, Mount Sinai Hospital, Toronto, Canada.
2 Department of Nutritional Sciences, University of Toronto, Toronto, Canada.
aAddress correspondence to this author at: Pathology and Laboratory Medicine, Mount Sinai Hospital, 600 University Ave., Toronto, Ontario, M5G 1X5 Canada. Fax 416-586-8628; e-mail rvieth{at}mtsinai.on.ca.
 |
Abstract
|
|---|
Background: Radioactive reagents are used in most assays for measurement of 1,25-dihydroxyvitamin D [1,25(OH)2D]. We evaluated a 1,25(OH)2D enzyme immunoassay (EIA) from IDS Ltd. that uses solid-phase immunoextraction and colorimetric detection and compared results to those of the thymus radioreceptor assay (RRA) for 1,25(OH)2D.
Methods: We collected serum samples (n = 145) representing an even distribution (0–200 pmol/L) of 1,25(OH)2D concentrations and Vitamin D External Quality Assessment Scheme (DEQAS) proficiency survey samples from 2004 surveys (n = 15) and stored them at –20 °C. We analyzed all samples with both EIA and RRA methods. We calculated imprecision using 5 QC samples in quadruplicate in each run (n = 6), including both pooled patient material used for QC with the RRA and QC material included in the EIA reagent set. We evaluated calibration stability by analyzing calibrators from different lots on the same plate and determining if calculated sample values drifted significantly.
Results: Deming linear regression between IDS EIA and RRA methods yielded slope 1.25 (95% CI 1.13–1.37), y-intercept –3 (95% CI –18 to 12), R2 = 0.74. DEQAS proficiency survey samples for 2004 were all within 30% of the all-methods-trimmed mean. Imprecision CVs were 12%–16% within-run and 15%–20% between-run.
Conclusions: We find no evidence of inferiority to the classic calf-thymus receptor assay for 1,25(OH)2D and no disadvantage in the results generated by the IDS EIA using samples from the major proficiency survey for 1,25(OH)2D. According to the product insert, however, the IDS EIA underestimates 1,25(OH)2D2 compared with the D3 form.
|
Introduction
|
|---|
Among the many reported metabolites of vitamin D (1), 2 are measured to address clinical questions: 25-hydroxyvitamin D [25(OH)D, calcidiol]1
, representative of vitamin D nutritional status, and 1,25-dihydroxyvitamin D [1,25(OH)2D, calcitriol], the bioavailable hormone involved in the regulation of calcium metabolism.
The measurement of 1,25(OH)2D presents several analytical challenges. The compound circulates at picomole per liter concentrations and is highly lipophilic. Furthermore, the structurally similar metabolic precursor 25(OH)D circulates at nanomole per liter concentrations, making assay specificity a constant analytical concern. Measurement of 1,25(OH)2D has undergone many improvements (2). A radioreceptor assay (RRA) based on the competitive binding of 1,25(OH)2D and tritiated tracer to its nuclear receptor isolated from calf thymus was introduced in the mid-1980s (3). The calf-thymus receptor assay is attractive because it involves purification and binding to the natural receptor for 1,25(OH)2D and adjustment for recovery. Commercial methods for 1,25(OH)2D do not adjust for recovery, and they involve the use of antibodies. Our laboratory has continued to use the calf-thymus receptor method, and is 1 of the last 2 laboratories reporting data with the method to the Vitamin D External Quality Assurance Scheme (DEQAS). Most clinical laboratories who offer 1,25(OH)2D assays use commercially available methods (4). Currently, the most commonly used methods for 1,25(OH)2D quantification are competitive RIAs using a 125I tracer. The methods involve extraction, but they do not adjust for between-sample variation in extraction efficiency. Two commercially available assays sold by DiaSorin and IDS dominate the market. They primarily differ in their respective 1,25(OH)2D extraction methods before assay and in sensitivity to 1,25(OH)2D2. The DiaSorin method uses a solid-phase extraction and silica purification using organic solvents (acetonitrile, methanol, methylene chloride, hexane, and isopropanol) while the IDS method employs a solid phase immunoextraction, avoiding organic solvents, but involving overnight incubation and a 2-day assay procedure. For the 1,25(OH)2D2 metabolite, the DiaSorin and IDS RIA methods state 100% and 91% specificity, respectively.
