Acute Hyperglycemia and Spatial Working Memory in Adolescents With Type 1 Diabetes

Research Design and Methods

The primary outcome of this open-label clinical trial (NCT03188757, National Medical Ethics Committee 0120–476/2017/2) was the difference in sWM capacity during acute hyperglycemia compared with euglycemia in adolescents with T1D.

Inclusion criteria were age 11–19 years, duration of T1D 5–10 years, and no severe hypoglycemic episode or diabetic ketoacidosis within the previous year. The control group consisted of age-matched healthy adolescents, not related to participants with T1D.

After an overnight stay to maintain stable glycemia, participants with T1D started with an intravenous euglycemic clamp, during which functional MRI with an sWM task and MRI spectroscopy were performed. Capillary blood glucose concentration was measured (CONTOUR NEXT; Ascensia, Parsippany, NJ) every 15 min and maintained between 5 and 10 mmol/L (90–180 mg/dL). Afterward, the blood glucose level was raised by an intravenous glucose bolus, and a hyperglycemic clamp (6) was maintained between 15 and 25 mmol/L (270–450 mg/dL) for 2 h, during which participants repeated the scanning protocol.

To control for the effects of learning and fatigue, healthy control subjects followed the same experimental protocol, but without any manipulation of glycemia.

Participants were familiarized with the spatial variation of the Sternberg item-recognition task (7) prior to the scanning sessions. Group differences and the effect of induced hyperglycemia on sWM capacity estimates were assessed using a two-way mixed-design ANOVA with within-subject factor session (first vs. second session) and between-subject factor group (participants with T1D vs. healthy control subjects). The analysis was performed using the ez-package (8) separately for sWM loads of two and four items.

Neuroimaging data were acquired with an Achieva 3.0T TX scanner (Philips, Best, the Netherlands). T1- and T2-weighted high-resolution images were acquired, and in each functional session, three T2*-weighted blood oxygen level–dependent images concurrent with sWM task performance, followed by spectroscopy scan using SV-PRESS sequence, were obtained.

Initial image preprocessing followed the Human Connectome Project minimal preprocessing pipeline (9). Further analyses were conducted using FreeSurfer (https://surfer.nmr.mgh.harvard.edu) (10) and in-house code using MATLAB 2014a (https://mathworks.com) and R (11) software packages.

The analysis of functional data was performed on a priori selected regions of interest (ROIs) (frontal eye fields, supplementary motor area, anterior insula, anterior intraparietal area, medial intraparietal area, inferior frontal junction, dorsolateral prefrontal cortex, and medial temporal lobe) (8). The selected ROIs are involved in the spatial information forming and its maintenance and executive control (12,13). Separate estimates of brain activation during encoding, delay, and response phase for each session and sWM load were obtained using a general linear model. Individual activation estimates for each ROI were then submitted to a second-level analysis. The effects of interest on mean ROI responses were investigated using a mixed-effects ANOVA with within-subject factors session (first vs. second), load (2 vs. 4), task phase (encoding, delay, and response), hemisphere (left vs. right), ROI, and a between-subject factor group (participants with T1D vs. healthy control subjects).

In the frontal cortex, ratios of magnetic resonance relaxation–corrected surface areas of N-acetyl aspartate, choline, and creatine were compared within each group and between the groups after functional MRI sessions. The Wilcoxon rank-sum test (between groups) and Wilcoxon signed-rank test (paired difference inside groups between the first and second measurements) were used; correlation between the blood glucose concentration and metabolites was assessed with linear mixed-effects model, all using R.