Background Observational studies have suggested that selenium (Se) may have beneficial effects against certain cardiovascular outcomes, with a possible U-shaped association. We assessed the hypothesis that blood Se concentration might be inversely associated with the prevalence of stroke and the relationship would be nonlinear. Methods and Results Data collected from adult participants (aged ≥20 years) in the Canadian Health Measures Survey ( CHMS 2007-2011, n=7065) and the US National Health and Nutrition Examination Survey ( NHANES 2011-2012, n=5030) were analyzed. A total of 82 (1.16%) and 202 (4.02%) stroke cases were identified in CHMS and NHANES . Respondents with stroke had lower Se levels than those without stroke, with a mean difference of 16 μg/L and 12 μg/L for CHMS and NHANES , respectively. Respondents with high blood Se concentration (tertile 3) had a lower prevalence of stroke compared with those with low Se concentration (tertile 1). The adjusted odds ratios were 0.38 (95% CI : 0.15, 0.92) and 0.57 (95% CI : 0.31, 1.03) for CHMS and NHANES , respectively. A continuous decreasing trend of stroke with whole blood selenium was observed in CHMS , whereas the curve plateaued starting at 190 μg/L for NHANES , based on the cubic restricted spline regression. Sensitivity analysis using the serum and urinary Se concentrations demonstrates that our results were consistent across different selenium biomarkers. Conclusions We observed inverse cross-sectional associations between whole blood Se and the prevalence of stroke in representative samples of the Canadian and the US population.
Keywords: Canadian Health Measures Survey; National Health and Nutrition Examination Survey; diet; environment; selenium; stroke.