The limitations of somatosensory evoked potentials during aortic occlusion stimulated us to evaluate the sensitivity and specificity of spinal (T10 and L4) evoked potentials (SpEPs) in predicting neurologic function after aortic occlusion. Thirty-six swine were assigned randomly to three equal groups (Group 1, control; Group 2, cerebrospinal fluid (CSF) drainage; Group 3, CSF drainage with 20-mg intrathecal papaverine). After induction of anesthesia and initiation of physiologic monitoring, a left-sided thoracotomy was performed to provide access to the descending aorta. SpEPs were generated by stimulating the thoracic spinal cord and recording the conducted response at the T10 and L4 level. After baseline measurements were recorded, the descending aorta was occluded 1 cm distal to the left subclavian artery. SpEPs were recorded every 2.5 min and physiologic variables every 5 min. The aorta was unclamped 10, 15, or 20 min after loss of the L4 SpEP. If the L4 SpEP was not lost, the aortic occlusion interval was terminated at 90 min. Attenuation of the SpEPs occurred earlier at the L4 level. Group 1 experienced the earliest loss of the L4 SpEP (18.3 +/- 7.8 min, P < 0.005). Loss of the L4 SpEP in Group 2 (49.3 +/- 27.8 min) was earlier than in Group 3 (73.7 +/- 26.1 min, P < 0.05). Early postoperative motor function (modified Tarlov scale) correlated with time from loss of the L4 SpEP until reperfusion of the distal aorta (r = 0.93). The sensitivity, specificity, and accuracy of the L4 SpEP in predicting neurologic dysfunction was 92.8% (13 abnormal/14 predicted), 90.9% (20/22), and 91.7% (33/36).(ABSTRACT TRUNCATED AT 250 WORDS)