A 2014 meta-analysis attempts to answer this question. Here is their relevant plot:
One conclusion: AD-related pathology (neuritic plaques and NFTs) has a stronger relationship with dementia and cognitive impairment (CI) than vascular-related pathology.
Interesting article from Mapstone et al on using blood levels of metabolites and LASSO analysis to predict progression from normal cognition to MCI/AD in 2-3 years. Of an initial set of 1000’s of metabolites, they focused in on these 10, which had lower blood levels in the subset of people who would eventually go on to develop MCI/AD:
- Phosphatidylcholine diacyls aa C36:6, aa C38:0, aa C38:6, aa C40:1, aa C40:2, PC aa C40:6, and ae C40:6
- Lysophophatidylcholine a C18:2
- Propionyl Acylcarnitines C3 and C16:1-OH
t Supplemental Figure 4; NC = Normal Control, doi:10.1038/nm.3466
As you can see, their relative levels validated well on an independent sample. Strange, though, that the predictive ability is better for normal controls vs preconverters (AUC 0.92) than it is for normal controls vs MCI/AD patients (AUC 0.77), although that might be due to statistical power considerations.
All of these are lipids and changes in their relative levels are probably associated with cell membrane damage. It’s hard to draw too much mechanistic insight into AD pathogenesis from the fact that they are lipids, however, without knowing precisely why these 10 metabolites were chosen. E.g., if it was due to convenience of sampling (certainly a valid criterion) or if they had some of the highest independent predictive ability.