Quantitative coronary angiography has been proposed as a means of reducing observer variability in the interpretation of coronary angiograms, especially before and after percutaneous transluminal coronary angioplasty (PTCA). Analysis of 13 consecutively acquired biplane digital subtraction angiograms before and after PTCA was undertaken to determine intra- and interobserver variability of absolute lesion diameter, relative videodensitometric cross-sectional area, automated percent diameter stenosis and visual percent diameter stenosis using a new fully automated quantitative computer program. The reliability of single-view measurements was also assessed. Both before and after PTCA, measures of absolute diameter showed less interobserver variability than densitometry, percent automated diameter stenosis and percent visual diameter stenosis measurements (before, r = 0.95, 0.83, 0.86, 0.70; after, 0.95, 0.88, 0.81, 0.62, respectively). Relative videodensitometric cross-sectional area correlated poorly with images from the orthogonal view (r = 0.46). These data suggest that quantitative angiography reduces variability from visual estimates; of all quantitative angiographic measurements, the highest interobserver reproducibility is achieved using absolute lesion diameter both before and after PTCA, probably because no operator interaction is needed to identify a "normal" segment. Unselected, single-view quantitative arteriography is poorly reproducible using videodensitometry. Therefore, automated determination of absolute lesion diameter in at least 2 projections provides the most reproducible evaluation of coronary lesions both before and after PTCA.