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 2 Adjustment for Differences in Patient Characteristics between Study Arms: Sensitivity to Prior Distributions

		Type of Prior
Dataset	Posterior Estimate Median OR (95% credible interval)	Base Case Analysis: " Reasonably Vague" (Grines)	Sensitivity Analysis: "Fairly Unrestrictive" (Prevost)	Sensitivity Analysis: " Vaguest"
	Overall (μ)	0.37 (0.17,0.77)	0.37 (0.23,0.60)	0.37 (0.18,1.25)
Base Case 3 Covariates ^a (k = 44)	Randomised (θ ₁ )	0.35 (0.17,0.63)	0.36 (0.21,0.59)	0.34 (0.13,0.74)
	Non-Randomised (θ ₂ )	0.39 (0.25,0.61)	0.38 (0.25,0.57)	0.40 (0.23,0.68)
	Overall (μ)	0.45 (0.20,0.95)	0.47 (0.28,0.74)	0.44 (0.13,1.31)
Imputed 5 Covariates ^b (k = 79)	Randomised (θ ₁ )	0.42 (0.18,0.78)	0.46 (0.25,0.73)	0.39 (0.14,0.87)
	Non-Randomised (θ ₂ )	0.49 (0.33,0.72)	0.49 (0.33,0.71)	0.49 (0.32,0.74)

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

ISSN: 1471-2288