Serum samples from 74 obese women were assayed for 25-hydroxyvitamin D [25(OH)D] concentrations using an automated immunoassay [Architect (Abbott)] and ELISA (Alpco Diagnostics), and results were compared with the LC/MS/MS reference method (Quest Diagnostics). The Architect values were significantly lower (mean difference: -13 nmol/L; 95% limits: -54; 28 nmol/L) and the ELISA values were significantly higher (mean difference: 24 nmol/L; 95% limits: -36; 84 nmol/L) than the LC/MSIMS values. The slope of the Passing-Bablok regression line relative to LC/MSIMS was 0.5 [95% confidence interval (CI): 0.41; 0.60] and 1.17 (95% CI: 0.87; 1.56) for Architect and ELISA, respectively, with an intercept of approximately 16 for both assays. Using 50 nmol/L as the cut-point for deficiency, Architect and ELISA misclassified 20 and 27% of the subjects, respectively. In subjects with low circulating 25-hydroxylated ergocalciferol [25(OH)D2] (<10 nmol/L), a Bland-Altman plot and Kappa statistics (Kappa = 0.73; 95% CI: 0.49-0.97) showed good agreement between Architect and LC/MS/MS. However, in subjects with high circulating 25(OH)D2 (>or=10 nmol/L), Architect demonstrated poor agreement (Kappa = 0.40; 95% CI: 0.16-0.65). ELISA demonstrated a higher degree of overestimation in women with minimal 25(OH)D2 than those with high 25(OH)D2, suggesting that ELISA overestimates 25-hydroxylated cholecalciferol [25(OH)D3] but underestimates 25(OH)D2.