Repeat coronary arteriography was performed in 100 patients with angiographically proven coronary artery disease, because of worsening symptoms (n = 67), persistent stable angina (n = 22) or other reasons (n = 11). The mean interval between the studies was 34.6 months (range 6-99). No patient had interim coronary artery bypass surgery. Progression of coronary artery disease was demonstrated in 60 patients and was correlated in bivariate analysis with the time interval between coronary arteriograms (P less than 0.0001) and with the interim clinical evolution (P less than 0.01). The incidence of progression was similar in the coronary arteries. Moderate lesions (40-70% narrowing) had the highest progression percentages. New significant lesions in previously normal arterial segments were rare. A multivariate analysis of 18 possible predictors revealed three independent variables correlated with progression of coronary artery disease: time interval (P less than 0.0001) and change in functional class (P less than 0.02) between coronary arteriograms, as well as the presence of clinical diabetes (P less than 0.03). Other risk factors and a change of individual risk factors between coronary arteriograms were not correlated with progression of coronary artery disease. Using the results of the multivariate analysis, the evolution of coronary artery disease was correctly predicted in only 72 patients. Following the repeat coronary arteriogram, coronary artery bypass surgery was proposed to 62 patients, 26 of whom needed more distal anastomoses for optimal myocardial revascularization. Because of the important therapeutic implications and the lack of reliable predictors, an aggressive follow-up of patients with known coronary artery disease seems warranted.