Table 2.

Adjusted marginal structural models estimating controlled direct effects (CDE) of inflammatory markers (i.e., high-sensitivity C reactive protein (CRP), tumour necrosis factor (TNF), and Interleukin 6 (IL-6)) on systolic blood pressure (SBP) measures in the presence and absence of vitamin D (VD) insufficiency [25(OH)D<50 nmol/L] among Indigenous Cree adults from the Multi-Community Environment-and-Health Study, 2005–2009

CDE for SBP measures
	VD insufficiency present (m=1)		P value*	VD insufficiency absent (m=0)		P value*
	†E(Y(a, m(1) – Y(a*, m(1)))			†E(Y(a, m(0) – Y(a*,m(0))
	Estimate β	95% CI		Estimate β	95% CI
CRP
Model 1	−2.32	−15.19 to 10.54	0.724	0.78	−5.30 to 6.85	0.803
Model 2	−1.06	−14.50 to 12.37	0.877	0.90	−5.72 to 7.56	0.789
Model 3	1.76	−14.75 to 18.29	0.839	0.41	−6.93 to 7.75	0.913
TNF-α
Model 1	−13.67	–23.78 to –3.56	0.008*	−5.70	−11.47 to 0.08	0.053
Model 2	−13.61	–23.54 to –3.07	0.011*	−5.92	–11.65 to –0.18	0.043
Model 3	−13.61	–24.42 to –2.80	0.014*	−5.73	−11.59 to 0.13	0.052
IL-6
Model 1	5.79	−15.26 to 26.85	0.589	1.85	−7.94 to 11.64	0.711
Model 2	7.81	−13.59 to 29.22	0.474	1.54	−7.70 to 10.80	0.743
Model 3	13.9	−13.32 to 41.15	0.317	2.62	−10.75 to 15.99	0.701

*Statistical significance (p<0.05).

†CDE for the each inflammatory marker exposure level indicating the average direct effect change on SBP and diastolic blood pressure outcomes if the exposure is contrasted (ie, a=1, and a*=0), while uniformly fixing the intermediate variable (m) to either designated level, VD insufficiency or VD sufficiency (ie, m(1) and m(0), respectively) in the population. Model 1: age, sex; Model 2: model 1, plus smoking status; Model 3: model 2, plus body mass index.