Abstract
Cerebrospinal fluid (CSF) biomarkers can enhance the early and accurate etiologic detection of Alzheimer's disease (AD) even when symptoms are very mild, but are not yet widely available for clinical testing. There are a number of reasons for this, including the need for an experienced operator, the use of instruments mostly reserved for research, and low cost-effectiveness when patient samples do not completely fill each assay plate. Newer technology can overcome some of these issues through automated assays of a single patient sample on existing clinical laboratory platforms, but it is not known how these newer automated assays compare with previous research-based measurements. This is a critical issue in the clinical translation of CSF AD biomarkers because most cohort and clinicopathologic studies have been analyzed on older assays. To determine the correlation of CSF beta-amyloid 1-42 (Aβ42) measures derived from the automated chemiluminescent enzyme immunoassay (CLEIA, on Lumipulse(®) G1200), a bead-based Luminex immunoassay, and a plate-based enzyme-linked immunoassay enzyme-linked immunosorbent assay (ELISA), we analyzed 30 CSF samples weekly on each platforms over 3 weeks. We found that, while CSF Aβ42 levels were numerically closer between CLEIA and ELISA measurements, levels differed between all three assays. CLEIA-based measures correlated linearly with the two other assays in the low and intermediate Aβ42 concentrations, while there was a linear correlation between Luminex assay and ELISA throughout all concentrations. For repeatability, the average intra-assay coefficient of variation (CV) was 2.0%. For intermediate precision, the inter-assay CV was lower in CLEIA (7.1%) than Luminex (10.7%, p = 0.009) and ELISA (10.8%, p = 0.009), primarily due to improved intermediate precision in the higher CSF Aβ42 concentrations. We conclude that the automated CLEIA generated reproducible CSF Aβ42 measures with improved intermediate precision over experienced operators using Luminex assays and ELISA, and are highly correlated with the manual Aβ42 measures.