To evaluate intra- and interobserver variability of an on-line quantitative coronary angiographic system, 2 independent observers measured 166 primary lesions excluding total occlusions before and after coronary angioplasty. Each observer repeated his measurement 3 times at 14 days interval. The average percent diameter stenosis results obtained by observer 1 and 2 were almost identical, before (62.2% +/- 12.0% and 62.6% +/- 11.4%, NS) and after (27.1% +/- 12.0% and 26.9% +/- 11.3%, NS) angioplasty. Variability was expressed as 95% limits of agreement (mean difference +/- 2 x SD). The intra-observer variability of observer 1 ranged from -6.6% to 6.6% before angioplasty and from -9.6% to 9.6% after angioplasty. The corresponding limits of observer 2 were -8.0% to 7.5% and -8.3% to 8.5%, respectively. The interobserver variability ranged from -10.4% to 9.6% before versus -12.5% to 13.1% after angioplasty. This variability was not influenced by vessel size. The widening of the limits observed after angioplasty was largely due to an increased variability in the measurements of the absolute minimal luminal diameter but not of the reference segment. We conclude that the intra- and interobserver variability of measurements obtained with an on-line quantitative angiographic system used for guiding coronary interventions is acceptable and without systematic bias in any direction for a wide range of primary coronary stenoses. However, the variability increases when images are acquired immediately after angioplasty.