Limitations of platform assays to measure serum 25OHD level impact on guidelines and practice decision making

Abstract

Context: Liquid Chromatography Mass Spectroscopy (LC-MS/MS) is the preferred method to measure 25 hydroxyvitamin D (25OHD) levels, but laboratories are increasingly adopting automated platform assays.

Objective: We assessed the performance of commonly used automated immunoassays, with that of LC-MS/MS, and the National Institute of Standards and Technology (NIST) reference values, to measure 25OHD levels.

Methods/setting: We compared serum 25OHD levels obtained from 219 elderly subjects, enrolled in a vitamin D trial, using the Diasorin Liaison platform assay, and the tandem LC-MS/MS method. We also assessed the performance of the Diasorin and Roche automated assays, expressed as mean % bias from the NIST standards, based on the vitamin D External Quality Assessment Scheme (DEQAS) reports, from 2013 to 2017.

Results: Serum 25OHD levels were significantly lower in the Diasorin compared to LC-MS/MS assay at baseline, 18.5 ± 7.8 vs 20.5 ± 7.6 ng/ml (p < 0.001), and all other time points. Diasorin (25OHD) = 0.76 × LC-MS/MS (25OHD) + 4.3, R2 = 0.596. The absolute bias was independent of 25OHD values, and the pattern unfit for any cross-calibration. The proportion of subjects considered for vitamin D treatment based on pre-set cut-offs differed significantly between the 2 assays. There also was wide variability in the performance of both automated assays, compared to NIST reference values.

Conclusion: The performance of most widely used automated assays is sub-optimal. Our findings underscore the pressing need to re-consider current practices with regard to 25OHD measurements, interpretation of results from research studies, meta-analyses, the development of vitamin D guidelines, and their relevance to optimizing health.

Trial registration: ClinicalTrials.gov NCT01315366.

Conflict of interest statement

All authors state that they have no conflict of interest

Copyright © 2018 Elsevier Inc. All rights reserved.

Figures

Figure 1

Panel A: Bland-Altman plot of the baseline difference between Diasorin Liaison and LC-MS/MS 25-OHD levels vs. baseline LC-MS/MS 25OHD in 219 subjects (symbols). Solid horizontal line, mean difference between the two assays (or bias); dashed horizontal lines, 95% confidence interval. Panel B. Bland-Altman plot of the baseline percentage difference [(LC-MS/MS - Diasorin Liaison)/LC-MS/MS*100]. Solid horizontal line, mean difference between the two assays (or bias); dashed horizontal lines, 95% confidence interval.

Figure 2. Proportion (Panel A) or frequency (Panel B) of subjects with 25OHD levels in the total cohort (n 516) above or at indicated assay specific 25OHD level.

Diasorin Liaison (gray) and the LC-MS/MS (black) assays.

Figure 3. Comparison of Diasorin Liaison vs. LC-MS/MS 25OHD levels in the overall data set (n 516) subjects.

Symbols represent individual data points using both assays. Gray light circles represent concordant points at increasing 25OHD cut-offs [A (20 ng/ml). Dashed line, Line of identity; solid line, linear regression model fit with its equation.

Figure 4:

Frequency distribution of the percentage difference in 25-OHD levels in the overall data set 516 subjects.

Figure 5

(A and B): Method specific mean % bias from NIST lowest (panel A) and highest (panel B) 25OHD) target values, for the 2 most commonly used platform assay methods/manufacturers and for LC-MS/MS methods. The number of laboratories varied by DEQAS report cycle, as shown in legend, depending on the cycle. Figure 5 (C and D): Method specific % bias from NIST lowest (Panel C) and highest (Panel D) 25OHD target values, at 2 AUB laboratories. The mean lowest 25(OH)D NIST target value (panels A and C) obtained by deriving the mean using the lowest 25(OH)D NIST target value for each cycle, over the DEQAS report period 2013–2017: 33.6 ± 8.1 nmol/L (13.4±3.2 ng/ml). Mean highest 25(OH)D NIST target value (panels B and D) obtained by deriving the mean using the highest 25(OH)D NIST target value for each cycle, over the report period 2013–2017: 100.07 ± 16.5 nmol/L (40.0±6.6 ng/ml).