Long-term outcome of stents versus bypass surgery in diabetic and nondiabetic patients with multivessel or left main coronary artery disease: a pooled analysis of 5775 individual patient data.

BACKGROUND Coronary artery bypass grafting and percutaneous coronary intervention (PCI) are alternative treatments for complex multivessel or left main disease. However, the relative treatment effects in diabetic and nondiabetic patients remain uncertain. METHODS AND RESULTS We performed a pooled analysis of 5775 patients from 3 clinical studies comparing PCI and coronary artery bypass grafting for multivessel or left main disease and compared adverse outcomes (death; a composite outcome of death, Q-wave myocardial infarction, or stroke; and repeat revascularization) according to the diabetic status. Over a median follow-up of 5.5 years, risk-adjusted mortality after PCI relative to coronary artery bypass grafting was not different in diabetic (hazard ratio [HR], 1.15; 95% CI, 0.88-1.51) and nondiabetic (HR, 1.15; 95% CI, 0.88-1.50) patients. The adjusted risks of the composite outcome of death, Q-wave myocardial infarction, or stroke were also not different in diabetic (HR, 1.00; 95% CI, 0.79-1.26) and nondiabetic (HR, 0.99; 95% CI, 0.78-1.26) patients. However, PCI was significantly associated with higher risk of repeat revascularization both in diabetic (HR, 3.56; 95% CI, 2.62-4.83) and in nondiabetic (HR, 3.55; 95% CI, 2.61-4.83) patients. Significant interactions were absent between diabetic status and revascularization strategies for death (P=0.27), composite outcome of death, Q-wave myocardial infarction, or stroke (P=0.97), and repeat revascularization (P=0.08). CONCLUSIONS For multivessel or left main disease, the long-term risks of mortality and composite serious outcomes were not different between PCI and coronary artery bypass grafting in diabetic and nondiabetic patients, but PCI was associated with higher risk of repeat revascularization. These relative treatment effects were not modified by diabetic status. These results are hypothesis generating and should be addressed in a randomized trial.