The dog stops at the wrong side of the door. She has opened that door a thousand times, nosed it, leaned on it, waited by it, and now she stands at the hinge and waits for something to happen. Her owner fills in a questionnaire at the vet's office, a validated instrument called the Canine Cognitive Dysfunction Rating scale, and a tube of blood goes off for testing. Years later, after she has gone, her brain is cut and stained in a laboratory in Colorado and read by a neuropathologist who usually reads human ones. That sequence, antemortem score and blood draw first, then the tissue itself, is the whole point of a paper published on 6 October 2026 in npj Dementia by Abdullatif Alsulami, Julie A. Moreno and colleagues at Colorado State University, the University of Washington and the University of Arizona, working with the Dog Aging Project Consortium.
The researchers took postmortem brain tissue from 24 client-owned senior dogs, donated after euthanasia and necropsy with owner consent, and applied the staging systems that define Alzheimer's disease in humans: Thal phases for the regional spread of amyloid beta, Braak-style regional assessment for phosphorylated tau (pTau181), and CERAD criteria for neuritic plaques. Seven regions were sampled for amyloid, including the middle frontal gyrus, primary visual cortex, transentorhinal and entorhinal cortex, hippocampus, neostriatum, midbrain and cerebellum. A region counted as positive if even one amyloid-positive plaque turned up. Nobody had done this systematically in dogs before; behavioural decline in aging dogs is well documented, but the map of what is actually happening in the tissue, drawn to human specifications, was missing.
What the dogs showed was an early chapter rather than a whole book. Amyloid deposition was consistent with early Thal-like regional involvement. Phosphorylated tau was present but variable and regionally restricted. And neuritic plaques, the dense, degenerating-neurite-laden lesions that CERAD scoring exists to count and that sit at the centre of the human diagnosis, were absent. Dogs with worse cognition carried more amyloid than dogs that were cognitively intact, but when the authors looked for what best predicted regional amyloid, age won. Age correlated moderately with plaque burden in the visual cortex (r = 0.57, p = 0.003) and not significantly in the prefrontal cortex (r = 0.35, p = 0.08), while the owner-reported cognitive scores themselves showed no significant correlation with amyloid burden in either region. In an earlier preprint version of the work, posted to Research Square in March 2026, the team reported tau pathology in 71 percent of the animals, with 17 percent showing atypical patterns that did not follow the orderly hippocampus-to-neocortex march Braak staging assumes.
The blood is where this gets interesting for people. Of the plasma markers measured while the dogs were alive, glial fibrillary acidic protein, GFAP, a marker of reactive astrocytes, was positively correlated with amyloid plaque burden in the brain. The others were not. GFAP is one of the blood markers now being studied hard in human Alzheimer's clinics, and here is an animal that was never engineered to get this disease, that developed it on its own schedule, sharing a home and an air supply and a sofa with a human family, in which a blood marker tracks tissue pathology confirmed at autopsy. That is the kind of validation a model has to earn, and most rodent models, built from human mutations inserted into a mouse, cannot earn it in the same way.
For a family reading this with a diagnosis in the house, the honest translation is modest and it is not about your dog's medication or your own. It is this: the companion dog is being carefully measured against the human yardstick rather than assumed to fit it, and the measurement shows both a real resemblance and a real gap. The resemblance is amyloid that accumulates with age in predictable places and a blood marker that sees it. The gap is the missing neuritic plaques and the patchy, out-of-order tau. If dogs stall at an early stage of the same process, they may turn out to be the best available window on the years before symptoms, which is exactly the window in which human treatment is now aimed, and exactly the window that is hardest to study in people because almost nobody donates a brain at that point. The authors are explicit that what they built is a comparative framework, not a validated canine staging system, and twenty-four animals, assessed once, cannot carry more weight than that.