All patients undergoing resection of thoracic or thoracoabdominal aneurysms at Mount Sinai Hospital since November 1993 had spinal cord function monitored with somatosensory-evoked potentials as part of a multimodality approach to reducing spinal cord injury. In the segment to be resected, each pair of intersegmental vessels was sequentially clamped, and they were subsequently sacrificed only if no change in somatosensory evoked potentials occurred within 8 to 10 minutes after occlusion. Adjunctive protective measures included mild hypothermia (31 degrees to 33 degrees C), distal perfusion, corticosteroids, maintenance of high normal blood pressures, avoidance of nitroprusside, and cerebrospinal fluid drainage. Ninety-five consecutive patients operated on since 1993 (group II) were compared with 138 earlier patients (group I). Preoperative characteristics such as age, sex, etiology of aneurysm, emergency operation, and reoperation did not differ between groups, nor did operative variables such as incidence of rupture and extent of resection. Group I had slightly more smokers and slightly fewer hypertensive individuals. Group II patients had a significantly better outcome with respect to in-hospital mortality (10.5% vs 18%, p = 0.045) and paraplegia (2% vs 8%, p = 0.008). By multivariate analysis, rupture and diabetes were associated with significantly higher in-hospital mortality, and smoking greatly increased the incidence of paraplegia. The extent of the aneurysm was a major determinant of mortality and paraplegia. The low paraplegia rate in group II was achieved without reattachment of a single intercostal or lumbar artery. No patient with fewer than 10 intersegmental arteries severed had paraplegia, and spinal cord ischemia was reversible in three patients after adjunctive maneuvers were performed to improve perfusion, suggesting that spinal cord blood supply is unlikely to depend on a single "artery of Adamkiewicz."