Abstract
Purpose: :
Changes in retinal vasculature are possible risk markers of subclinical stroke and may predict incident stroke. We examined the association between retinal microvascular signs with incident stroke, and the incremental usefulness of adding retinal microvascular signs for stroke prediction in an Asian population.
Methods: :
We conducted a prospective, population-based cohort study among Asian Malay persons 40-80 years at baseline. Retinal microvascular signs were measured from baseline retinal photographs; these include quantitative retinal measures using semi-automated computer software (retinal vascular caliber, branching angle, tortuosity and fractal dimension) and qualitative signs (focal arteriolar narrowing, arteriovenous nicking, opacification of the arteriolar wall) and retinopathy, comprising microaneurysms and retinal hemorrhages. Systemic measures and C-reactive protein (CRP) level were measured at baseline. Incident stroke was defined as newly diagnosed of clinical stroke documented by National Registry of Diseases Office after the baseline examination. Cox regression models and the C statistic were calculated.
Results: :
A total of 3,189 participants were free of prevalent stroke at baseline, of which 2,648 (83.0% of 3,189 participants) had complete information on incident stroke events, retinal microvascular, systemic and CRP measurements. During the follow-up (median 3.41 years), 44 (1.66%) participants developed a stroke. In Cox proportional-hazards models adjusting for established risk factors (age, systolic blood pressure, HbA1c, body mass index, smoking, cholesterol and CRP), presence of retinopathy (hazard ratio [HR] 2.17, 95% confidence interval (CI), 1.09 to 4.34) and wider retinal venular caliber (HR 2.04, 95% CI, 0.88 to 4.75, comparing 4th vs. 1st quartiles) were associated with risk of stroke. The C statistic increased marginally when these two retinal microvascular signs were incorporated into the model (C statistic 0.805 vs. 0.798, p=0.290), particularly among persons with diabetes (C statistic 0.824 vs. 0.727), p=0.040). Other retinal microvascular signs were not associated with risk of stroke.
Conclusions: :
Retinopathy and wider retinal venular caliber predict the risk of stroke, and a retinal assessment may be useful for stroke risk stratification particularly in diabetic persons.
Keywords: imaging/image analysis: clinical • clinical (human) or epidemiologic studies: prevalence/incidence • retina