4 popular vitamin D testers gave very different results (no surprise)

Comparison of four routinely used vitamin D automated immunoassays

J Med Biochem . 2021 Jun 5;40(3):277-285. doi: 10.5937/jomb0-27531.

Jindra Windrichova 1, Pavel Broz 1, Radka Fuchsova 2, Ondrej Topolcan 1, Ladislav Pecen 1, Otto Mayer 3, Radek Kucera 1

1 University Hospital Pilsen, Department of Immunochemistry Diagnostics, Czech Republic.

2 University Hospital Pilsen, Institute of Clinical Biochemistry and Hematology, Czech Republic.

3 University Hospital Pilsen, Second Internal Clinic, Czech Republic.

1. This study measured the Vitamin D level from 100 blood samples on 4 different testers * Vitamin D testing accuracies, including dried blood spot – Jan 2020 * Problems with Vitamin D Testing – chapter – Aug 2019 * Vitamin D deficiency of a group - 15 pcnt to 48 pcnt (depends on tester used) - Nov 2017 * Big differences (~10 ng) in vitamin D test results, even when using the same type of tester – Aug 2015 * * Huge differences in % deficient (< 30ng) depending on the tester used** * * Vitamin D measurements vary from the same blood sample – March 2014 **poor repeatability * * 1. # Opinion: Due to both accuracy and repeatability variations a Vitamin D test should never be trusted to be within 5 ng 1. Tests for Vitamin D contains the following overview/opinion {include} --- Note: For many years there has been an attempt at standardizing Vitamin D measurments used for research It appears that most hospitals/testing facilities are not "standardized" * *📄 Download the Standarization PDF from VitaminDWiki * *

📄 Download the PDF from VitaminDWiki

Background: To compare four automated immunoassays for the measurement of 25(OH)-vitamin D (25-OHD) and to assess the impact on the results obtained from a healthy population.

Methods: We analysed 100 serum samples on Unicel DxI 800 (Beckman Coulter), Architect i1000 (Abbott), Cobas e411 (Roche) and Liaison XL (DiaSorin). Passing-Bablok regression and Bland-Altman plots were used for method comparison. In order to categorise the obtained values, results were categorised into the following groups: 0-25 nmol/L, 25-50 nmol/L, 50-75 nmol/L and above 75 nmol/L and compared. The percentage of samples below 75 nmol/L, and below 50 nmol/L was then calculated for every method.

Results: According to paired comparisons, each method differs from others (p<0.0001) except Cobas vs Architect, which do not show a statistically significant difference (p=0.39). The strongest correlation was found between Liaison and Architect (ρ=0.94, p<0.0001).

The percentage of samples below the recommended value of 75 nmol/L were:

The percentage of samples below the value of 50 nmol/L were:

Conclusions: The observed differences stem from the use of different analytical systems for 25-OHD concentration analysis and can result in different outcomes. The recommended values should be established for each assay in accorda