Interval-specific likelihood ratios for clinical interpretation of plasma pTau217 across assays

Abstract Background Plasma phosphorylated tau217 (pTau217) is a highly accurate blood-based biomarker for Alzheimer's disease (AD). However, absolute pTau217 concentrations differ substantially between assays, and results are commonly interpreted categorically using one or two assay-specific cut-offs, reducing a continuous biomarker into two or three diagnostic categories and limiting comparison across assays. We documented the performance of the HISCL™ pTau217 assay and valuated whether test result interval-specific likelihood ratios (LRs) could provide a common framework for interpreting plasma pTau217 across assays. Methods A total of 187 participants from the Cognitive Center UZ Ghent (CCUG) biobank were classified as AD (n = 103) or non-AD cognitive disorders (n = 84) using CSF biomarkers. Samples were analysed using 3 pTau217 assays: HISCL™ (Sysmex), Lumipulse G (Fujirebio) and Elecsys (Roche). Assay-specific concentration thresholds were derived from predefined sensitivity and specificity criteria, and interval-specific LRs were calculated for each assay. Results HISCL™ pTau217 showed a high diagnostic accuracy (AUC 0.952), comparable to Lumipulse (AUC 0.950, p-value = 0.931 ) and Elecsys (AUC 0.939, p-value = 0.357 ). Despite substantial differences in absolute concentrations, interval-specific LRs were closely aligned. Using 95% sensitivity and 95% specificity thresholds, LRs were 0.06–0.08 below the lower threshold, 0.40–0.85 for intermediate results, and 15.9–22.5 above the upper threshold. Subdivision by 90%, 95% and 99% specificity showed a stepwise increase, with LR from 0.10–0.13 up to 28.4–29.8, again with close agreement across assays. Conclusion Test result interval-specific LRs allow assay-specific pTau217 concentrations to be expressed on a common scale of diagnostic evidence, enabling comparison across assays while retaining information on result magnitude and allowing integration with the pre-test probability of AD. This proof-of-concept approach complements single- and two-cut-off strategies and warrants validation in larger, multicentre cohorts.