Long-term safety and efficacy of drug-eluting stents remains controversial. The CREDO-Kyoto registry cohort-2 is a physician-initiated non-company sponsored multi-center registry enrolling consecutive patients undergoing first coronary revascularization in 26 centers in Japan. We compared 3-year outcome between patients treated with sirolimus-eluting stent (SES) only (5092 patients) and bare-metal stent (BMS) only (5405 patients). SES-use as compared with BMS-use was associated with significantly lower adjusted risk for all-cause death [hazard ratio (HR) [95% confidence interval (CI)] 0.72 (0.59–0.87), P = 0.0007], which was mainly driven by the reduction in non-cardiac death [HR (95% CI) 0.64 (0.48–0.85), P = 0.002]. The risk of cardiac death [HR (95% CI) 0.82 (0.63–1.07), P = 0.15], myocardial infarction [HR (95% CI) 0.73 (0.51–1.03), P = 0.07] and definite stent thrombosis [HR (95% CI) 0.62 (0.35–1.09), P = 0.1] was not different between the two groups. Despite longer duration of thienopyridine administration, SES-use was associated with significantly lower risk for bleeding [HR (95% CI) 0.75 (0.6–0.95), P = 0.02] and similar risk for stroke [HR (95% CI) 1.0 (0.75–1.34), P = 1.0]. The risk for target-lesion revascularization (TLR) was markedly lower in the SES group [HR (95% CI) 0.42 (0.36–0.48), P < 0.0001]. The direction and magnitude of the effect of SES relative to BMS in patients presenting acute myocardial infarction (AMI) were similar to those in patients presenting otherwise. In conclusion, SES-use as compared with BMS-use was associated with marked reduction of TLR without any increases in death, myocardial infarction, stent thrombosis, stroke and bleeding in real world clinical practice regardless of clinical presentation including AMI.