In June 2026, Alzheimer's & Dementia published the ANGMAR cohort study (Fernández-Lebrero et al., Hospital del Mar Research Institute, Barcelona).
The problem: cerebral amyloid angiopathy (CAA) and AD frequently coexist, and mixed pathology changes both clinical progression and treatment risk. CSF is invasive — lumbar puncture is hard to run routinely in a population at high bleeding risk — so there was no non-invasive way to separate "pure CAA" from "CAA with AD" within a CAA cohort.
What they did: 231 participants in four groups (healthy controls, pure CAA, CAA with AD, AD without CAA); plasma and CSF were measured in parallel on the fully automated Lumipulse platform.
Results:
- After adjustment for age and sex, plasma p-tau217 showed a clean stepwise pattern: lowest in healthy controls, only mildly elevated in pure CAA, no meaningful difference between CAA-with-AD and pure AD, with a between-group effect size up to 3.07;
- Using CSF A+T+ as the gold standard, the AUC for detecting AD co-pathology in CAA was 0.920;
- Two prespecified cut-offs: 0.27 pg/mL balances sensitivity and specificity; 0.34 pg/mL is more specific;
- The key negative finding: cerebral microbleeds, cortical superficial siderosis, white matter hyperintensities and the total CAA small-vessel disease score showed no statistically significant association with plasma p-tau217. The marker does not reflect cerebrovascular amyloid burden — it points only to AD-type tau pathology.
What I want to discuss: with two cut-offs, 0.27 and 0.34, if only one could go on the report, which would you pick? The balanced point means a batch of false positives to explain; the higher point means missing the people in the intermediate band — and the paper reports that only a minority of pure CAA patients fall inside it (0.22–0.34).
Note: the paper states the design is cross-sectional observational. It supports association and statistical mediation, not causal ordering.