A smell test and a blood test are converging on early Alzheimer's

The nose knows first

The earliest signal may be the cheapest one to measure. In a multi-cohort preprint study of cognitively normal older adults, those who tested positive for amyloid scored significantly lower on odor identification than those who tested negative. Each 10-percentage-point increase in odor identification score was associated with a 16% lower hazard of developing mild cognitive impairment. That makes a smell test a candidate screening step, not a diagnosis — a cheap way to decide who needs the expensive test. m:…1025

The size of the effect is what makes it interesting. A 16% lower hazard per 10 points of odor score is not a subtle statistical whisper; it is large enough to sort people into risk groups, which is exactly what a screening step needs to do. The test itself — identifying odors — needs no blood draw, no scanner, no lab. The question the paper leaves open is not whether the signal exists but whether it holds up the same way in every cohort, and on that point the authors found trouble: the association between parahippocampal tau burden and odor identification was inconsistent across cohorts. m:…1025

A blood protein sits in the middle of the cascade

A second preprint looks at plasma GFAP, a blood protein, and places it squarely in the middle of the amyloid cascade. Plasma GFAP statistically mediates the association between soluble amyloid-beta 42/40 and amyloid PET scans, and between amyloid PET and tau PET. The protein networks associated with GFAP are enriched for extracellular matrix and immune-related pathways — the structural and immune context of the disease, not just the amyloid itself. m:…4395

Mediation, in plain terms, means GFAP sits on the path between the earlier and later stages: the soluble amyloid signal reaches the PET scan partly through GFAP, and the PET signal reaches tau partly through GFAP. That makes it more than a bystander marker — it is a measurable step in the chain, and a blood test that reads it is reading the cascade itself rather than a distant echo of it. The enrichment for extracellular matrix and immune pathways adds a second point: the cascade runs through the brain's structural and immune environment, not through amyloid alone. m:…4395

Proteins in plasma map the stages

A third study connects plasma proteins to what brain scans show. Machine learning on plasma proteomic signatures classified Alzheimer's stages with accuracy between 0.836 and 0.963 by area under the curve, and specific proteins tracked specific brain changes: higher plasma cortisol was associated with lower volumes in the left precuneus and superior parietal regions when comparing mild cognitive impairment with Alzheimer's, higher osteopontin with lower left paracentral and accumbens volumes in the same comparison, and higher vitronectin with lower posterior corpus callosum volume when comparing cognitively normal adults with Alzheimer's. Blood is starting to read like a rough map of the brain. m:…9383

The pattern of the three proteins matters as much as the headline accuracy. Cortisol, osteopontin and vitronectin do not all point at the same place; each is tied to a different region, and the comparison that matters shifts with the stage — mild cognitive impairment against Alzheimer's for two of them, cognitively normal against Alzheimer's for the third. That is the shape of a staging tool rather than a single yes-or-no test: a blood draw that says not just whether the disease is present, but roughly where in its course a person sits. m:…9383

What is still shaky

None of this is settled. The link between parahippocampal tau burden and odor identification was inconsistent across cohorts in the smell study, and the GFAP paper's interaction findings were not consistently replicated across its two cohorts. The direction is consistent — earlier, cheaper signals — but any single marker still needs to survive replication before it can be trusted in the clinic. m:…1025 m:…4395