Evaluation of two automated immunoassays for 25-OH vitamin D: comparison against LC-MS/MS

Abstract

Background/methods: A total of 95 human serum specimens, and a 12 specimen precision panel, were measured by 2 automated immunoassays (investigation use only DiaSorin LIAISON(®) 25 OH Vitamin D TOTAL Assay [LSN], and Siemens ADVIA Centaur(®) Vitamin D Total (VitD) assay [Centaur]) and the results compared against LC-MS/MS [LCMS] used as the reference method (Esoterix Inc.). For functional sensitivity and precision, 12 serum specimens [range 1.2-148ng/mL] were run in six replicates [N=30] or four replicates [N=20], respectively, for 5 consecutive days.

Results: Passing-Bablok fit and Difference plot analysis [N=92] showed that although both immunoassays had comparable correlation coefficient [r] values to LCMS (0.936 and 0.933), the LSN assay results were statistically equivalent to those given by LCMS (slope 0.93, intercept -2.5), whereas the results of the Centaur assay showed overall significant assay bias compared to LCMS (slope 1.30, intercept -15.8) and this bias was more significant for doses <30ng/mL by LCMS [bias -30.4%; 95% limits of agreement -72.4% to 11.7%]. For specificity, based on 25-OHD2 and 25-OHD3 levels assessed by LCMS, we divided the specimens into 2 groups, one with detectable 25-OHD2 [Group 1, N=41] and the other with no detectible 25OHD2 [Group 2, N=51]. The 2 groups showed comparable correlation coefficient [r] values between the methods, but showed significant differences in slope: Centaur [1.48 with group 1 and 1.18 with group 2] compared to LSN [0.91 with Group 1 and 0.96 with Group 2]. LSN demonstrated better precision [total CV range 5.5-10.0%] compared to Centaur [total CV range 11.0-16.3%]. Functional sensitivity was calculated per EP-17A: 2.15ng/mL by LSN and 4.57ng/mL by Centaur.

Conclusions: Though there was good overall correlation, substantial bias was present in Centaur. Although LSN had a slope and intercept that was not significantly different than LCMS, Centaur had a significantly higher slope in specimens containing measurable 25-OHD2 levels, a large negative intercept and a significant negative dose bias for doses <30ng/mL by LCMS, suggesting the Centaur assay would report a higher frequency of patients with apparent vitamin D insufficiency/deficiency at the low end and apparent vitamin D toxicity at the high end compared against LCMS. This article is part of a Special Issue entitled 'Vitamin D Workshop'.