The association between calcification of the coronary arteries and coronary artery narrowing is well established. However, fluoroscopic visualization of coronary calcifications has been insufficiently sensitive to be useful as a screening test. Since digitization of radiographic images permits the subtraction of noncardiac structures from moving cardiac structures, such subtraction might increase the sensitivity of coronary fluoroscopy. To determine whether coronary calcifications were better visualized with digital subtraction fluoroscopy than with conventional fluoroscopy, we taped diseased human coronary arteries to a pulsating water balloon inside the thorax of a dog cadaver and studied this model with both fluoroscopic techniques. Calcific atherosclerotic plaques were more easily identified with digital subtraction fluoroscopy than with conventional fluoroscopy. We tested the method clinically by submitting 191 subjects without history or electrocardiographic evidence of previous myocardial infarction who were referred for coronary arteriography to both fluoroscopic studies. For at least one, at least two, and three calcified coronary arteries, digital fluoroscopy was more sensitive (92%, 66%, and 40%) than conventional fluoroscopy (63%, 21%, and 2%) (all p less than .001) for the prediction of significant coronary obstructions (greater than 50%). Although digital fluoroscopy was less specific than conventional fluoroscopy (digital: 65%, 89%, and 97%; conventional: 81%, 98%, and 100%) (all but last, p less than .01), receiver operating curve analysis revealed a significantly larger area under the curve, indicating higher accuracy for the digital technique (p = .03). Digital subtraction fluoroscopy was more accurate in younger than in older patients.(ABSTRACT TRUNCATED AT 250 WORDS)