The importance of adjusting for potential confounders in Bayesian hierarchical models synthesising evidence from randomised and non-randomised studies: an application comparing treatments for abdominal aortic aneurysms

Table 1 Covariate Data: Average Imbalance between Study Arms

	Base Case 3 Covariates^a (k = 4 randomised and 40 non-randomised)	Imputed data 5 Covariates^b (k = 4 randomised and 75 non-randomised)
Study Type Non-randomised	Average Difference (EVAR-OSR)	Average Difference (EVAR-OSR)
Male (proportion)	0.09	0.10
Age (years)	2.40	2.53
Cardiac disease (proportion)	0.12	0.14
Pulmonary disease (proportion)	not considered as missing in 43% of the 75 non-randomised studies	0.10
Renal disease (proportion)	not considered as missing in 54% of the 75 non-randomised studies	0.05
Randomised
Male (proportion)	0.05	0.05
Age (years)	0.82	0.82
Cardiac disease (proportion)	0.05	0.05
Pulmonary disease (proportion)	not considered as missing in 25% of the 4 randomised studies	0.13
Renal disease (proportion)	not considered as missing in 50% of the 4 randomised studies	0.07

a.male, age, cardiac disease, b.male, age, cardiac disease, pulmonary disease, renal disease

ISSN: 1471-2288