Association of diabetes, glycemic status and vascular brain lesions in patients with atrial fibrillation

Introduction The global prevalence of atrial fibrillation (AF) and diabetes mellitus (DM) is rising, contributing to increased risk of cerebrovascular complications. While brain lesions in AF patients are primarily associated with cardioembolic mechanisms, DM-related brain lesions are linked to small vessel disease. However, the association of DM with vascular brain lesions in patients with AF remains unclear. Purpose This study aims to investigate the associations between diabetes and glycemic status with vascular brain lesions in AF patients. Methods A total of 1,706 patients from a large national prospective AF cohort study with standardized brain MRI and HbA1c measurements were included in this cross-sectional analysis. Patients were categorized by HbA1c levels into normoglycemia (<5.7%), prediabetes (5.7-6.4%) and diabetes (≥6.5%). MRI-detected brain lesions included large cortical and noncortical infarcts (LNCCI), small noncortical infarcts (SNCI), white matter hyperintensities (WMH) and microbleeds (MB), and their respective volume. Associations of DM and brain lesions were assessed by logistic and linear regression models. Results Mean age was 72 ± 8 years (27% female; 46% paroxysmal AF; 16% prior DM diagnosis). According to baseline HbA1c levels, 500 (29.3%) had normoglycemia, 922 (54%) prediabetes and 284 (16.6%) diabetes. The prevalence of each brain lesion type increased stepwise across HbA1c categories (Figure 1). Median [IQR] brain lesion volume (mm3) across the three categories was 1230 [257-7338], 1685 [232-6983] and 2001 [259-7021] for LNCCI; 51 [28.5-111], 66 [30-173] and 63 [36-183] for SNCI; and 2788 [1168-7500], 4173 [1563-10146] and 5093 [2005-12109] for WMH. While DM categories were associated with the prevalence of most brain lesions in non-adjusted logistic regression models, none of the categories were associated with prevalent brain lesions after multivariable adjustment. In multi-adjusted linear regression models, DM categories were associated with SNCI volume (Beta=0.49, 95% CI 0.09-0.89, p=0.016), but not with other brain lesion volumes. Conclusion Patients with prediabetes and diabetes have a higher prevalence of vascular brain lesions. After multivariable adjustment for age, sex and comorbidities, glycemic status was not independently associated with prevalent vascular brain lesions in patients with AF. Diabetes remained significantly associated with SNCI volume in multivariable-adjusted models. These findings warrant further investigation in longitudinal studies.