Supplemental Table 1. Demographics and risk factors comparison between subjects included and excluded.

Variables (n, % or M ± SD)	Total (n = 6,906)	Inclusion (n = 3,678)	Exclusion (n = 3,228)	p-value
Age (year)	61 ± 14	62 ± 13	59 ± 15	< 0.001
Male	3,089 (44.7)	1,583 (43.0)	1,506 (46.7)	0.003
Risk factors
Hypertension	3,653 (53.7)	2,127 (58.4)	1,526 (48.2)	< 0.001
Diabetes	1,260 (18.4)	816 (22.2)	444 (14.0)	< 0.001
Hyperlipidemia	3,178 (53.0)	2,235 (60.8)	943 (40.0)	< 0.001
Atrial filtration	114 (2.3)	85 (2.4)	29 (2.1)	0.409
CAD	530 (7.8)	333 (9.1)	197 (6.2)	< 0.001
Previous stroke history	585 (8.5)	330 (9.0)	255 (7.9)	0.110
Smoking	2,090 (33.1)	1,104 (30.0)	986 (33.1)	0.007
Drinking	1,960 (29.2)	968 (26.3)	992 (31.7)	< 0.001
BMI (kg/m2)	25.57 ± 3.77	25.78 ± 3.83	25.26 ± 3.68	< 0.001

CAD: coronary heart disease

In summary, subjects included were much older (mean age 62 vs 59, p < 0.05), more likely to have hypertension (58.4% vs 48.2%, p < 0.05), diabetes (22.2% vs 14.0%, p < 0.05), hyperlipidemia (60.8% vs 40.0%, p < 0.05), and coronary disease (9.1% vs 6.2%, p < 0.05), but the proportion of male is much lower than in those excluded (43.0% vs 46.7%, p < 0.05), accordingly, the percentage of smoking (30.0% vs 33.1%) and drinking (27.0% vs 31.7%).

Several factors contribute to the unmatched difference. First, some risk factors are age-related, thus older subjects have a much higher percentage of risk factors; Second, subjects with more risk factor seems to more willing to participate the biochemistry testing and vascular assessment. Third, the risk factors diagnosis was based on self-reported medical history in the exclusion group, lack of laboratory test might possibly result in the underestimation of diagnosis, such as diabetes and hyperlipidemia. Finally, the gender difference between the two groups could explain part of the unbalance of smoking and drinking.