The reliability of interpretation of coronary arteriography as a diagnostic tool was investigated in a sub-study of the VA Cooperative Study of Surgical Treatment for Coronary Arterial Occlusive Disease. Twenty-two physicians with varying levels of experience read 13 cine angiograms -- blind -- on two different occasions. Analysis of inter- and intraobserver variability showed that angiographic items about which observers were most inconsistent from one reading to the other had the largest interobserver disagreement as well. They were the distal portions of the left anterior descending and left circumflex arteries. Among the items on which there was most consistent agreement -- namely, the right main coronary artery and presence of ventricular aneurysm -- there was most often agreement between observers as well. When individual readers were evaluated, some observers were far more consistent in their own readings of all the angiographic items than others. This intraobserver agreement in turn correlated fairly well with how often they agreed with the other observers and with how much experience they reported having in reading coronary cineangiograms.
Detre et al. (1975) studied this question.