The efficacy of combined methods of spinal cord protection during thoracoabdominal aortic reconstruction was evaluated because a recent clinical study failed to substantiate the value of cerebrospinal fluid drainage when used alone in the prevention of paraplegia. The effect of cerebrospinal fluid drainage and aortofemoral shunting were analyzed with regard to neurologic outcome and spinal cord blood flow in a model of thoracic aortic occlusion. In addition, we studied the use of motor-evoked potentials as compared with somatosensory-evoked potentials in monitoring cord perfusion. Thirty-two dogs underwent proximal and distal thoracic aortic occlusion for 60 minutes. The control group (n = 8) underwent thoracic aortic cross-clamping only. Spinal cord protection was used in three groups: cerebrospinal fluid drainage alone (n = 8), aortofemoral shunting alone (n = 8), and cerebrospinal fluid drainage and aortofemoral shunting (n = 8). Neurologic outcome improved in all treatment groups as compared with controls (p less than 0.001). The addition of cerebrospinal fluid drainage to aortofemoral shunting did not further improve neurologic outcome. Spinal cord blood flow measured with microspheres in the lumbar gray matter was significantly higher in the dogs with aortofemoral shunting (+/- cerebrospinal fluid drainage) as compared with those with cerebrospinal fluid drainage alone (p less than 0.05) or the controls (p less than 0.001). Aortofemoral shunting also prevented the development of acidosis and hyperglycemia. Loss or changes in amplitude and latency of motor-evoked potentials did not distinguish between the groups. Loss of somatosensory-evoked potentials had a high sensitivity (92%) but lower specificity (68%) in predicting neurologic injury, whereas loss of motor-evoked potentials had a high specificity (100%) but a very low sensitivity (16%). We conclude that cerebrospinal fluid drainage or aortofemoral shunting significantly improve spinal cord blood flow and neurologic outcome. The greatest increase in spinal cord blood flow was seen with aortofemoral shunting, which also prevented metabolic disturbances of reperfusion. Although the addition of cerebrospinal fluid drainage to aortofemoral shunting was the only group in which no neurologic injury occurred, this group did not have a significant improvement in outcome when compared with aortofemoral shunting alone. Spinal cord ischemia was more accurately detected with somatosensory-evoked potentials when aortofemoral shunting was used, whereas motor-evoked potentials recorded from the spinal cord were not sensitive enough to predict neurologic injury.