Observer variance in the qualitative evaluation of left ventricular wall motion and the quantitation of left ventricular ejection fraction using rest and exercise multigated blood pool imaging.

Abstract

Multigated blood-pool imaging (MBPI) at rest and with exercise is widely used for the evaluation of left ventricular regional wall motion and ejection fraction. Because the precision of these tests depends on interobserver and intraobserver variations in interpretation, the following study was performed. Patients (59) had MBPI at rest and during peak supine bicycle exercise and left ventriculography (LV gram) at rest during cardiac catheterization. Patients (49) had significant coronary artery disease and 10 did not. Rest MBPI, exercise MBPI and LV gram regional wall motion were graded by 3 independent observers. For scoring purposes the left ventricular wall was subdivided into anterolateral, apical, inferior, septal, apical-inferior and posterior walls. Wall motion for the tracer and contrast studies was scored subjectively on a 5-point scale: 3 = normal, 2 = mild hypokinesis, 1 = moderate-to-severe hypokinesis, 0 = akinesis, -1 = dyskinesis. Ejection fraction was determined by 2 independent observers 6 times for rest and exercise MBPI using a counts technique that allowed frame-by-frame variations in the regions of interest. Ejection fraction was determined by 2 independent observers 3 times for the LV gram using an area-length method. Results were analyzed by a 2-way analysis of variance and expressed as .+-.2 SD. Using the 5-point scoring system, the estimated interobserver variance for regional wall motion scores ranged from .+-.0.60 to .+-.1.02 grade for the rest MBPI, .+-.0.94 to .+-.1.46 grade for the exercise MBPI and .+-.0.66 to .+-.0.98 grade for the LV gram, depending on the wall analyzed. The estimated interobserver variance for regional wall motion scores ranged from .+-.0.56 to .+-.1.08 grade for a change between rest and exercise MBPI, depending on the wall analyzed. Interobserver variance for the rest MBPI was greater than that for the LV gram for only the septal wall (P < 0.01). Intraobserver and interobserver variance for ejection fraction determinations were .+-.5.8 and .+-.6.0%, respectively, for the rest MBPI, .+-.9.2 and .+-.9.6% for the exercise MBPI and .+-.11.0 and .+-.11.4% for a change in ejection fraction between rest and exercise MBPI. When a single observer determined the ejection fraction twice for both the rest and the exercise MBPI, and then compared the averaged rest with the averaged exercise ejection fraction, the intraobserver variance was reduced to .+-.4.6%. Intraobserver and interobserver variance for the LV gram ejection fraction were .+-.6.0 and .+-.11.6%. Although intraobserver variance for the rest MBPI and LV gram ejection fraction were not significantly different, interobserver variance for the rest MBPI ejection fraction was significantly less than that for the LV gram (P < 0.005). Interobserver variance for rest MBPI regional wall motion, and interobserver and intraobserver variance for rest MBPI ejection fraction, are comparable to those for the LV gram, except for the septal wall. The results offer objective criteria by which exercise-induced changes in left ventricular regional wall motion and ejection fraction can be interpreted. To minimize the error due to observer variance, exercise-induced changes in ejection fraction should be determined by comparing averaged ejection fractions derived from at least 2 determinations for both the rest and the exercise MBPI.