OBJECTIVE Salvage of thrombosed prosthetic dialysis shunts can be performed using surgical or endovascular techniques. A prospective randomized trial was designed to compare the efficacy of these two methods in restoring dialysis access function. METHODS One hundred fifteen patients with thrombosed dialysis shunts were randomized prospectively to surgical (n = 56) or endovascular (n = 59) therapy. In the surgical group, salvage was attempted with thrombectomy alone in 22% and with thrombectomy plus graft revision in 78%. In the endovascular group, graft function was restored with mechanical (82%) or thrombolytic (18%) graft thrombectomy followed by percutaneous angioplasty. RESULTS Stenosis limited to the venous anastomotic area was the cause of shunt thrombosis in 55% of patients, and long-segment venous outflow stenosis or occlusion was the cause in 30%. In 83% of the surgical group and in 72% of the endovascular group, graft function was immediately restored (p = NS). The postoperative graft function rate was significantly better in the surgical group (p < 0.05). Thirty-six percent of grafts managed surgically remained functional at 6 months and 25% at 12 months. In the endovascular group, 11% were functional at 6 months and 9% by 12 months. Patients with long-segment venous outflow stenosis or occlusion had a significantly worse patency rate than those with venous anastomotic stenosis (p < 0.05). CONCLUSIONS Neither surgical nor endovascular management resulted in long-term function for the majority of shunts after thrombosis. However, surgical management resulted in significantly longer primary patency in this patient population, supporting its use as the primary method of management in most patients in whom shunt thrombosis develops.