Late-Life Incident Stroke in the Atherosclerosis Risk in Communities Study: Etiology and Prediction

BACKGROUND: – As life expectancy rises, identifying causes and risk factors for incident acute ischemic stroke (AIS) among the oldest-old (≥80 years) is increasingly important. We examined whether the effect of age at stroke on AIS subtype is mediated by embolic risk factors and whether these factors improve AIS prediction. METHODS: – Stroke-free participants from the ARIC study (Atherosclerosis Risk in Communities) who developed AIS between visit 5 (2011–2013) and visit 10 (2023) were included for causal analysis; Stroke-free participants at visit 5 were included for prediction analysis. In logistics regression models, the association between age at stroke-onset (≥80 versus <80 years) and adjudicated AIS subtype (embolic ischemic stroke versus thrombotic ischemic stroke) was determined. Bootstrapped mediation analyses (1000-iterations) tested whether atrial fibrillation, myocardial infarction, coronary heart disease, heart failure, and electro/echocardiogram measures mediated the age-AIS subtype relationship. C statistics were calculated for AIS prediction (Predicting Risk of Cardiovascular Disease Events, CHA₂DS₂-VASc) and compared preinclusion and postinclusion of embolic risk factors. RESULTS: – Of 6213 stroke-free participants at visit 5, 277 (4.4%) developed AIS during a median (Q1–Q3) of 5.1 (2.6–7.1) years (median [Q1–Q3] age: 76 [72–80] years; median [Q1–Q3] age at AIS: 81 [77–86] years; 62% female; 99 embolic ischemic stroke and 178 thrombotic ischemic stroke). Individuals with AIS ≥80 years had higher odds of embolic ischemic stroke (versus thrombotic ischemic stroke) compared with those aged <80 years (odds ratio, 1.90 [95% CI, 1.09–3.31]). The effect of age at stroke-onset on embolic ischemic stroke was mediated by atrial fibrillation (44%; P=0.03), an abnormal left atrium volume index (45%; P=0.048), or an abnormal P-wave axis (43%; P=0.04). The predictive performance for AIS ≥80 years using the Predicting Risk of Cardiovascular Disease Events equation (N=5702, C statistic, 0.49 [95% CI, 0.45–0.53]), or CHA₂DS₂-VASc score (N=5739, C statistic, 0.57 [95% CI, 0.55–0.59]) was poor, but inclusion of embolic risk factors improved the performance (Predicting Risk of Cardiovascular Disease Events: C statistics, 0.77 [95% CI, 0.74–0.80]; CHA₂DS₂-VASc: C statistics, 0.63 [95% CI, 0.59–0.67]). CONCLUSIONS: – These findings suggest that identification and control of embolic risk factors are critical to reduce stroke risk as people age, and better stroke-specific prediction tools are needed.