Comparison of left ventricular ejection fraction and volumes in heart failure by echocardiography, radionuclide ventriculography and cardiovascular magnetic resonance; are they interchangeable?

AIMS To prospectively compare the agreement of left ventricular volumes and ejection fraction by M-mode echocardiography (echo), 2D echo, radionuclide ventriculography and cardiovascular magnetic resonance performed in patients with chronic stable heart failure. It is important to know whether the results of each technique are interchangable, and thereby how the results of large studies in heart failure utilizing one technique can be applied using another. Some studies have compared cardiovascular magnetic resonance with echo or radionuclude ventriculography but few contain patients with heart failure and none have compared these techniques with the current fast breath-hold acquisition cardiovascular magnetic resonance. METHODS AND RESULTS Fifty two patients with chronic stable heart failure taking part in the CHRISTMAS Study, underwent M-mode echo, 2D echo, radionuclude ventriculography and cardiovascular magnetic resonance within 4 weeks. The scans were analysed independently in blinded fashion by a single investigator at three core laboratories. Of the echocardiograms, 86% had sufficient image quality to obtain left ventricular ejection fraction by M-mode method, but only 69% by 2D Simpson's biplane analysis. All 52 patients tolerated the radionuclude ventriculography and cardiovascular magnetic resonance, and all these scans were analysable. The mean left ventricular ejection fraction by M-mode cube method was 39+/-16% and 29+/-15% by Teichholz M-mode method. The mean left ventricular ejection fraction by 2D echo Simpson's biplane was 31+/-10%, by radionuclude ventriculography was 24+/-9% and by cardiovascular magnetic resonance was 30+/-11. All the mean left ventricular ejection fractions by each technique were significantly different from all other techniques (P<0.001), except for cardiovascular magnetic resonance ejection fraction and 2D echo ejection fraction by Simpson's rule (P=0.23). The Bland-Altman limits of agreement encompassing four standard deviations was widest for both cardiovascular magnetic resonance vs cube M-mode echo and cardiovascular magnetic resonance vs Teichholz M-mode echo at 66% each, and was 58% for radionuclude ventriculography vs cube M-mode echo, 44% for cardiovascular magnetic resonance vs Simpson's 2D echo, 39% for radionuclide ventriculography vs Simpson's 2D echo, and smallest at 31% for cardiovascular magnetic resonance-radionuclide ventriculography. Similarly, the end-diastolic volume and end-systolic volume by 2D echo and cardiovascular magnetic resonance revealed wide limits of agreement (52 ml to 216 ml and 11 ml to 188 ml, respectively). CONCLUSION These results suggest that ejection fraction measurements by various techniques are not interchangeable. The conclusions and recommendations of research studies in heart failure should therefore be interpreted in the context of locally available techniques. In addition, there are very wide variances in volumes and ejection fraction between techniques, which are most marked in comparisons using echocardiography. This suggests that cardiovascular magnetic resonance is the preferred technique for volume and ejection fraction estimation in heart failure patients, because of its 3D approach for non-symmetric ventricles and superior image quality.