Head‐to‐head comparison of CSF AD biomarker assays: fully‐automated high‐throughput platforms (ECLIA and CLEIA) vs ELISA

Abstract

AbstractBackgroundBiomarkers in cerebrospinal fluid (CSF) can help to diagnose Alzheimer’s disease (AD) and classify patients according to the amyloid/tau/neurodegeneration (ATN) research framework.MethodWe directly compared the performance of three different testing systems, i..e. traditional solid‐phase sandwich ELISA assays (Innotest®, Fujirebio Europe N.V., Gent, Belgium, and IBL, Minneapolis, USA), a fully automated electrochemiluminescence immunoassays (ECLIA) platform (Elecsys® CSF assays measured on the cobas e 411 analyzer, Roche Diagnostics International Ltd, Rotkreuz, Switzerland), and a fully automated chemiluminescent enzyme immunoassays (CLEIA) platform (Lumipulse® G600II, Fujirebio Europe N.V., Gent, Belgium), in 50 early AD patients (64±8.1y) and 25 cognitively unimpaired (CU) controls (65±9.4y)] from the same cohort. Individual CSF biomarkers Aβ42, Aβ40, pTau181 and tTau were measured using all systems. Ratios for pTau181/Aβ42, tTau/Aβ42, Aβ42/Aβ40, pTau181/Aβ42/Aβ40, tTau/Aβ42/Aβ40, pTau181/Aβ40, and tTau/Aβ40 were calculated.ResultAcross all systems, biomarkers and their ratios accurately distinguished AD patients from CU in a comparable manner: ROC: AUCs >0.96 (upper bound of CI:1.0)], except for ELISA Aβ42/Aβ40 [AUC = 0.948 (CI: 0.901‐0.995)]; according to Youden Index (YI)‐derived optimal cut‐offs, sensitivity and specificity were comparably high. YI cut‐offs and manufacturer (MA) cut‐offs were highly consistent. However, for the ECLIA platform (Elecsys®), the YI cut‐off led to significantly higher sensitivity for tTau (97.8% [0.88‐1.00%]/ 80.9%) and pTau181 (97.8% [0.88‐1.00%]/ 83.0%), and higher specificity for Aβ42 (96.0% [79‐100%]/ 76.0%), than MA cut‐off, maybe due to high prevalence of AD. The two fully automated platforms showed markedly reduced maximal coefficients of variation (CV) than ELISA.ConclusionIn addition to increased time efficiency, fully automated platforms showed more consistent results than ELISA assays which favor single analyte measurements. Improved accuracy using YI cut‐offs suggests that in‐house cut‐offs might be important for optimal performance. The two fully automated platforms had comparable accuracy. Thus, the platform of choice might depend on the user’s needs; for instance, the pTau181/Aβ42 ratio (Elecsys®; (AUC: 1.0 [1.0‐1.0], sensitivity 100%, specificity 100%)) combines the two AD hallmarks (A and T) in one test. For ATN classification, the Aβ42/Aβ40 ratio (Lumipulse®, AUC: 1.0 [1.0‐1.0], sensitivity 100%, specificity 100%) might be a more accurate marker for amyloid positivity (A+) than Aβ42 alone.