Impact of Cardiac Magnetic Resonance-Derived Right Ventricular Ejection Fraction on Adverse Outcomes: A Robust Bayesian Model-Averaged Meta-Analysis.

Abstract

Background: There are few meta-analyses examining the prognostic value of right ventricular ejection fraction (RVEF) for a specific type of cardiovascular disease (CVD). The aim of this study was to compare the association of cardiac magnetic resonance (CMR)-derived RVEF with adverse outcomes for several specific types of CVD, using a robust Bayesian model-averaged meta-analysis.

Methods: Three databases were searched for CMR articles reporting hazard ratios (HRs) of RVEF restricted to a specific type of CVD. For each specific type of CVD, Bayesian model-averaged meta-analyses with and without publication bias adjustments were conducted to evaluate the strength of evidence for RVEF according to the Bayes Factor (BF).

Results: Among 108 articles (21,166 patients) analyzing 11 CVD types, pooled HR for 5% reduction in RVEF assessed by publication bias-unadjusted, Bayesian model-averaged meta-analysis offered moderate or strong evidence of an association with outcomes for all types of CVD (HR: 1.07-1.37, BF₁₀: 4.3-9.6*10⁷). In contrast, a robust Bayesian model-averaged meta-analysis, adjusted for publication bias, found moderate or strong evidence in favor of an association of RVEF with outcomes only in hypertrophic cardiomyopathy (HR: 1.19, 95% CrI: 0.98-1.42, BF₁₀: 5.0), dilated cardiomyopathy (HR: 1.16, 95% CrI: 1-1.22, BF₁₀: 23.3), pulmonary hypertension (HR: 1.05, 95% CrI: 1-1.12, BF₁₀: 3.0), and aortic stenosis (HR: 1.15, 95% CrI: 0.97-1.34, BF₁₀: 4.2). There was weak evidence for an association of RVEF with adverse outcomes in seven other CVDs.

Conclusions: In a Bayesian meta-analysis adjusted for publication bias, there was moderate or strong evidence for an association of RVEF with outcomes for only four CVDs. Additional data may strengthen evidence regarding other CVDs.