A blood test must work for the people it will reach

# A blood test must work for the people it will reach OCTOBER 15, 2026 | LONGEVITY SPECIAL A blood test for Alzheimer's disease may work about equally well across racial groups at flagging amyloid buildup in the brain, but the typical level of the marker differs from group to group, which raises a question about whether one cutoff can fit everyone. That cutoff question is our own inference, not a tested result. A diagnostic marker becomes more useful when its performance is understood across the populations who may receive it. This analysis brings racial and ethnic variation into that evaluation. The finding comes from a cross-sectional study in JAMA Neurology (2026). That is an observational design: it looks at people at one point in time and can show links, not cause and effect. The researchers analysed screening data from 15,612 adults aged 55 to 80 who were being screened for a trial called AHEAD 3-45. The data were collected from July 2020 to October 2024 at 106 sites across four world regions. The trial targets preclinical Alzheimer's disease. That means amyloid, a protein that builds up in the brain in Alzheimer's, is present, but there are no cognitive symptoms yet. The blood test measures the percentage of a form of tau protein called p-tau217. The study asked whether that percentage differs by race and ethnicity, and whether its link to thinking scores and to amyloid on brain scans (PET scans) differs too. Levels did differ. According to the study, %p-tau217 was lower in Asian, Black and Hispanic participants than in White and non-Hispanic participants. Participants were grouped by self-reported race (Asian, Black or White) and ethnicity (Hispanic or non-Hispanic). Carrying the APOE ε4 gene variant, a known Alzheimer's risk factor, was linked to higher %p-tau217 in every group. The size of that link differed by race and by ethnicity, according to the authors' statistical tests. Of the participants, 5,755 (36.9%) were APOE ε4 carriers. The link between the blood marker and amyloid on PET scans showed no significant pairwise differences across racial groups. The same was true for the link with a cognitive score called the PACC5. The authors say this consistency supports using a blood test first, with a PET scan to confirm. Ethnicity told a more complicated story. The links between the blood marker and both the cognitive score and the PET scans differed between Hispanic and non-Hispanic participants. The authors say these ethnicity-related findings need further investigation. The abstract does not explain why the differences appear. For people without symptoms, the practical meaning is limited. This study does not show that a blood result predicts who will develop dementia. It looked at screening for a trial, not at health outcomes. The abstract also does not give the specific cutoffs used, so it cannot say how many people would be misclassified under a single threshold. The study's own framing is that levels vary across groups, and that this may influence who is eligible for trials. A fixed cutoff applied to groups with different typical levels could, in principle, treat them differently. That is where the argument is weakest: the abstract does not test group-specific cutoffs or report how any cutoff performed, so the concern remains a reasoned possibility rather than an established problem. Researchers will need to test whether group-specific or adjusted thresholds improve accuracy, and to follow up the ethnicity findings in further studies. The task is to establish thresholds and interpretations that work reliably. Broader representation can improve the tool itself, as well as the reach of its benefits. Source: JAMA Neurology (2026) JAMA Neurology (2026): https://doi.org/10.1001/jamaneurol.2026.3598 ---