A new enzyme immunoassay (EIA) for the measurement of 1,25(OH)2D uses solid-phase immunoextraction and colorimetric detection. The manufacturer’s claim that this method has only 39% reactivity of 1,25(OH)2D2 compared with 1,25(OH)2D3 is a concern because vitamin D2 is commonly used clinically. We present our data for comparison of the EIA to our in-house RRA, as well as an evaluation of lot-to-lot calibration stability and potential compliance with external proficiency surveys.
|
Materials and Methods
|
|---|
thymus rra
Calf-thymus receptor assay, involving purification of analyte on Bond Elut C18OH cartridges (Varian) and an internal tritiated calibrator to correct for losses during purification, has been described (5). Instead of an antibody, this method is based on competition for the nuclear 1,25(OH)2D receptor extracted from calf thymus.
ids eia
Reagent sets were donated by IDS Ltd., and we followed manufacturer instructions. We delipidated samples with dextran sulfate and magnesium chloride solution, centrifuged them to pellet debris, and extracted analytes using immunocapsules (100 µL sample/capsule, assayed in duplicate) containing monoclonal antibody to 1,25(OH)2D linked to solid-phase particles in suspension with vitamin D binding protein inhibitor. Immunocapsules were agitated on a rocker-shaker for 90 min at room temperature (18 to 25 °C), washed 3 times with deionized water, and extracted analyte was eluted with 3 successive applications of 150 µL ethanol. Eluates were evaporated under a gentle flow of nitrogen or by SpeedVac evaporator (Savant). We reconstituted calibrators (lyophilized BSA-phosphate buffer containing 1,25(OH)2D and 0.9 g/L sodium azide) immediately before assay or thawed them from frozen. Dry, immunopurified samples were resuspended with assay buffer (BSA-phosphate buffer with 0.9 g/L sodium azide). We added primary antibody solution (sheep anti–1,25(OH)2D in BSA-phosphate buffer with 0.9 g/L sodium azide) to all tubes, and they were incubated overnight (16–20 h) at 2–8 °C. The next day, we applied samples to appropriate wells of the antisheep IgG–linked 96-well microplate. We selected the plate layout to minimize sample positional bias. Calibrators were run in duplicate and placed on opposite ends of the plate. Controls were run in duplicate on half plates and in quadruplicate on full plates and placed on the ends and middle of the plate. All samples were placed so that no replicate was ever placed in an immediately adjoining well. The plate was incubated on an orbital shaker at 18–25 °C for 90 min, then 1,25(OH)2D biotin solution was added to all wells except the substrate blanks, and the plate was incubated on an orbital shaker (500–750 rpm) at 18–25 °C for 60 min. We washed the plate 3 times, added avidin linked to horseradish peroxidase to all wells except the substrate blanks, and incubated the plate at 18–25 °C for 30 min. We repeated the washes and added tetramethylbenzidine substrate to all wells including the substrate blanks. The reaction was developed at 18–25 °C for 30 min and stopped with 0.5 mol/L hydrochloric acid. Plates were read at 450 nm (reference 630 nm) using a microplate reader within 30 min of adding the HCl solution.
We calculated percentage binding (B/B0) of each calibrator, control, and unknown sample as (mean absorbance – mean absorbance substrate blank) x 100/(mean absorbance for 0 calibrator – mean absorbance substrate blank). We prepared a calibration curve by plotting B/B0% on the ordinate against log concentration of 1,25(OH)2D on the abscissa. We fitted a 4-parameter logistic model to the curve and interpolated sample data using GraphPad software.
samples
Serum samples (n = 145) for method comparison were previously assayed by thymus RRA in our laboratory and selected to represent an even distribution of typical (40–140 pmol/L), low (<40 pmol/L), and high (>140 pmol/L) concentrations. We included 5 QC samples in each run (Table 1
) and pooled both patient material routinely used for QC with the in-house thymus RRA and the QC material from the manufacturer (lyophilized human serum containing 1,25(OH)2D and 0.9 g/L sodium azide). DEQAS survey samples included tubes 141–155 from the 2004 survey year. All samples were stored at –20 °C between runs, and freeze-thaw cycles were minimized.
