Short-interval variability and test-retest reliability of Alzheimer’s disease plasma biomarkers in a clinically heterogenous cohort

BACKGROUND: Assessment of short-term within-subject variability and test-retest reliability of Alzheimer's disease (AD)-related plasma biomarkers is important for real-world clinical implementation and interpretation in research studies. In this study we aimed to assess short-term variability and reliability of core AD plasma biomarkers, proposed ratios, and evaluate a broader proteomic panel. METHODS: Two serial plasma samples were collected from each of the 92 participants enrolled in observational research studies over intervals ranging from 30 minutes to three months. Test-retest reliability and within-subject variability were assessed for 131 biomarkers measured on three immunoassays (Lumipulse G plasma p-tau217; Fujirebio, NULISAseq CNS Disease Panel and NULISAqpcr AD 5-Plex; Alamar Biosciences). We subsequently evaluated the stability of a dual cut-point approach for AD probability diagnostic categorisation and examined whether within-subject variability was associated with inter-sample interval (time between serial blood sampling). RESULTS: All core AD plasma biomarkers assessed in this study (Aβ42, BD-p-tau181, BD-p-tau217, BD-p-tau231, p-tau181, p-tau217, p-tau231, GFAP and NfL) demonstrated strong correlations (r ≥ 0.79) and high reliability (ICC ≥ 0.84). Within-subject variability was moderate to low for core AD plasma biomarkers. Using a validated dual cut-point approach for plasma p-tau217, most individuals (92%) remained within the same diagnostic category (high/indeterminate/low) across both timepoints, and no individual moved across two boundaries (i.e. from high to low AD probability categorisation or vice versa). For the majority of analytes, within-subject variability was not associated with inter-sample interval after adjustment for age and sex, however, the core AD plasma biomarker, NfL did have a significant association. CONCLUSIONS: While core AD plasma biomarkers had limited variability and high reliability over short-intervals, other central nervous system (CNS)-relevant biomarkers showed greater variability, which may be relevant for interpretation in association studies and clinical applications.