Limited data are available on the relative effect of drug-eluting versus bare-metal stents on the requirement for subsequent coronary artery bypass grafting (CABG). The aim of this study was to evaluate the incidence and predictors of CABG after bare-metal and paclitaxel-eluting coronary stent implantation. A patient-level, pooled analysis was conducted of 2,736 patients from 3 double-blind, randomized trials comparing the slow-release paclitaxel-eluting Taxus stent with an otherwise identical bare-metal stent control in single de novo coronary lesions, with 5-year follow-up. The rate of target lesion revascularization by CABG (TLR-CABG) was reduced from 4.1% in patients with bare-metal stents to 1.4% in those with Taxus stents (p <0.001). The use of the Taxus stent was the strongest predictor of freedom from TLR-CABG on multivariate analysis (hazard ratio 0.33, p <0.001). Significant reductions in TLR-CABG with Taxus compared with bare-metal stents were seen in the treatment of left anterior descending artery lesions (6.1% vs 1.8%, p <0.001) and non-left anterior descending artery lesions (2.8% vs 1.3%, p = 0.037), in patients with diabetes (6.0% vs 1.0%, p <0.01), and in those without diabetes (3.5% vs 1.6%, p <0.01). In conclusion, referral to CABG is significantly less common after stenting single coronary lesions with Taxus compared with bare-metal stents. The relative reductions in TLR-CABG of 54% in patients without diabetes, 87% in patients with diabetes, 70% in left anterior descending artery lesions, and 54% in non-left anterior descending artery lesions with Taxus compared with bare-metal stents should be considered during stent selection.