Biomarkers

In vivo detection of intermediate and high Alzheimer's disease neuropathologic changes (ADNC) is key for disease staging and clinical trial enrolment. While a visually positive tau-PET scan with the FDA/EMA-approved tracer [¹⁸F]flortaucipir can be considered a reliable proxy of high ADNC (Fleisher et al., 2024), plasma biomarkers such as tau phosphorylated at threonine 217 (p-tau217) are emerging as minimally-invasive alternatives. This study evaluates the performance of commercially-available plasma p-tau217 in identifying intermediate and high ADNC defined by neuropathologic assessments at autopsy as well as by visual reads of FDA/EMA-approved in vivo tau-PET scans.

Postmortem neuropathologic evaluations and antemortem plasma p-tau217 and Aβ42/Aβ40 measurements (Lumipulse G1200 assay) from 110 ADNI participants were included. In addition, plasma biomarkers were assessed in 625 additional ADNI participants and for 403 A4 participants plasma p-tau217 (MesoScale ECL assay) was included. All participants underwent [¹⁸F]flortaucipir PET. Three trained readers rated the [¹⁸F]flortaucipir scans as visually positive/negative using the FDA/EMA-approved interpretation method. Receiver operating characteristic (ROC) curve analysis assessed the biomarkers' discriminative accuracy across ADNC levels.

Plasma p-tau217 biomarkers demonstrated high accuracy (AUC>0.93) for discriminating between "no/low/intermediate" and "high" ADNC at autopsy (Figure 1A). When applying externally derived two-tier cut-points for plasma p-tau217 (negative/intermediate/positive), it identified "no" or "high" ADNC accurately (80-93%) (Figure 1B-C). However, for detecting low/intermediate ADNC, high variability was observed, with only 40-65% of the low ADNC cases classified as negative and 35-57% of the intermediate ADNC cases classified as positive (Figure 1C). Findings in the in vivo cohort examined with tau-PET supported these results. Plasma biomarkers reliably identified high ADNC as reflected by the PET-based A+T+ profile, but the majority of the A+T- individuals were classified as intermediate or negative (Figure 2C-E and Figure 3C-E). Plasma biomarkers yielded higher accuracy in cognitively impaired (Figure 3A-B) than in unimpaired individuals (Figure 2A-B).

Plasma p-tau217 markers effectively classify ADNC extremes but are less reliable for accurately assessing low/intermediate ADNC levels, reflecting their limited utility for biological disease staging. This weaker performance in preclinical cases, including failure to reliably identify A+T- cases as positive, challenges their potential role in early AD detection. Complementary imaging may be needed to improve reliability in early stages.