Highlights
In the ARIC cohort, incident ischemic stroke occurring at age 80 years or older was more likely to be embolic than thrombotic.
Atrial fibrillation, abnormal left atrial volume index, and abnormal P-wave axis each explained a substantial share of the age-related shift toward embolic stroke, supporting an atrial disease pathway.
Standard prediction tools performed poorly on their own in this oldest-old population, but adding embolic risk factors markedly improved discrimination, especially for the Predicting Risk of Cardiovascular Disease Events equation.
The findings reinforce a practical message for late-life stroke prevention: age alone is not the whole story; atrial substrate matters.
Background
Stroke remains a leading cause of death and long-term disability in older adults, and the burden rises steeply in the oldest-old. As survival into the eighth, ninth, and even tenth decade becomes more common, clinicians increasingly face two related problems: first, understanding why ischemic stroke occurs later in life, and second, identifying which older adults are most likely to experience it. These tasks are not the same. Chronological age captures overall risk accumulation, but it does not specify the biological pathway that leads to stroke.
In late life, embolic ischemic stroke is often linked to atrial fibrillation and broader atrial cardiopathy, a constellation of structural and electrical abnormalities of the left atrium that can promote clot formation and embolization even when atrial fibrillation is intermittent or not yet documented. By contrast, thrombotic ischemic stroke more often reflects local atherosclerotic disease. Distinguishing these mechanisms matters because prevention strategies differ: anticoagulation, rhythm monitoring, and recognition of atrial disease are more relevant to embolic mechanisms, whereas aggressive vascular risk factor control remains central for atherosclerotic disease.
The Atherosclerosis Risk in Communities study offers a valuable setting for this question because it is a long-running community-based cohort with adjudicated stroke outcomes and detailed cardiovascular phenotyping. The present analysis asked whether the effect of very old age at stroke onset is mediated by embolic risk factors and whether these factors improve prediction beyond commonly used cardiovascular scores.
Study design
This was an observational cohort analysis nested within ARIC. For the causal analysis, investigators included stroke-free participants at visit 5 (2011-2013) who subsequently developed incident acute ischemic stroke between visit 5 and visit 10 (2023). For the prediction analysis, stroke-free participants at visit 5 were used to evaluate how well available risk tools identified future ischemic stroke occurring at age 80 years or older.
The primary comparison for subtype analysis was age at stroke onset of 80 years or older versus younger than 80 years, with adjudicated stroke subtype classified as embolic ischemic stroke or thrombotic ischemic stroke. The authors used logistic regression to estimate the association between late-life stroke onset and stroke subtype. They then performed bootstrapped mediation analyses with 1000 iterations to test whether atrial fibrillation, myocardial infarction, coronary heart disease, heart failure, and electrocardiographic or echocardiographic measures explained part of the age-subtype relationship.
For prediction, the authors examined the Predicting Risk of Cardiovascular Disease Events equation and the CHA2DS2-VASc score, then assessed whether adding embolic risk factors improved the c statistic. In other words, they asked whether the models could better distinguish participants who would later have ischemic stroke at age 80 years or older from those who would not.
Key findings
Among 6213 stroke-free participants at visit 5, 277 (4.4%) experienced incident acute ischemic stroke over a median follow-up of 5.1 years. The cohort had a median age of 76 years at baseline, and the median age at stroke was 81 years. Women comprised 62% of the sample. Of the 277 strokes, 99 were embolic ischemic strokes and 178 were thrombotic ischemic strokes.
The main subtype result was clinically important: participants who developed ischemic stroke at age 80 years or older had higher odds of embolic ischemic stroke, relative to thrombotic ischemic stroke, than participants whose stroke occurred before age 80. The odds ratio was 1.90 with a 95% confidence interval of 1.09 to 3.31. This indicates nearly a doubling of the odds of embolic rather than thrombotic stroke in the oldest-old, although the confidence interval reminds us that the estimate is modest in precision.
The mediation analysis is the most mechanistically interesting part of the study. The age-related association with embolic stroke was partially mediated by atrial fibrillation, which explained 44% of the effect (P=0.03), by abnormal left atrial volume index, which explained 45% (P=0.048), and by abnormal P-wave axis, which explained 43% (P=0.04). These are not identical variables; they capture complementary aspects of atrial structure and electrical function. Together, they point toward atrial cardiopathy as a plausible bridge between advanced age and embolic stroke.