statistical analyses
All statistical analyses were performed using GraphPad Prism software. To compare different lines (Figs. 1
and 2
), we plotted the data as appropriate and determined statistical differences between lines or curves by F-test on the residuals of data points around the calculated line or curve. We determined the limit of detection by triplicate extraction and measurement of 4 serial 2-fold dilutions using patient pool material at low concentration (i.e., <20 pmol/L). We calculated the limit of detection, defined as the upper 95% CI for values at theoretical 0 concentration, based on the SD of residuals (Sy|x).
View larger version (20K):
[in this window]
[in a new window]
|
Figure 1. Comparison of the thymus RRA and IDS EIA assays.
Deming (model II) linear regression analysis of 145 human serum samples measured for 1,25(OH)2D concentrations. (A), each sample was measured once by thymus RRA and in duplicate by IDS EIA (means shown). 95% CI of the slope = 1.128– 1.370 and y-intercept = –18.19 to 12.07. (B), the same data plotted by lot number. The slopes of the 2 lines are significantly different (F-test, P <0.0001). 95% CI of the slope for lot 1 = 1.2828–1.860 and lot 2 = 1.067–1.309. 95% CI of the y-intercept for lot 1 = –29.81 to 30.65 and lot 2 = –19.59 to 11.64.
|
|
View larger version (12K):
[in this window]
[in a new window]
|
Figure 2. Calibrator shift between lots.
Calibrators from both lots were analyzed in parallel on the same plate in the same run. Data are plotted using a 4-parameter logistic curve fit. The curves are significantly different (F-test, P = 0.0135).
|
|
|
Results
|
|---|
Method comparison by Deming linear regression of the EIA and thymus RRA methods yielded slope 1.25 (95% CI 1.13–1.37), y-intercept –3 (95% CI –18 to 12), R2 = 0.74 (Fig. 1A
). We were initially supplied with 1 lot of EIA reagent, which was divided into 2 runs and showed promising correlation with the thymus RRA method: slope 1.57 (95% CI 1.28–1.86), y-intercept 0.4 (95% CI –30 to 31), R2 = 0.84 (Fig. 1B
). With a 2nd lot of EIA reagents, however, results for in-house QC material values were lower, also corresponding to a significant shift in the slope vs the RRA: slope 1.19 (95% CI 1.07–1.31), y-intercept 0.4 (95% CI –20 to 11), R2 = 0.75 (Fig. 1B
). Troubleshooting efforts eliminated factors such as mixing/reconstitution problems and variability from different drying techniques (SpeedVac concentrator vs nitrogen stream) and ultimately revealed a shift in calibration when calibrators from different lots were run side-by-side on the same plate. The manufacturer subsequently informed us that they had changed their calibration between lots; the curve fits in Fig. 2
demonstrate the curve shift. Additionally, we calculated patient data collected from the same plate using both calibration curves, and paired t-test demonstrated a significant shift in sample values (P <0.0001).
We measured imprecision with 5 QC samples in quadruplicate in each run (n = 6), including both pooled patient material used for QC with the RRA and QC material included with the EIA reagent set. Samples were analyzed once by RRA and twice by EIA in duplicate on the same plate. Duplicates for the EIA method were divided at the time of extraction, as opposed to pipetting duplicate wells from the same sample tube as is commonly used with ELISA techniques not requiring an extraction step. Imprecision (CV) values were 12%–16% within-run and 15%–20% between-run (Table 1
). The cumulative distribution of between-replicate variation (Fig. 3A
) demonstrated a median between-replicate CV of 3.7% (25th percentile 1.8%, 75th percentile 7.3%, 95th percentile 18.6%). The distribution of between-replicate variation as a function of analyte concentration (Fig. 3B
) demonstrates that variability is minimized within the reference interval for 1,25(OH)2D (40–140 pmol/L). To calculate the relative contribution of extraction and EIA to the variability of the method, we repeated analysis of between-replicate variation using calibrators from each run (data not shown). Assuming that total error can be expressed as 
2total = 2extraction + 2EIA, and that we measure total error at mean between-replicate CV, then the total error would be ±5.9%, the EIA assay error would be ±2.4%, and the extraction error would be ±5.4%. Thus, the extraction procedure contributes significantly more to the between-replicate error than the detection protocol of the EIA itself.
