Reproducibility in the analysis of multigated radionuclide studies of left ventricular ejection fraction

The authors determined the reproducibility (the standard deviation [SD]) in the analysis of multigated radionuclide studies of left ventricular ejection fraction (LVEF). Radionuclide studies from a consecutive series of 38 patients suspected of ischemic heart disease were analyzed independently by four nuclear medicine physiologists and four laboratory technicians. Each study was analyzed three times by each of the observers. Based on the analyses of the eight observers, the SD could be estimated by the use of a variance component model for LVEF determinations calculated as the average of the analyses of an arbitrary number of observers making an arbitrary number of analyses. This study presents the SDs for LVEF determinations based on the analyses of one to five observers making one to five analyses each. The SD of a LVEF determination decreased from 3.96% to 2.98% when an observer increased his number of analyses from one to five. A more pronounced decrease in the SD from 3.96% to 1.77% was obtained when the LVEF determinations were based on the average of a single analysis made by one to five observers. However, when dealing with the difference between LVEF determinations from two studies, the highest reproducibility was obtained if the LVEF determinations at both studies were based on the analyses made by the same observer. No significant difference was found in the reproducibility of analyses made by nuclear medicine physicians and laboratory technicians. Our study revealed that to increase the reproducibility of LVEF determinations, special efforts should be made to standardize the outlining of the end-systolic region interest.