Observer variability in evaluating extracranial carotid artery stenosis.

One hundred twenty eight cervical carotid arteriograms were twice viewed by three readers for the evaluation of atherosclerotic disease at the carotid bifurcation. Stenoses were estimated using calipers to the nearest 5% and lesions were qualitatively characterized as smooth, irregular, or ulcerated. The intraobserver correlation coefficient between estimates of percent stenosis was .94 overall and .98 for the internal carotid artery. The average intraobserver variability in estimating percent stenosis was 5.23% for all vessels and 6.04% with a standard deviation of 8.09% for the internal carotid artery. The intraobserver percent agreement at a fixed stenosis is defined as the percent of the time one reader on two readings would read at least the fixed percent stenosis among cases that might be read as having the fixed percent stenosis. The intraobserver percent agreement rate for the internal carotid artery was 95.9% at greater than 0% stenosis, 90.4% for 50% or greater stenosis, and 96.8% for 100% stenosis (total occlusion). The interobserver correlation coefficient between readers was .92 overall and .97 for the internal carotid artery. The absolute difference in percent stenosis between readers was 7.21% for all vessels and 8.64% for the internal carotid artery with a standard deviation of 9.5%. The interobserver agreement rate for the internal carotid artery at greater than 0% stenosis was 93.0%, 85.4% for 50% or greater stenosis and 96.8% at 100% stenosis. The addition of oblique views had no statistical effect on estimates of percent stenosis but increased the frequency with which irregularity and ulceration were diagnosed in the internal carotid artery.