By contrast, the study also tested myocardial infarction, coronary heart disease, heart failure, and other electrocardiographic or echocardiographic measures, but the abstract highlights AF and left atrial markers as the clearest mediators. This does not mean that ischemic heart disease or heart failure are unimportant in the elderly; rather, it suggests that they did not explain as much of the specific age-to-embolic-stroke relationship in this analysis.
Prediction performance
The prediction results underscore how poorly traditional cardiovascular scores alone perform when the outcome is late-life ischemic stroke.
| Prediction model | Sample size | C statistic before embolic factors | C statistic after embolic factors |
|---|---|---|---|
| Predicting Risk of Cardiovascular Disease Events equation | 5702 | 0.49 (95% CI, 0.45-0.53) | 0.77 (95% CI, 0.74-0.80) |
| CHA2DS2-VASc score | 5739 | 0.57 (95% CI, 0.55-0.59) | 0.63 (95% CI, 0.59-0.67) |
A c statistic of 0.49 is essentially no better than chance, while 0.57 reflects only modest discrimination. Adding embolic risk factors improved performance, especially for the Predicting Risk of Cardiovascular Disease Events equation, which rose to 0.77. CHA2DS2-VASc improved more modestly, to 0.63. The gain suggests that the model was missing clinically meaningful information about atrial and embolic biology. It also reinforces a key limitation of general cardiovascular risk tools: they were not built specifically to predict stroke subtype in the oldest-old.
In practical terms, the study suggests that if clinicians want to anticipate late-life ischemic stroke more accurately, they may need to look beyond age, hypertension, diabetes, and broad vascular history, and pay closer attention to atrial rhythm, left atrial size, and related electrocardiographic markers.
Expert commentary
This study is notable for moving the conversation from simple risk association toward mechanism and prediction. The central message is not merely that older adults have more stroke, which is already well known, but that the nature of late-life stroke appears to shift toward embolic disease, with atrial fibrillation and atrial cardiopathy accounting for much of that shift. That has direct implications for clinical practice.
First, older adults with new neurologic symptoms and no obvious large-artery cause should prompt a careful search for occult atrial disease. Prolonged rhythm monitoring, review of prior electrocardiograms, and echocardiographic assessment of left atrial size may be particularly useful in selected patients. Second, the findings support the broader concept that atrial cardiopathy may be clinically relevant even when atrial fibrillation has not yet been documented. Third, the study argues for stroke-specific prediction tools rather than relying on general cardiovascular scores.
At the same time, the analysis has limitations that matter for interpretation. It is observational, so mediation analysis can support plausibility but cannot prove causality in the experimental sense. The embolic versus thrombotic subtype assignment depends on adjudication, which is a strength, but residual misclassification can still occur. The oldest-old are also a selected population: many individuals with severe disease may have already experienced events or died before reaching visit 5, which can affect generalizability. Finally, the improved prediction seen after adding embolic risk factors may partly reflect variables closely tied to the outcome rather than a ready-made bedside tool.
These caveats do not weaken the core message. Instead, they define the next step: prospective studies should test whether atrial cardiopathy markers and rhythm surveillance can improve prevention decisions, especially when paired with contemporary anticoagulation strategies and better stroke phenotyping.
Conclusion
In ARIC, ischemic stroke occurring at age 80 years or older was more likely to be embolic than thrombotic, and much of that association was explained by atrial fibrillation and markers of left atrial disease. Standard risk scores were weak predictors of late-life ischemic stroke, but embolic risk factors substantially improved performance. For clinicians, the study strengthens the case for looking carefully at the atrium when assessing stroke risk in very old adults.
Funding and clinicaltrials.gov
Funding details were not provided in the abstract supplied here. This was an observational cohort analysis and therefore does not have a clinicaltrials.gov registration number in the usual sense of an interventional trial.
References
1. Wang J, Egle M, Jin Z, Lakshminarayan K, Ndumele CE, Coresh J, Gottesman RF, Johansen MC. Late-Life Incident Stroke in the Atherosclerosis Risk in Communities Study: Cause and Prediction. Stroke. 2026 Apr 30. PMID: 42059062.
2. Kleindorfer DO, Towfighi A, Chaturvedi S, et al. 2021 Guideline for the Prevention of Stroke in Patients With Stroke and Transient Ischemic Attack. Stroke. 2021;52:e364-e467.