Table 5.

The effect of overweight on systolic blood pressure, adjusted for alcohol consumption. 2nd column: linear models in which the confounder-exposure and/or confounder-outcome associations are modelled as linear. 3rd column: spline models in which the confounder-exposure and/or confounder-outcome associations are modelled as non-linear

	Linear models	Non-linear spline models
	β (95% CI)	β (95% CI)
Multivariable regression analysis	3.589 (0.686; 6.493)	3.022 (0.136; 5.908)
Covariate adjustment using the PS	3.739 (0.822; 6.656)	3.062 (0.164; 5.960)
Stabilized IPW	4.121 (1.110; 7.132)	3.813 (0.807; 6.819)
DR estimation	3.983 (1.262; 6.704)	3.585 (0.879; 6.291)

Abbreviations: PS: propensity score, IPW: inverse probability weighting, DR: double robust, β: regression coefficient, CI: confidence interval