Cerebral microbleeds and white matter hyperintensities in cardioembolic stroke patients due to atrial fibrillation: single-centre longitudinal study.

Cerebral microbleeds (CMBs) are a potential predictor of future stroke risk with clinical relevance for antithrombotic treatments, especially in ischaemic stroke patients with atrial fibrillation. However, prospective data on CMBs and risk of stroke in this particular stroke population remain scarce. We therefore performed a single centre longitudinal study to investigate CMBs and white matter hyperintensities (WMH) and the risk of future stroke. Consecutive acute stroke patients, admitted during 2008-2012 for presumed cardioembolic stroke due to non-valvular atrial fibrillation with available follow-up for the occurrence of recurrent stroke were included in our study. The rate of future stroke between patients with vs. without CMBs and moderate to severe WMH at baseline MRI was compared in separate survival and multivariable Cox regression analyses. A total of 119 cardioembolic stroke patients (49% female, median age: 76; IQR: 68-82years) were included. CMBs were found at baseline in 26/119 (21.8%; 95% CI: 14.8-30.4%) patients. Moderate to severe WMH were present in 27/119 (22.7%; 95% CI: 15.5-31.3%) cases. During a median follow-up time of 17months (IQR: 3-50months), 17 of 119 patients experienced a symptomatic stroke: 14 patients had an ischaemic stoke and 3 had intracerebral haemorrhage. The overall incidence rate for ischaemic stroke and intracerebral haemorrhage was 4.2 (95% CI: 2.3-7.1) and 0.9 (95% CI: 0.5-2.6) per 1000 patient-year of follow-up respectively. In multivariable Cox regression analysis the hazard ratio for total CMB number and the risk of stroke during follow-up was 1.05 (95% CI: 0.99-1.11; p=0.137, per each additional CMB increase), after adjusting for CHAD2S. A similar regression analysis demonstrated that moderate to severe WMHs were independently associated with increased risk of symptomatic stroke at follow-up, after adjusting for CHAD2S (HR: 2.99; 95% CI: 1.01-8.30; p=0.036). Despite the small sample size, our study provides useful data to guide power calculations and likely effect sizes relevant for ongoing and future larger studies and clinical trials.