Background: With a 10-year sustained increase in 25-hydroxyvitamin D [25(OH)D] testing, laboratories have swapped their LC-MS/MS methods for high-throughput automated immunoassays. Although it is generally well-known that immunoassays have poor recoveries for 25-hydroxyvitamin D2 [25(OH)D2], the frequency and extent to which this impacts total 25(OH)D have not been previously demonstrated. We evaluated 3 automated immunoassays against the first FDA-cleared CDC/NIST-traceable LC-MS/MS method.
Methods: Method comparison was performed for the Siemens ADVIA Centaur, Roche Elecsys Cobas, and Abbott Architect 25(OH)D immunoassay methods in real patient samples (n = 105). We calculated the mean bias in samples containing >20 ng/mL 25(OH)D2 and estimated the percent 25(OH)D2 cross-reactivities. We determined the prevalence of appreciable concentrations of 25(OH)D2 in our patient population through random sampling (n = 120) and projected the frequency of inaccurate 25(OH)D immunoassay results.
Results: Linear regression for 25(OH)D was y = 1.09x - 4.44 (Centaur), y = 0.84 + 0.43 (Cobas), and y = 0.83x - 0.48 (Architect). The mean biases of 25(OH)D concentrations were 5.6 (11.0) ng/mL (Centaur), -17.5 (7.2) ng/mL (Cobas), and -20.3 (9.8) ng/mL (Architect) in samples containing >20 ng/mL 25(OH)D2. The observed percent cross-reactivities for 25(OH)D2 were 115% (Centaur), 52% (Cobas), and 44% (Architect). We estimate that 8% of our population has >20 ng/mL 25(OH)D2, thereby compromising the accuracy of 25(OH)D results in >3000 samples annually.
Conclusions: We demonstrate that immunoassay manufacturer package inserts indicate much better recoveries of 25(OH)D2 than what is observed in unadulterated real patient samples. We estimate the frequency of inaccurate total 25(OH)D determination by these immunoassay methods to be largely dependent on the concentration of 25(OH)D2 in each sample.
© 2017 American Association for Clinical Chemistry.