Accuracy of cardiac CT, radionucleotide and invasive ventriculography, two- and three-dimensional echocardiography, and SPECT for left and right ventricular ejection fraction compared with cardiac MRI: a meta-analysis.

AIMS Left ventricular ejection fraction (LVEF) and right ventricular ejection fraction (RVEF) are important tools in clinical decision-making. We hypothesized that two-dimensional echocardiography (2DE), three-dimensional echocardiography (3DE), radionucleotide ventriculography (RNV), cardiac computed tomography (CT), gated single-photon emission CT (SPECT), and invasive cardiac cine ventriculography (ICV) provide variable accuracy for LVEF and RVEF when using cardiac magnetic resonance imaging (MRI) as a gold standard. METHODS AND RESULTS We systematically searched published databases for studies comparing LVEF and RVEF measured by CT, 3DE, 2DE, RNV, ICV, and SPECT compared with MRI. We utilized meta-analytic methods to determine the pooled bias (mean weighted difference), limits of agreement (LOA), and correlation coefficient for each modality. For LVEF, 174 studies (7047 patients) were included. For RVEF, 46 studies (1720 patients) were included. Pooled LOA for LVEF were different between modalities: CT and 3DE had smaller LOA than 2DE, SPECT, ICV, and RNV. 2DE showed the largest LOA and a weaker correlation for LVEF (-13.3 to 12.1%, r = 0.660). For RVEF, CT and 3DE have the best data to support their use with a bias <5% and tight LOA and correlation coefficients with (r) >0.75. CONCLUSION For LVEF, CT and 3DE had the lowest bias and the best agreement with MRI. Compared with MRI, CT and 3DE comparably estimate RVEF and have the most evidence to support their use.