In a cohort of 990 participants from the Bio-Hermes-001 study, plasma phosphorylated tau-217 (pTau-217) was measured using two platforms—a Lilly immunoassay and a University of Gothenburg mass spectrometry assay. Gaussian mixture modeling identified three latent biomarker classes in both assays, with monotonically increasing rates of amyloid-PET positivity and clinical Alzheimer's diagnosis across classes. PET-anchored thresholds (90% specificity for amyloid-β PET) demonstrated higher sensitivity but lower specificity than distribution-based thresholds (mean + 2 SDs of the lowest latent class). Agreement between methods was only moderate—Cohen's kappa of 0.678 and 0.575 for the two assays—with disagreement concentrated almost entirely in the intermediate biomarker class.
This finding carries real clinical weight. Plasma pTau-217 has rapidly emerged as one of the most promising blood-based Alzheimer's biomarkers, potentially enabling scalable screening without costly PET imaging. Yet this preprint—not yet peer-reviewed—highlights a underappreciated problem: the choice of positivity threshold matters more than which laboratory platform you use. The intermediate class, where disagreement clusters, likely represents the early or transitional amyloid accumulation phase—precisely the population where early intervention has the greatest theoretical benefit. Clinicians and trialists adopting pTau-217 for screening or enrollment decisions must recognize that threshold methodology is not a neutral technical detail but a substantive clinical choice that shapes who gets flagged, monitored, or treated. This work is incremental but practically important, and independent validation in more diverse cohorts is needed before these findings influence guideline development.
