Restenosis prediction from published studies is hampered by inadequate sample size and incomplete angiographic follow-up. The prediction of restenosis with the existing variables is poor. The aim of the present study was to include the clinical and angiographic variables commonly associated with angiographic restenosis and develop a prediction model for restenosis from the PRESTO database.Methods and Results—
This study included 1312 patients with a single lesion enrolled in the angiographic substudy of the PRESTO trial. We constructed 2 risk scores. The first used preprocedural variables (female gender, vessel size [≤2.5 mm, 2.5 to 3 mm, 3 to 3.5 mm, 3.5 to 4 mm, >4 mm], lesion length >20 mm, diabetes, smoking status, type C lesion, any previous percutaneous coronary intervention [PCI], and unstable angina) derived from previous studies. Estimated restenosis rates and corresponding variability for each possible level of the resultant risk score were obtained via bootstrapping techniques. The area under the receiver-operator characteristic (ROC) curve was 0.63, indicating modest discriminatory ability to predict restenosis. The second approach constructed a multiple logistic regression model considering significant univariate clinical and angiographic predictors of restenosis identified from the PRESTO database (treated diabetes mellitus, nonsmoker, vessel size, lesion length, American College of Cardiology/American Heart Association type C lesion, ostial location, and previous PCI). The area under the ROC curve for this risk score was also 0.63.Conclusions—
The preprocedural clinical and angiographic variables from available studies and from the PRESTO trial have only modest predictive ability for restenosis after PCI.