View larger version (12K):
[in this window]
[in a new window]
|
Figure 3. Analysis of between-replicate variability.
CV values were calculated from mean difference between replicate measurements performed by IDS EIA. (A), cumulative frequency plot of between-replicate %CV (curve fit: LOWESS). (B), between-replicate %CV as a function of analyte concentration (curve fit: LOWESS).
|
|
The limit of detection, defined as the upper 95% CI for values at theoretical 0 concentration, was determined by use of replicates of extractions at low concentration. The Sy x value was 4.7 nmol/L. Thus, the limit of detection was 9.4 pmol/L, consistent with our practice of reporting low values as <10 pmol/L.
DEQAS was established in 1989 for 25(OH)D and extended in 1997 to include 1,25(OH)2D. DEQAS distributes 5 serum samples 4 times per year to over 200 participants who measure 1,25(OH)2D) in 24 countries worldwide. The current analytical target for 25(OH)D is for 80% of samples to be within 30% of the all-methods-trimmed mean (ALTM) (4). In 2003, 59% of participants met the 25(OH)D target—a somewhat disappointing statistic. There are currently no analytical targets set for 1,25(OH)2D assay performance by DEQAS, but for the purpose of comparison among methods, we used the following criterion for 25(OH)D: 80% of samples to be within 30% of the ALTM (4). Two methods dominate the calculation of the 1,25(OH)2D ALTM: Diasorin RIA (30% of participants) and IDS RIA (59% of participants). IDS EIA performance by single measurement (Fig. 4A
) shows that 14 of 15 samples for 2004 were within 30% of the ALTM. The in-house RRA method also shows that 14 of 15 samples were within 30% for 2004 (Fig. 4
, B–D). EIA performance would be expected to improve if duplicates had been measured, although most laboratories routinely report clinical results based on singleton measurement.
View larger version (41K):
[in this window]
[in a new window]
|
Figure 4. Performance of IDS EIA method for 1,25(OH)2D DEQAS data 2004.
Values (141–155) on ordinate axis represent survey sample vial designations. Shaded area represents ±30% from the ALTM. (A and B), we measured DEQAS samples for 1,25(OH)2D in singleton using IDS EIA and our thymus RRA. (C and D), performance of method mean values for Diasorin RIA and IDS RIA methods as reported for the 2004 DEQAS sample set for 1,25(OH)2D.
|
|
|
Discussion
|
|---|
The IDS EIA and our in-house thymus RRA compare well with other reports about the performance of 1,25(OH)2D assays (6)(7)(8)(9). Although between-run imprecision seems high in the general context of clinical chemistry, it is equally within expected performance characteristics of assays of this analyte. While we observed between-run CVs of 15%–20% in our evaluation, historical DEQAS data show that other methods are unlikely to provide better performance. Between-laboratory CVs in the 2004 DEQAS report were as follows: IDS RIA 11%–26%, Diasorin RIA 12%–30%, and overall between-method 15%–28%. The relative imprecision of this assay in general can in large part be attributed picomolar concentrations of analyte. However, we found that the extraction contributes more variability than the immunoassay component of the procedure.
The shift in calibration was discussed with the IDS technical staff, and they have made assurances that this observation was attributable to an early preproduction lot of reagents being distributed to us for evaluation. Nevertheless, it would be prudent for users to approach lot changes cautiously. This issue may be of less concern for laboratories that intend to use a predetermined number of reagent sets for clinical trials that have a defined number of patients and may not require the long-term analytical stability of a clinical service laboratory. In those cases, reagents from a single lot for a given trial would be optimal. It should be noted that calibrators are not extracted in the EIA method or the RRA method, but the RRA method uses a tracer for monitoring extraction efficiency whereas the EIA method has no such adjustment. This process would putatively make the EIA method more difficult to troubleshoot should there be a discrepancy between lots, as it would have to be determined initially if the problem was with extraction or with calibration.
