Research Design and Methods
Sampling
We used a cross-sectional study design, and sampling methods have been described previously. Participants were included from two studies: the Cognition and Diabetes in Older Tasmanians study (CDOT) and the Tasmanian Study of Cognition and Gait. Participants with T2DM aged ≥55 years were recruited into CDOT between January 2008 and January 2010 using the National Diabetes Service Scheme database as a sampling frame. The Tasmanian Study of Cognition and Gait sample was recruited by mailing approach letters to eligible registrants aged ≥55 years, living in the same Southern Tasmanian postcodes as those in the CDOT study, and has been described previously. The phenotype of T2DM was based on self-report and confirmed using a single plasma glucose level according to standard criteria (fasting plasma glucose ≥7.0 mmol/L, random plasma glucose ≥11.1 mmol/L, and HbA1c >6.5% [48 mmol/mol]). People living in a nursing home, those with insufficient English for cognitive testing, or contraindication to magnetic resonance imaging (MRI) were excluded. The Southern Tasmanian Health and Medical Human Research Ethics Committee and the Monash University Human Research Ethics Committee approved the study, and we obtained written, informed consent.
Measurements
Saf. We used the AGE reader (DiagnOptics BV, Groningen, the Netherlands) to measure SAF. The spectrometer reader uses a light source to illuminate ~4 cm of skin on the volar surface of the right arm 10 cm below the elbow fold. SAF is calculated as the ratio between the emission light and reflected excitation light, multiplied by 100 and expressed in arbitrary units. In our laboratory, the test-retest reliability for SAF was high (intraclass correlation coefficient 0.93; n = 11) when individuals were measured 5 days apart.
MRI Scans: MRI scans were obtained using a single 1.5T General Electric scanner with the following sequences: high-resolution T1-weighted spoiled gradient echo (GRE; TR 35 ms; TE 7 ms; flip angle 35°; field of view 24 cm; 120 contiguous slices; and isotropic voxel size 1 mm); T2-weighted fast spin echo (repetition time [TR], 4,300 ms; echo time [TE], 120 ms; number of excitations, 1; turbo factor, 48; and voxel size, 0.90 × 0.90 × 3 mm); fluid attenuated inversion recovery (TR, 8,802 ms; TE, 130 ms; TI, 2,200 ms; and voxel size, 0.50 × 0.50 × 3 mm); and GRE (TR, 0.8 ms; TE, 0.015 ms; flip angle, 30°; and voxel size, 0.9 × 0.9 × 7 mm).
Brain Volumes: Three-dimensional T1 and axial GRE sequences were registered into standard Montreal Neurological Institute space using Functional Magnetic Resonance Imaging of the Brain's Linear Image Registration Tool. A multispectral segmentation process was applied using three-dimensional T1 and GRE sequences using Statistical Parametric Mapping software version 5 to produce tissue probability maps of gray and white matter. Tissue maps were smoothed using an isotropic 8-mm Gaussian kernel. A single expert manually segmented both hippocampi using established methods known to have high test-retest reliability in our laboratory (intraclass correlation coefficient 0.97). Tissue volumes of the segmented areas (total gray, white matter, and hippocampal) were calculated using standard voxel-counting algorithms.
Other Measurements: Standardized questionnaires were used to record demographic and clinical information. Weight, height, waist and hip circumferences, habitual physical activity using a pedometer worn over 1 week, and blood pressure (BP) in a sitting position as an average of three recordings from the right arm were measured and BMI calculated. A standardized cognitive battery was used to test domains of memory, speed, and executive and visuospatial function (Supplementary Table 1 http://diabetes.diabetesjournals.org/content/64/1/279/suppl/DC1) as described previously. Diagnosis of cognitive impairment was assigned, blinded to T2DM status, if function in any of the domains was <1.5 SDs from age-, sex-, and education-adjusted norms.
Data Analysis
The analyses were conducted on a complete dataset consisting of those in whom both measures of SAF and brain imaging were available.
Logistic regression was used to describe the associations of T2DM and brain atrophy with cognitive impairment. Linear regression was used to estimate the associations of SAF and T2DM with measures of brain atrophy. Covariates for age, sex, total intracranial volume (TICV), and other variables were added to the regression models for brain atrophy if their inclusion produced a statistically significant increase in model fit or changed the coefficient of the covariate for T2DM by >10%. Putative factors considered were hypertension (defined as mean BP >140/90 mmHg or previous diagnosis), ever smoked tobacco, creatinine, mean steps per day, history of ischemic heart disease, stroke, hyperlipidemia, BMI and waist-to-hip ratio, and the use of specific medications that have been shown to influence AGE levels (pravastatin, irbesartan, and metformin). We then examined whether SAF mediated the associations estimated between T2DM and brain atrophy. For this, we entered SAF into the model relating T2DM to brain volume outcome measures adjusting for age, sex, smoking, serum creatinine, and TICV. If the introduction of SAF substantially attenuated the regression coefficient of the binary covariate for T2DM, and the coefficient of SAF remained largely unchanged from its value without T2DM in the model, it was considered a potential mediator. We also investigated any modifying effect (interaction) of SAF using a test of significance of the coefficient of a covariate formed as the product of the covariates for T2DM and SAF. Statistical analyses were carried out using STATA version 11.1 (StataCorp, College Station, TX).