Table 3.

Threshold effect of dietary selenium intake on risk of stroke.

Dietary selenium intake (100 μg/day, n = 39,438)	Model 1	Model 2	Model 3
Linear model
OR (95% CI) p-value	0.71 (0.64, 0.78) <.001	0.81 (0.72, 0.90) <.001	0.86 (0.74, 1.00) . 051
Non-linear model
Breakpoint (K)	2.00	1.20	1.05
OR1(95% CI), <K p-value	0.63 (0.55, 0.71) <.001	0.62 (0.49, 0.77) <.001	0.61 (0.44, 0.85) . 004
OR2(95% CI), >K p-value	1.22 (0.88, 1.71) .237	0.99 (0.83, 1.18) .889	0.97 (0.81, 1.16) .723
p-Value for non-linearity	.002	.007	.025

The piece-wise linear regression model was applied to show the threshold effect of dietary selenium intake on the risk of stroke. Linear model: model that presumes the correlation between dietary selenium intake and the risk of stroke is linear. Non-linear model: model that presumes the correlation between dietary selenium intake and the risk of stroke is non-linear and has a breakpoint. p-Value for non-linearity <.05 means that the non-linear model may better describe the correlation. Model 1 was adjusted for age, sex, and race. Model 2 was adjusted for education level, marital status, poverty-income ratio, body mass index, smoking, alcohol use, hypertension, diabetes, physical activity based on Model 1. Model 3 was adjusted for levels of haemoglobin, uric acid, total cholesterol, HDL-cholesterol, triglyceride, glycohemoglobin, daily intake of total energy and cholesterol from the diet based on Model 2. OR (95% CI): odds ratio and 95% confidence interval.