BACKGROUND: We used a large patient-level dataset including six angiographic randomized trials (RCTs) on coronary artery bypass conduits to explore incidence and determinants of coronary graft failure. METHODS: Patient-level angiographic data of six RCTs comparing long-term outcomes of the radial artery and other conduits were joined. Primary outcome was graft occlusion at maximum follow-up. The analysis was divided as follows: 1) left anterior descending coronary (LAD) distribution, 2) non-LAD distribution (circumflex and right coronary artery). To identify predictors of graft occlusion, mixed model multivariable Cox regression including all baseline characteristics with stratification by individual trials was used. RESULTS: 1091 patients and 2281 grafts were included (921 left internal mammary arteries, 74 right internal mammary arteries, 710 radial artery and 576 saphenous veins; all left internal mammary arteries were used on the LAD, the other conduits were used on the non-LAD distribution; mean angiographic follow up: 65±29 months). Occlusion rate was 2.3%, 13.5%, 9.4%, 17.5% for the left internal mammary arteries, right internal mammary arteries, radial artery and saphenous veins, respectively. At multivariable analysis type of conduit used, age, female gender, left ventricular ejection fraction<50% and use of the Y graft were significantly associated with graft occlusion in the non-LAD distribution. CONCLUSIONS: Our analyses showed that failure of the left internal mammary arteries to LAD bypass is a very uncommon event. For the non-LAD distribution, the non-use of radial artery, age, female gender, left ventricular ejection fraction<50% and use of the Y graft configuration were significantly associated with mid-term graft failure.