Non-linearity in lipase assays: A multicentric comparison on different analysers
Résumé
OBJECTIVES: Non-linearity in lipase assays and the ensuing gaps in results distribution have been described on Roche analysers, but have yet to be studied on other analysers. DESIGN AND METHODS: Eighteen lithium-heparinized plasma pools of lipase activities decreasing from 1700 to \textless4 U/L were prepared for multicentric evaluation on several analysers. Non-linearity was modelled as the difference between the polynomial regression of lipase activities depending on relative dilutions over the primary measuring range, and the linear regression of the same variables above the manufacturer's limit of linearity (MLL). Gaps in lipase distribution resulting from non-linearity were graphically evidenced through histograms. Upper limits of gaps were calculated, which are lipase activities where non-linearity biases no longer impact the diluted lipase results. RESULTS: MLLs and lipase (U/L) calculated at MLL (%biases versus MLL) were respectively: 1200 and 1124 (-6.3%) on the Architect C16000 (Abbott); 300 and 248 (-17.3%) on the Cobas c503 (Roche); 1500 and 1458 (-2.8%) on the Dimension Vista (Siemens); and 700 and 659 (-5.9%) on the Atellica CH930 (Siemens). Using Sentinel Lipase reagents on Abbott analysers, these measurements were respectively: 300 and 294 (-2.0%) on the Architect C16000, and 300 and 298 (-0.7%) on the Alinity. Setting Randox Lipase reagents on the Alinity, MLL and lipase at MLL were 953 and 776 (-18.6%), respectively. CONCLUSIONS: Considering the desirable (±14.2 %) and optimal (±7.1 %) allowable total error for lipase (EFLM/EuBIVAS), biases at manufacturer's limit of linearity were acceptable, except for Roche Cobas c503 method and Randox method on Abbott Alinity.