Purpose: This report summarizes our experience with epidural cooling (EC) to achieve regional spinal cord hypothermia and thereby decrease the risk of spinal cord ischemic injury during the course of descending thoracic aneurysm (TA) and thoracoabdominal aneurysm (TAA) repair. Methods: During the interval July 1993 to Dec. 1995, 70 patients underwent TA (n = 9, 13%) or TAA (n = 61) (type I, 24 [34%], type II, 11 [15%], type III, 26 [37%]) repair using the EC technique. The latter was accomplished by continuous infusion of normal saline (4°C) into a T11-12 epidural catheter; an intrathecal catheter was placed at the L3-4 level for monitoring of cerebrospinal fluid temperature (CSFT) and pressure (CSFP). All operations (one exception, atriofemoral bypass) were performed with the clamp-and-sew technique, and 50% of patients had preservation of intercostal vessels at proximal or distal anastomoses (30%) or by separate inclusion button (20%). Neurologic outcome was compared with a published predictive model for the incidence of neurologic deficits after TAA repair and with a matched (Type IV excluded) consecutive, control group (n = 55) who underwent TAA repair in the period 1990 to 1993 before use of EC. Results: EG was successful in all patients, with a 1442 ± 718 ml mean (range, 200 to 3500 ml) volume of infusate; GSFT was reduced to a mean of 24°± 3°C during aortic cross-clamping with maintenance of core temperature of 34°± 0.8 ± C. Mean CSFP increased from baseline values of 13 ± 8 mm Hg to 31 ± 6 mm Hg during cross-clamp. Seven patients (10%) died within 60 days of surgery, but all survived long enough for evaluation of neurologic deficits. The EC group and control group were well-matched with respect to mean age, incidence of acute presentations/aortic dissection/aneurysm rupture, TAA type distribution, and aortic cross-clamp times. Two lower extremity neurologic deficits (2.9%) were observed in the EC patients and 13 (23%) in the control group (p < 0.0001). Observed and predicted deficits in the EC patients were 2.9% and 20.0% (p = 0.001), and for the control group 23% and 17.8% (p = 0.48). In considering EC and control patients (n = 115), variables associated with postoperative neurologic deficit were prolonged (>60 min) visceral aortic cross-clamp time (relative risk, 4.4; 95% CI, 1.2 to 16.5; p = 0.02) and lack of epidural cooling (relative risk, 9.8; 95% CI, 2 to 48; p = 0.005). Conclusion: EC is a safe and effective technique to increase the ischemic tolerance of the spinal cord during TA or TAA repair. When used in conjunction with a clamp-and-sew technique and a strategy of selective intercostal reanastomosis, EC has significantly reduced the incidence of neurologic deficits after TAA repair.