According to 2003 DEQAS participation data (4), 59% of participants currently use the IDS RIA method for 1,25(OH)2D, which uses the identical immunoextraction method. Equally, the IDS RIA method for 1,25(OH)2D does not include calibrator extraction as part of the protocol. These results indicate that the laboratories currently using the IDS RIA and wishing to switch to a nonradioactive method could do so with relative ease. Our evaluation demonstrates that the IDS EIA would easily meet the DEQAS goal of 80% of survey samples within 30% of the all-methods mean (Fig. 4
).
In theory, the IDS EIA may underestimate 1,25(OH)2D2; however, our data using clinical samples reveal no evidence of inferiority to the classic calf-thymus receptor assay for 1,25(OH)2D, and no disadvantage in the results generated by the IDS EIA using samples from the major proficiency survey for 1,25(OH)2D.
|
Acknowledgments
|
|---|
Grant/funding support: None declared.
Financial disclosures: IDS Ltd. contributed assay reagent sets.
Acknowledgements: We thank Flor Reyes, Syeda Hasani, and Shirley Siu for assay of serum samples by calf-thymus receptor assay.
|
Footnotes
|
|---|
1 Nonstandard abbreviations: 25(OH)D, 25-hydroxyvitamin D; 1,25(OH)2D, 1,25-dihydroxyvitamin D; RRA, radioreceptor assay; DEQAS, Vitamin D External Quality Assessment Scheme; EIA, enzyme immunoassay; ALTM, all-methods-trimmed mean. 
|
References
|
|---|
- Bouillon R, Okamura WH, Norman AW. Structure-function relationships in the vitamin D endocrine system. Endocr Rev 1995;16:200-257.[CrossRef][ISI][Medline]
[Order article via Infotrieve]
- Hollis BW. Detection of vitamin D and its major metabolites. Pike JW Glorieux FH Feldman D eds. Vitamin D 2004:931-950 Academic Press San Diego. .
- Reinhardt TA, Horst RL, Orf JW, Hollis BW. A microassay for 1,25-dihydroxyvitamin D not requiring high performance liquid chromatography: application to clinical studies. J Clin Endocrinol Metab 1984;58:91-98.[Abstract]
- Carter GD, Carter CR, Gunter E, Jones J, Jones G, Makin HL, Sufi S. Measurement of Vitamin D metabolites: an international perspective on methodology and clinical interpretation. J Steroid Biochem Mol Biol 2004;89–90:467-471.[CrossRef]
- Hollis BW, Kilbo T. The assay of circulating 1,25(OH)2D using nonend-capped C18 silica: performance and validation. Norman AW Schaefer K Grigoleit H-G von Herrath D eds. Vitamin D: Molecular, Cellular and Clinical Endocrinology 1988:710-719 W deGruyter Berlin. .
- De Leenheer AP, Bauwens RM. Comparison of a cytosol radioreceptor assay with a radioimmunoassay for 1,25-dihydroxyvitamin D in serum or plasma. Clin Chim Acta 1985;152:143-154.[CrossRef][ISI][Medline]
[Order article via Infotrieve]
- Bertelloni S, Baroncelli GI, Benedetti U, Franchi G, Saggese G. Commercial kits for 1,25-dihydroxyvitamin D compared with a liquid-chromatographic assay. Clin Chem 1993;39:1086-1088.[Abstract/Free Full Text]
- Manolagas SC, Reitz R, Horst R, Haddad J, Deftos LJ. Multicentre comparison of 1,25-dihydroxycholecalciferol measurements in human serum. Lancet 1983;i:191-192.
- Jongen MJ, Van Ginkel FC, van der Vijgh WJ, Kuiper S, Netelenbos JC, Lips P. An international comparison of vitamin D metabolite measurements. Clin Chem 1984;30:399-403.[Abstract]
Top
Abstract
Introduction
