AbstractIntroductionThe identification of early, preferably presymptomatic, biomarkers and true etiologic factors for Alzheimer’s disease (AD) is the first step toward establishing effective primary and secondary prevention programs. Consequently, the search for a relatively inexpensive and harmless biomarker for AD continues. Despite intensive research worldwide, to date there is no definitive plasma or blood biomarker indicating high or low risk of conversion to AD.MethodsMagnetic resonance imaging and β -amyloid (Aβ) levels in three blood compartments (diluted in plasma, undiluted in plasma and cell-bound) were measured in 96 subjects (33 with mild cognitive impairment, 14 with AD and 49 healthy controls). Pearson correlations were completed between 113 regions of interest (ROIs) (45 subcortical and 68 cortical) and Aβ levels. Pearson correlation analyses adjusted for the covariates age, sex, apolipoprotein E (ApoE), education and creatinine levels showed neuroimaging ROIs were associated with Aβ levels. Two statistical methods were applied to study the major relationships identified: (1) Pearson correlation with phenotype added as a covariate and (2) a meta-analysis stratified by phenotype. Neuroimaging data and plasma Aβ measurements were taken from 630 Alzheimer’s Disease Neuroimaging Initiative (ADNI) subjects to be compared with our results.ResultsThe left hippocampus was the brain region most correlated with Aβ(1 - 40) bound to blood cell pellets (partial correlation (pcor) = − 0.37, P = 0.0007) after adjustment for the covariates age, gender and education, ApoE and creatinine levels. The correlation remained almost the same (pcor = − 0.35, P = 0.002) if phenotype is also added as a covariate. The association between both measurements was independent of cognitive status. The left hemisphere entorhinal cortex also correlated with Aβ(1 - 40) cell-bound fraction. AB128 and ADNI plasma Aβ measurements were not related to any brain morphometric measurement.ConclusionsAssociation of cell-bound A β(1 - 40) in blood with left hippocampal volume was much stronger than previously observed in A β plasma fractions. If confirmed, this observation will require careful interpretation and must be taken into account for blood amyloid-based biomarker development.