HYPERGLYCEMIA AND NEURAL NETWORK FUNCTIONAL CONNECTIVITY: PEAKS AND THEN TROUGHS IN MID- TO LATE LIFE FOR PRECLINICAL AD AND ACROSS THE AD SPECTRUM

In normal aging, pre- and type 2 diabetes (“diabetes”) are associated with less functional connectivity within and between AD-sensitive neural networks, which may be an early AD marker. To date, however, metabolic dysfunction and intrinsic neural network functional connectivity have not been systematically examined in preclinical AD, or both MCI and AD, compared to normal aging. Data were downloaded from 13,186 UK Biobank subjects (47-81y) with no memory impairment at baseline, and 157 aged subjects from ADNI at 5 visits spanning 2 years. Mean intrinsic functional connectivity was computed and Z-scored for Group ICA derived orthogonalized networks including default mode, central executive, emotion regulation, salience, and recognition memory networks. Mixed model regression tested the main effect of diabetes vs. no diabetes and 2-way interactions with AD parental family history or Apolipoprotein E ε4 (APOE4) status. Further interactions with age in years were explored. Similar ADNI models were tested with log-transformed fasting glucose and genetic risk factors or baseline clinical diagnosis (normal, Mild Cognitive Impairment or MCI, AD). Interactions with change over Time were explored. In UK Biobank, AD Parental History or APOE4 x Diabetes x Age interactions were seen for all networks (p<.05 to p<.01, corrected). For example, in default mode network (Figure 1), among adults with no AD parental history (Figure 1, left graph), having diabetes (“red” bar) vs. no diabetes (“gray” bar) was linked to less network connectivity regardless of age. By contrast, in adults with AD parental history (Figure 1, right graph), diabetes in middle vs. old age was related to 1.5SD more vs. 2SD less network connectivity. In ADNI, using default mode as an example, higher glucose in APOE4 carriers predicted less intrinsic functional connectivity (R2=0.208, p<.001). A Diagnosis x Glucose interaction (p<.05, Figure 2) showed that higher glucose was, over 2 years, consistently related to more neural network connectivity in stable MCI and MCI progressors to AD, but less network connectivity in AD.