AbstractBackgroundThe cognitive reserve (CR) hypothesis was proposed to explain inter‐individual differences in the association between observed cognitive function and estimated neurodegenerative burden, e.g. in Alzheimer’s disease (AD). Latent CR marker approaches were suggested as a promising method to predict reserve.MethodAmyloid‐β (Aβ)‐negative, cognitively normal and Aβ‐positive, cognitive impaired (Clinical Dementia Rating global score >= 0.5) individuals participating in the AD Neuroimaging Initiative (ADNI) were classified as healthy controls (HC, N=138) and AD spectrum (ADS, Baseline N=462, Follow‐Up N=211), respectively. A latent CR marker (CRM) was obtained by multilinear regression analysis in the entire study cohort, defined as the residual global cognitive performance remaining after accounting for neuropathological burden (CSF Aß42 and total Tau levels and APOE ε4 carrier status) and demographic variables. Mean standardized uptake value ratio (SUVR) values for regions of interests (ROIs) of left and right angular gyrus (L/R‐ANG), bilateral posterior cingulate (PCC), left and right inferior temporal gyrus (L/R‐TEMP) were derived from FDG‐PET data. In separate analyses for HC and ADS, the interactive effects of the memory cognitive composite (MEM) and latent CR marker on FDG‐PET SUVR values were tested for the five cortical ROIs at baseline and 24‐months follow‐up data.ResultIn ADS, higher CRM attenuated the relationship between MEM and FDG‐PET SUVR in L‐TEMP, R‐TEMP and L‐ANG at baseline, and in L‐TEMP, R‐TEMP, L‐ANG and R‐ANG in the longitudinal analyses. In HC, however, no interaction was observed.ConclusionPatients with higher CR maintained their memory performance better in the face of AD‐related glucose hypometabolism . Therefore, the present study supports the use of a residual CRM to improve our understand of inter‐individual differences in the resilience against neurodegeneration, both cross‐sectionally and longitudinally.