Results:

A total of 3,289 participants (48.8% men) were included and followed over 9.0 ± 1.9 years. Eighty-four strokes (75 ischemic, 9 hemorrhagic) occurred during follow-up among 84 subjects (2.55%; 48 men, 36 women). Compared with subjects not having a stroke, subjects exhibiting a stroke were older; had a higher waist circumference; more often revealed arterial hypertension, diabetes mellitus, and atrial fibrillation; and had higher triglycerides and a higher Framingham risk score. Subjects experiencing a stroke had higher CAC and CIMT values, and lower ABI values. They more frequently received medications for hypertension and diabetes. In multivariable Cox proportional hazard regression models, CAC (hazard ratio, 1.45 [95% confidence interval, 1.11–1.88] per standard deviation [SD] increase in ln(CAC+1); SD, 2.40), CIMT (1.34 [1.08–1.66] per SD increase; SD, 0.127 mm), and ABI (1.55 [1.32–1.82] per SD decrease; SD, 0.148) were associated with stroke, with established risk factors included in the models. When combined with each other, ln(CAC+1)’s hazard ratio remained similar when CIMT (1.41 [1.09–1.83]) was inserted into the multivariable model, but slightly decreased when ABI (1.31 [1.01–1.72]) or CIMT and ABI (1.29 [0.99–1.68]) were included. Although CAC alone did not significantly elevate the area under the curve in Harrell’s c-statistics (by 0.009; p = 0.379) in addition to established risk factors, the combination of CAC and ABI increased area under the curve (by 0.029; p = 0.047), as did ABI (by 0.025; p = 0.038), but not CIMT (by 0.002; p = 0.795) alone. The combination of CAC and ABI also resulted in significant category-free net reclassification and integrated discrimination improvement.