Computer-assisted quantitative coronary arteriography (QCA) has gained widespread acceptance in assessing changes in coronary dimensions over time, but little is known about the utility of QCA in patients having undergone coronary bypass surgery. As a validation study, we analyzed the accuracy and precision of QCA in a subset of the baseline angiograms of a clinical trial in 395 post-bypass men with low HDL cholesterol concentrations who have been randomized to receive double-blind gemfibrozil or placebo for 2 1/2 years. Based on repeat measurements of the same cineframe, the average diameter of a segment (ADS) had a mean coefficient of variation (CV) of 3.1%. The mean CVs of the minimum luminal diameter (MLD), percent diameter stenosis (PDS) and stenotic flow reserve of an obstruction were 8.6, 10.2 and 9.8%, respectively, but the area of the atherosclerotic plaque had an unacceptably high CV, 24.0%. When the measurements from two contrast injections into a native coronary artery during the same angiographic session were compared, precision (standard deviation of the differences) was 0.198 mm for ADS, 0.192 mm for MLD, and 7.37% for PDS. Variability was not substantially reduced when measurements from 3 or 5 consecutive cineframes were averaged. Comparable repeatability was found when venous bypass grafts were imaged twice, whether the grafts themselves or the grafted native vessels were analyzed. We conclude that QCA has an acceptable accuracy and precision in analyzing coronary dimensions in bypass-grafted patients. A change of 0.40 mm in ADS and MLD, and 20% in PDS represent true progression or regression of coronary atherosclerosis with more than 95% confidence.