Identifying and resolving analytical outliers in lipoprotein(a) immunoassay testing

OBJECTIVES: Lipoprotein(a) [Lp(a)] is associated with increased risk of atherosclerotic cardiovascular disease. Lp(a) is commonly measured using latex-enhanced immunoassays, which are susceptible to optical, reagent, and matrix interferences. To resolve sporadic, discrepant results (i.e., outliers) observed in testing for Lp(a), the following analyses and experiments were performed. DESIGN AND METHODS: A dataset from a sampling period in which duplicate Lp(a) testing was performed on clinical specimens [Roche Tina-quant Lp(a) Gen.2 assay on c503 (Roche) instruments] was retrieved. Additional Lp(a) experiments included evaluation of the Roche assay on c503 instruments and Randox and Diazyme Lp(a) assays on both c503 and AU5800 (Beckman Coulter) instruments. Outliers were defined based on the average of original and repeat results: absolute differences of greater than or equal to ± 6 mg/dL (±15 nmol/L) for averages < 30 mg/dL (<75 nmol/L), or percent differences of ≥ 20% for averages ≥ 30 mg/dL (≥75 nmol/L). Replicate pairs where one or both results were outside the analytical measuring range (AMR) were excluded. RESULTS: Of 26,637 specimens for which duplicate results were extracted, 20,832 (78.2%) had both original and repeat results within the AMR. 0.2% of these pairs (n = 49) were classified as outliers. Clinically, discrepancies were resolved by additional replicate testing. Assay optimization using Diazyme reagent was conducted to ensure thorough onboard specimen pipetting and sample/reaction mixing and washing, yielding decreases in the number of outliers. CONCLUSIONS: Immunoturbidimetric Lp(a) assays are subject to outliers that are minimized with optimization of reagent/sample mixing and wash steps.