Background: Computerized quantitative coronary angiography (QCA) has fundamentally altered our approach to the assessment of coronary interventional techniques and strategies aimed at the prevention of recurrence and progression of stenosis. It is essential, therefore, that the performance of QCA systems, upon which much of our scientific understanding has become integrally dependent, is evaluated in an objective and uniform manner.
Methods and results: We validated 10 QCA systems at core laboratories in North America and Europe. Cine films were made of phantom stenoses of known diameter (0.5 to 1.9 mm) under four experimental conditions: in vivo (coronary arteries of pigs) calibrated at the isocenter or by use of the catheter as a scaling device and in vitro with 50% contrast and 100% contrast. The cine films were analyzed by each automated QCA system without observer interaction. Accuracy and precision were taken as the mean and SD of the signed differences between the phantom stenoses, and the measured minimal luminal diameters and the correlation coefficient (r), the SEE, the y intercept, and the slope were derived by their linear regression. Performance of the 10 QCA systems ranged widely: accuracy, +0.07 to +0.31 mm; precision, +/- 0.14 to +/- 0.24 mm; correlation (r), .96 to .89; SEE, +/- 0.11 to +/- 0.16 mm; intercept, +0.08 to +0.31 mm; and slope, 0.86 to 0.64.
Conclusions: There is a marked variability in performance between systems when assessed over the range of 0.5 to 1.9 mm. The range of accuracy, intercept, and slope values of this report indicates that absolute measurements of luminal diameter from different multicenter angiographic trials may not be directly comparable and additionally suggests that such absolute measurements may not be directly applicable to clinical practice using an on-line QCA system with a different edge detection algorithm. Power calculations and study design of angiographic trials should be adjusted for the precision of the QCA system used to avoid the risk of failing to detect small differences in patient populations. This study may guide the fine-tuning of algorithms incorporated within each system and facilitate the maintenance of high standards of QCA for scientific studies.