The UK National Institute for Health and Care Excellence has recommended thresholds for total vitamin B12 (B12) assays with interchangeable results for identifying B12 deficiency. We assessed the agreement between B12 methods, to evaluate whether the thresholds may be assumed applicable to all assays. A total of 19 External Quality Assessment (EQA) exercises (1791 determinations) based on human subjects-pool materials and 97 serum samples were retrieved to characterize the agreement between Alinity and Architect (Abbott Diagnostics), Access DXi (Beckman Coulter Diagnostics), Atellica and ADVIA Centaur (Siemens Healthcare Solution) compared to Cobas (Roche Diagnostics), considered as comparator because its calibrator traceability to the World Health Organisation (WHO) International Standard (IS) code 03/178. Ordinary least squares and Bland-Altman were used for this purpose. Abbott and Beckman methods overestimated and underestimated, respectively, B12 concentrations vs. Roche and the other methods. We reported similar systematic or proportional error patterns between EQA and serum samples. Only Beckman was affected by both errors. Due to the wide Limit of Agreement Interval, we cannot confidently conclude on the agreement between Roche and the other methods. However, the inter-method bias was well lower than the desirable goal of 9.4 % for Abbott Architect and also lower for Siemens ADVIA Centaur. The recommended thresholds for serum total B12 should not be assumed applicable to all assays, due to the poor agreement among the currently available methods, a limitation that persists despite the release of the WHO IS 03/178.
Cesana et al. (Tue,) studied this question.