The advent of graft thrombolysis has provided an objective means for evaluating the etiology of graft occlusion. Over a 10-year period, intra-arterial urokinase (102 cases) or streptokinase (seven cases) was used in 109 infrainguinal conduits (30 autogenous and 79 non-autogenous) that failed 30 days or more after implantation. Thrombolysis was not achieved in 19 additional graft occlusions; these cases were excluded from study because of an inability to define the mechanism of failure. Non-invasive laboratory data were available within 6 months of graft occlusion in 82 (75%) of the cases, with Doppler segmental studies in 80 cases (73%) and duplex ultrasonography studies in 39 cases (36%). Pre-failure non-invasive laboratory abnormalities were detected more frequently in autogenous grafts (21 of 24 patients, 88%), while non-autogenous grafts usually occluded without prior hemodynamic change (11 of 58 patients had abnormalities, 19%) (P < 0.001). Thrombolysis uncovered anatomic defects responsible for thrombosis in 27 (90%) of 30 autogenous grafts compared with only 32 (41%) of non-autogenous conduits (P < 0.001). The most common lesions underlying autogenous graft failure comprised stenoses within the body of the graft (11 cases, 37%), while the most common lesions in failed non-autogenous grafts appeared to be stenoses at an anastomosis (21 cases, 27%). Thus, the mechanisms underlying the late failure of autogenous and non-autogenous grafts differ markedly; autogenous grafts most commonly fail as a result of the gradual development of lesions intrinsic to the graft, while non-autogenous grafts fail precipitously, presumably as a result of some non-anatomic mechanism.