Habit-learning and decision-making circuits are susceptible to glycemic variability in type 2 diabetes: a...

Type 2 diabetes mellitus (T2DM) is associated with lower gray matter (GM) volumes. However, little is known about the impact of glycemic control on brain atrophy, especially in highly susceptible regions. Therefore, we aim to identify the effect of glycemic variability (GV) on long-term changes in brain volume among individuals with T2DM.

A longitudinal clinical, biochemical, and imaging assessment was conducted at a baseline visit on 170 individuals (85 with T2DM), from which 29 (15 with T2DM) were evaluated at a 7-year follow-up visit. Brain regional volumes were evaluated with 3 T MRI, using the FreeSurfer 7 longitudinal pipeline. GV metrics such as SD, M-value, MAG (mean absolute glucose change), MAGE (mean amplitude of glycemic excursion), and CoV (coefficient of variation) were calculated in both visits.

Statistically significant negative correlations between GV metrics and symmetrized percent change (SPC) of GM volumes were found in specific cortical and subcortical regions of individuals with T2DM. MAGE was correlated with regionally specific atrophy on the temporal lobe (r = -0.63, p = 0.021), insula (ρ = -0.62, p = 0.022), thalamus (r = -0.64; p = 0.024), hippocampus (r = -0.59; p = 0.034), and putamen (ρ = -0.65, p = 0.017). Concerning the hippocampal subregions, the presubiculum was significantly correlated with MAGE (r = -0.73; p = 0.005). Baseline GV was consistently associated with temporal lobe SPC. Linear regression analysis showed that, for each increase of 1 mmol/L in MAGE value, the SPC of the temporal lobe decreases on average by 1.2% (higher atrophy rate).

The relationship between longitudinal GM atrophy and GV has a regionally specific pattern, suggesting localized brain susceptibility to intra-daily glucose fluctuations. Negative correlations between GV metrics and SPC volume of regions involved in habit-learning, decision-making, and memory highlight GV as a mediator of the neural impact of T2DM on the reward prediction-error circuits.