Commutability of two reference materials for two commercial lithium assays

is ensured by the availability in the JCTLM database of two secondary reference materials (RM), i.e. the NIST SRM 956c (electrolytes in frozen human serum) and the IRMM BCR-304 (lyophilised human serum). As these RM are intended for use in the calibration and traceability validation of commercial systems, information on their commutability is central. Here we tested the commutability of these RM using two commercial methods measuring serum lithium, based on direct potentiometry (Ion-Selective Electrodes Direct, Roche Cobas Integra 400) and on a colorimetric approach (Multigent Lithium, Abbott Architect c16000), respectively. Methods: A total of 27 leftover human serum samples were collected, aliquoted and stored at -80 °C until their use. We measured lithium concentrations with the two systems in each biological sample, in SRM 956c (3 levels) and in BCR-304 in duplicate in two different runs on the same day. Manufacturer’s control materials were used to validate analytical runs. The commutability of RM was estimated from Deming regression analysis of the measured results in native samples using the 95% prediction interval (95PI) and multiples of the standard error of regression (Sy-x), in accordance with the CLSI C53-A standard. Results: The SRM 956c results did not fall inside the 95PI based on the results for the native clinical samples. In addition, using an acceptance criterion for commutability of ±3 times the experimental Sy-x (±0.066), the relative residuals (rr) for SRM 956c (−2.956 level 1, −4.044 level 2, −3.209 level 3) were all outside the acceptable range. On the other hand, BCR- 304 results fall inside the 95PI, but its rr (−0.197) was not within the acceptable range. Conclusions: Our results show that SRM 956c was not commutable between the methods evaluated. BCR-304 showed better, although not perfect, commutability and should be preferred to align lithium assays to higher-order references. The uncertainty of BCR-304 (2.9%) is however relatively high and this may become an issue for fulfilling the goal of acceptable uncertainty of lithium measurement for clinical laboratories (±4.3%), as calculated assuming a time interval between doses of 12h and a drug average half-life of 24h.