Accuracy of individual and panel visual interpretations of coronary arteriograms: implications for clinical decisions

The visual interpretation of coronary arteriograms by individuals has been shown to be variable and inaccurate. To determine whether observer accuracy improves with experience or with use of the mean values obtained from a panel of observers, the visual readings of percent diameter stenosis and "normal" reference segment diameter were compared with the quantitative analyses of 13 randomly chosen coronary stenoses. Visual interpretation was also performed on cineangiograms of seven phantom stenoses ranging in severity from 17% to 83%. Repeated quantitative arteriography demonstrated good intraobserver variability for minimal stenosis diameter (r = 0.91, SD = 0.23 mm) and percent diameter stenosis (r = 0.93, SD = 6.4%). When the mean of the repeated quantitative analyses was used as the standard, visual interpretations of percent diameter stenosis were found to have considerable inaccuracy (r = 0.78, SD = 14.5%). Phantom percent diameter stenosis data were better correlated (r = 0.85), but accuracy remained poor (SD = 17.8%). Fifty percent narrowings were read over a range from 30% to 95%. Substantial inaccuracies were also found for observer assessment of normal reference segment diameter (r = 0.75, SD = 0.75 mm). Observer accuracy of percent stenosis did not correlate with prior angiographic experience but was progressively improved by taking the mean value of the interpretations of three and five experienced angiographers (r = 0.88, 0.89; SD = 11.3%, 8.3%, respectively). These findings suggest that arteriographic interpretations accurate enough for interventional decisions can only be obtained using quantitative arteriography or the mean value of data from a large panel of angiographers.