Background and aims: Vascular disease (VD), as assessed by history of myocardial infarction or peripheral artery disease or aortic plaque, increases stroke risk in atrial fibrillation (AF), and is a component of risk assessment using the CHA2DS2-VASc score. We investigated if systemic atherosclerosis as detected by ultrasound carotid plaque (CP) could improve the predictive value of the CHA2DS2-VASC score. Methods: We analysed data from the ARAPACIS study, an observational study including 2027 Ialian patents with non-valvular AF, in whom CP was detected using Doppler Ultrasonography. Results: VD was reported in 351 (17.3%) patients while CP was detected in 16.6% patents. Adding CP to the VD definition leaded to higher VD prevalence (30.9%). During a median [IQR] follow-up time of 36 months, 56 (2.8%) stroke/TIA events were recorded. Survival analysis showed that conventional VD alone did not increase the risk of stroke (Log-Rank: 0.009, p = 0.924), while addition of CP to conventional VD was significantly associated to an increased risk of stroke (LR: 5.730, p = 0.017). Cox regression analysis showed that VD + CP was independently associated with stroke (HR: 1.78, 95% CI: 1.05-3.01, p = 0.0318). Reclassification analysis showed that VD + CP allowed a significant risk reclassification when compared to VD alone in predicting stroke at 36 months (NRI: 0.192, 95% CI: 0.028-0.323, p = 0.032). Conclusions: In non-valvular AF patients the addition of ultrasound detection of carotid plaque to conventional VD significantly increases the pedictive value of CHA2DS2-VASc score for stroke.