. Author manuscript; available in PMC: 2021 Jul 27.

Published in final edited form as: J R Stat Soc Ser A Stat Soc. 2020 May 8;183(4):1705–1726. doi: 10.1111/rssa.12570

Table 3.

Model-based sensitivity analyses

	Mediators: brain hypoperfusion and white matter disease
	OR	90% CI	Risk ratio	90% CI
Model 1^†
Total effect	1.51	(1.22, 1.86)	1.27	(1.13, 1.42)
Direct effect	1.31	(1.01, 1.72)	1.20	(1.03, 1.39)
Indirect effect	1.15	(0.98, 1.35)	1.06	(0.97, 1.16)
Model 2^‡
Total effect	1.51	(1.22, 1.86)	1.27	(1.13, 1.42)
Direct effect	1.34	(1.02, 1.75)	1.20	(1.04, 1.40)
Indirect effect	1.13	(0.96, 1.32)	1.05	(0.96, 1.15)
Model 3^§
Total effect	1.51	(1.22, 1.86)	1.27	(1.13, 1.42)
Direct effect	1.38	(1.05, 1.81)	1.22	(1.05, 1.42)
Indirect effect	1.09	(0.93, 1.29)	1.04	(0.95, 1.14)
Model 4^§§
Total effect	1.54	(1.22, 1.95)	1.27	(1.12, 1.45)
Direct effect	1.30	(0.97, 1.75)	1.18	(1.00, 1.39)
Indirect effect	1.18	(0.99, 1.42)	1.08	(0.98, 1.20)

^†

Model inputs were kept the same as in the original analysis.

^‡

Model inputs were kept the same as in the original analysis, except that the white matter disease cut-off changed to greater than 1 out of 9 points.

^§

Model inputs were kept the same as in the original analysis, except that the white matter disease cut-off changed to greater than 3 out of 9 points.

^§§

Model inputs were kept the same as in the original analysis, plus additional terms for baseline log₁₀(interleukin-6 level), baseline log₁₀(C-reactive protein level) depressive symptoms, income level and physical activity.