The between-day CVs were between 5.1% and 12% at D 2 , D 3 , and total D concentrations of 12.9 -143 ng/mL (32.2-358 nmol/ L). The matrix effect was negligible. The linearity intervals for D 3 , D 2 , and total D were 3.0 -200 ng/mL (7.5-500 nmol/L), 1.0 -150 ng/mL (2.5-375 nmol/L), and 3.0 -350 ng/mL (7.5-875 nmol/L), respectively. The recoveries were 83.8%-109.2% (mean, 95.9%). The lower limits of quantification for D 3 and D 2 were 3.0 ng/mL (7.5 nmol/L) and 1.0 ng/mL (2.5 nmol/ L), respectively. We tested the 2011 College of American Pathologists proficiency testing set ABVD-01-05 with the assay, and all results were within Ϯ25% of target values (mean bias, Ϫ1.2%). Interference by 3-epi-25-OH-D 3 was negligible, because our study included only samples from adults (21-97 years) (3 ).We randomly chose samples from 149 adult patients submitted for 25-OH-D testing by the Abbott Architect assay. All samples were tested by the LC-MS/MS assay (63 contained D 2 ), and subsets of these samples were aliquoted for the fol- For the Liaison assay, the 2 scenarios yielded similar coefficients of determination; however, these values were different for all of the other assays. The regression slopes and r 2 values of the Architect and Cobas assays both de-
Letters to the Editor1274 Clinical Chemistry 59:8 (2013) creased when D 2 -containing samples were included. For the Centaur assay, on the other hand, the regression slope increased, and the r 2 decreased. The mean (SD) changes in bias from scenario (a) to (b) for the 4 assays are as follows: Liaison assay, Ϫ17.4% (6.0%) to Ϫ19.2% (4.6%); Architect assay, 0.9% (5.2%) to Ϫ7.8% (4.2%); Cobas assay, Ϫ11.4% (11.9%) to Ϫ15.5% (7.0%); Centaur assay, Ϫ27.7% (7.0%) to Ϫ17.0% (8.0%).The effect of D 2 % on assay bias is illustrated in Fig. 1B. The Liaison assay had an insignificant bias change as D 2 % increased. The Architect and Cobas assays showed an increasingly negative bias with increasing D 2 %, whereas the negative bias of the Centaur assay decreased and changed to a positive bias.We calculated D 2 crossreactivity normalized to D 3 reactivity (see Fig. 1 legend). The values for each assay in Although D 3 is the preferred supplement form for treating vitamin D deficiency (4, 5 ), we have observed a high prevalence of patient samples containing substantial amounts of D 2 , and we have demonstrated the different effects of D 2 content on the accuracies of commercial assays. Therefore, correctly interpreting total 25-OH-D results and monitoring patient compliance requires that vitamin D status, type of supplementation, and measurement method be considered.