A comparison of quantitative computerized and human panel coronary endpoint measures: implications for the design of...

The Monitored Atherosclerosis Regression Study was a double-blind, 2-year, placebo-controlled, randomized, serial angiographic trial which tested reduction of low density lipoprotein-cholesterol with monotherapy using lovastatin on the progression of coronary atherosclerosis. Angiographic outcome was evaluated both by a panel of human readers who visually inspected matched film pairs to arrive at a global change score and by automated computerized vessel edge finding and lesion measurement (quantitative coronary angiography, QCA). In this paper, we model the association between QCA measures of coronary artery lesion change and the panel-based global change score. QCA measures included: per-patient changes in percent diameter stenosis and minimum lumen diameter averaged over all lesions; per-patient changes in average diameter and percent involvement averaged over all segments; the numbers of progressing and regressing lesions and new total occlusions; and the development of any new lesions. We found that when evaluating coronary artery lesion change, panelists evaluate changes in percent diameter stenosis for both low grade (< 50% diameter stenosis at baseline) and high grade (> or = 50% diameter stenosis at baseline) lesions as well as new total occlusions and the number of progressing lesions. Although computerized quantification of the size of a lesion at baseline and as an endpoint may be a more precise measure than that by human panel interpretation, QCA fails to incorporate many other important aspects of coronary angiographic change visualized over the entire coronary artery tree by a panel of human interpreters. Thus, the global change score provides a "multiple endpoint" for coronary angiographic trials which does not suffer from the problems of statistical analysis and interpretation of multiple hypothesis tests which usually accompany true multiple endpoint measures. Choice of either or both endpoints in preparing angiographic trials depends on careful consideration of the desired information as well as the cost of carrying out the endpoint analysis.