Software Discrepancies in Radionuclide-Derived Left Ventricular Ejection Fraction.

Abstract

Gated equilibrium radionuclide angiography (ERNA), or multigated acquisition scanning, is a well-established technique to monitor left ventricular ejection fraction (LVEF) in patients treated with potentially cardiotoxic chemotherapy. To determine the results of a true change in LVEF, low inter- and intrareader variability is important. The aim of this study was to investigate inter- and intrareader variability in LVEF measurements using 2 different commercially available software packages with cardiac MR (CMR) as a reference standard. Methods: In 46 ERNA scans, LVEF was measured by 2 experienced nuclear medicine technologists, using the 2 software packages Xeleris and Corridor4DM. All patients had CMR performed within 1.5 h from ERNA. CMR-derived LVEF was measured by a cardiologist using cvi42 software. Eight patients were reanalyzed to investigate intrareader variability. Bland-Altman analysis was used to assess agreement between readers and software. Repeated-measures ANOVA was used to assess interactions between readers and software. Differences in mean LVEF were compared using a t test. The Lin concordance correlation coefficient (CCC) was used to test LVEF agreement between software packages and readers and the reference CMR results. Results: Corridor4DM had a significantly higher mean LVEF than did Xeleris. No significant interreader difference was observed within the same software. ANOVA found that readers did not influence LVEFs. The CCC between software packages was similar for both readers, at 0.409 for reader 1 and 0.418 for reader 2. Both software packages showed a significant LVEF bias compared with CMR (4% for Xeleris vs. 11% for Corridor4DM). For both readers, the CCC for correlation with MRI was higher for Xeleris (0.438/0.572) than for Corridor4DM (0.257/0.244). Conclusion: A high degree of variability was found between the 2 different software packages for the calculation of LVEF. No significant difference in LVEF was found between readers using the same software. Corridor4DM gave higher LVEF estimates than did Xeleris. Our findings suggest that different software programs for assessing LVEF in ERNA examinations are not interchangeable. The utmost caution is recommended if switching between different types of software.