Examining The Interplay Between Atrial Fibrillation, Anticoagulation, and Retinal Stroke - An Instrumental Variable Approach

Cardioembolic stroke (stroke coming from the heart) is the most lethal form of ischemic stroke, causing sudden, permanent deficits in strength, sensation, speech, and sight. The most common cause of cardioembolic stroke in older Americans is atrial fibrillation (AF), a chronic, cardiac arrythmia that can be clinically silent yet portend a risk of life-threatening stroke. There are approximately 700,000 people with undiagnosed AF in the United States. A recent NHLBI working group published recommendations for future directions in cardiac monitoring, a strategy to detect clinically silent AF and – ideally – intervene before it is too late. One facet of the recommendations was that individual clinical features should be used to target cardiac monitoring in patients who may be at highest risk of clinically silent AF. Retinal stroke (central retinal artery occlusion) is a disabling subset of ischemic stroke. It is not known whether AF is associated with retinal stroke. Knowing whether AF represents an independent risk factor for retinal stroke would be valuable, because it would motivate clinicians to pursue cardiac monitoring in patients with retinal stroke, and boost the odds of detection of AF in this vulnerable patient population. Compelling preliminary data that AF is a risk factor for retinal stroke is presented, however there are key limitations to its interpretation given the inherent biases in observational literature. The current application proposes to do approach this question from a distinct perspective. The central hypothesis of this proposal is that oral anticoagulation therapy modifies the observed association between AF and retinal stroke. Testing this hypothesis provides indirect evidence of an association between AF and retinal stroke, given that anticoagulation reduces thromboembolism in patients with AF. During this project, the principal investigator will continue to receive intensive mentorship with an added focus on instrumental variable analysis. An instrumental variable is a variable which is highly associated with a downstream exposure (in this case anticoagulation) but not associated with the key study end point (other than via its association with a downstream exposure) and not associated with unmeasured confounding. Specific Aim 1 will construct an instrumental variable – individual prescriber preference for anticoagulation use – and determine whether such a prescriber preference is associated with the hazard of retinal stroke. Specific Aim 2 with determine whether changes in anticoagulation exposure within individual Medicare beneficiaries is associated with the odds of retinal stroke, by means of a case-crossover design. Results of this project – whether positive, neutral or negative – will inform our understanding of the biology of retinal stroke, allow physicians to more appropriately counsel patients on the benefits and possible harms of anticoagulation, and allow the refinement of instrumental variable analysis when applied to time to event data. Completion of the proposed work will allow the PI to branch out into a new research direction informed by the results of the parent K23 and advance his long-term objective of becoming an independent investigator. Project Number: 1R03HL178686-01 | Fiscal Year: 2025 | NIH Institute/Center: National Heart Lung and Blood Institute (NHLBI) | Principal Investigator: Brian Mac Grory | Institution: DUKE UNIVERSITY, DURHAM, NC | Award Amount: $242,250 | Activity Code: R03 | Study Section: Special Emphasis Panel[ZHL1 CSR-Q (J1)] View on NIH RePORTER: https://reporter.nih.gov/project-details/1R03HL17868601