The Multi-Hospital Eastern Atlantic Restenosis Trial group obtained follow-up angiography in 510 patients with 598 successfully dilated coronary lesions who were enrolled in a controlled trial of the effects of a single dose of 1 g of methylprednisolone on restenosis after coronary angioplasty. The overall restenosis rate was 39.6%. The strongest univariate relations to the restenosis rate were found for lesion location (saphenous vein graft, 68%; left anterior descending artery, 45%; left circumflex artery and right coronary artery, 32%; p = 0.002); lesion length (less than or equal to 4.6 mm, 33%; greater than 4.6 mm, 45%; p = 0.001); percent stenosis before angioplasty (less than or equal to 73%, 25%; greater than 73%, 43%; p = 0.005), percent stenosis after angioplasty (less than or equal to 21%, 33%; greater than 21%, 46%; p = 0.017) and arterial diameter (less than 2.9 mm, 44%; greater than or equal to 2.9 mm, 34%; p = 0.036). Two multivariate models to predict restenosis probability were developed with use of stepwise logistic regression. The preprocedural model, which included only variables whose values were known before angioplasty, entered lesion length, vein graft location, left anterior descending artery location, percent stenosis before angioplasty, eccentric lesion and arterial diameter. The postprocedural model, which also included variables whose values were known after angioplasty was performed, was similar to the preangioplasty model except that it also entered postangioplasty percent stenosis and "optimal" balloon sizing but did not enter eccentric lesion. These data indicate that the probability of restenosis after angioplasty is determined predominantly by the characteristics of the lesion being dilated. They are consistent with the known intimal proliferative mechanism of restenosis, offer a means of identifying lesions at unusually high or low risk of restenosis, and of predicting the likelihood that a particular lesion will restenose after angioplasty and provide a rationale for stratification by restenosis probability in the design of future studies of restenosis.