To determine the importance of predilatation stenosis morphology on the risk of restenosis after percutaneous transluminal coronary angioplasty (PTCA), 500 procedures were randomly chosen for analysis from 3,839 consecutive successful PTCA procedures. Angiographic follow-up was available for 308 patients (62%) at a mean of 7.3 +/- 3.4 months after PTCA. One dilated site was randomly chosen per procedure. One hundred and one sites had documented restenosis (greater than or equal to 50% mean diameter stenosis from multiple projections) by quantitative angiography (33% of all sites restudied and 20% of all sites dilated). Twenty-eight morphologic variables and 20 other angiographic, clinical and procedural variables were analyzed by an observer blinded to clinical outcome. Univariate analysis found post-PTCA percent stenosis greater than 30% (p = 0.005), bend point location (p = 0.01), post-PTCA gradient greater than 15 mm Hg (p = 0.02), angina class III to IV (p = 0.03), age (p = 0.04) and the absence of dissection (p = 0.04) to predict restenosis. Multivariate analysis found only 2 significant (p less than 0.05) independent predictors of restenosis: post-PTCA percent stenosis greater than 30% and bend point location. Restenosis occurred in 41% of lesions located at an end-diastolic vessel angle greater than or equal to 45 degrees compared with 28% in lesions on lesser bends. Thus, only 1 predilatation morphologic characteristic, stenosis location at a bend point, was an important independent predictor of restenosis, and should be considered when assessing patients for PTCA.