Fig. 3.

A causal diagram in which C confounds the smoking/ischemic stroke, wider venules/ischemic stroke, and smoking/wider venules relation. We assume there is no unmeasured confounding, and that smoking does not affect C. Here C and U refer to the following list of measured person characteristics: age, sex, subcohort, education, systolic blood pressure, diastolic blood pressure, blood pressure lowering medication use, body mass index, total cholesterol, high-density lipoprotein cholesterol, white blood cell counts, diabetes mellitus, alcohol intake, carotid plaques, history of cardiovascular disease, and retinal arteriolar caliber