OBJECTIVES Low patency is a major concern when using separate tube grafts for intercostal artery reconstruction. Our goal was to elucidate the optimal size and length of grafts from their patency and the computational fluid dynamics (CFD). METHODS The patency, size and length of separate tube grafts were evaluated in 41 patients. Simulation of CFD was performed in a model derived from a patient with a patent 12-mm graft that was 15 mm long, with 2 simulation models with a smaller (8-mm) or longer (30-mm) graft. RESULTS A total of 49 grafts were used for intercostal artery reconstruction. There was 1 in-hospital death and 2 spinal cord injuries. The patency rate, which could be evaluated in 46 grafts, was 63% (29/46). It was 71% (24/34) in thoracoabdominal aortic replacement and 42% (5/12) in descending aortic replacement. Among 14 patients in whom all grafts were occluded, no patients developed spinal cord injury. All grafts longer than 25 mm were occluded (n = 5). Eight- and 10-mm grafts showed better patency than 12-mm grafts in thoracoabdominal aortic replacement (P = 0.008) when grafts were shorter than 25 mm. Simulation of CFD revealed vortical flow within the 12-mm graft, which did not reach the intercostal orifice, whereas helical flow was maintained throughout the cardiac cycle within the 8-mm graft. CONCLUSIONS Eight- and 10-mm grafts seemed better than 12-mm grafts, and grafts should be kept shorter than 25 mm. Simulation of CFD may shed light on the issue of the optimal intercostal artery reconstruction technique.