Spinal cord injury in experimental thoracic aortic occlusion: Investigation of combined methods of protection

Transcription

1 Spinal cord injury in experimental thoracic aortic occlusion: Investigation of combined methods of protection James R. Elmore, MD, Peter Gloviczki, MD, C. Michel Harper, Jr., MD, Michael J. Murray, MD, Qing Hua Wu, MD, Thomas C. Bower, MD, Peter C. Pairolero, MD, James M. Naessens, MPH, and Jasper R. Daube, MD, Rochester, Minn. 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 bcclusion for 60 minutes. The control group (n = 8) underwent thoracic aortic cross-damping 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 ~ treatment groups as compared with controls (p < 0.001). The addition of cerebrospina! fluid drainage to aortofemoral shuntingdid 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 (+ cerebrospina! fluid drainage) as compared with those with cerebrospinal fluid drainage alone (p < 0.05) or the controls (p < 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 somatosensoryevoked 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 somatoseusory-evoked potentials when aortofemoral shunting was used, whereas motor-evoked potentials recorded from the spinal cord were not sensitive enough to predict neurologic injury. (J VAsc SURG 1992;15: ) Paraplegia is a devastating complication after reconstruction of the thoracoabdominal aorta. Nu- From the Section of Vascular Surgery (Drs. Elmorc, Gloviczki, Wu, Bower, and Pairolero), Department of Neurology (Drs. Harper and Daube), Department of Anesthesiology (Dr. Murray), and Section of Biostatistics (Mr. Naessens), Mayo Cfinic and Foundation, Rochester. Presented at the Fifteenth Annual Meeting of the Midwestern Vascular Surgical Society, Chicago, Ill., Sept , Reprint requests: Peter Gloviczki, MD, Mayo Clinic, 200 1st St. SW, Rochester, MN /6/36504 merous methods have been used to decrease the incidence of neurologic injury, but none have consistently prevented the development of neurologic complications.l7 Cerebrospinal fluid (CSF) drainage and distal aortic perfusion with partial bypass or aortofemoral shunting are two methods frequently used in an attempt to reduce the incidence of neurologic injury, s,6 Crawford et al.s in a prospective randomized study failed to demonstrate benefit of CSF drainage in patients undergoing thoracoabdominal aortic reconstruction, it is unknown at present 789

2 790 Elmore et al. j~ournal of VASCULAR SURGERY whether CSF drainage combined with any other method is effective at reducing neurologic injury in these patients. Another problem associated with thoracoabdominal reconstruction is the lack of a reliable method to monitor perfusion of the spinal cord during operation. Limitations of somatosensory-evoked potential (SEP) monitoring has prompted interest in motorevoked potential (MEP) monitoring. 9,1 Motorevoked potentials have theoretic advantages over SEPs, since MEPs monitor potentials in the motor pathways and are not limited by peripheral nerve ischemia. Initial experimental experience in our laboratory has identified limitations of MEPs and did not confirm an advantage of MEPs as compared with SEPs. ll This study was therefore designed to investigate combined methods of spinal cord protection and to further evaluate evoked potentials. In particular, we evaluated the effects of CSF drainage with and without aortofemoral shunting on neurologic outcome and spinal cord blood flow. In addition, we evaluated MEPs as compared with SEPs in predicting neurologic outcome. MATERIALS AND METHODS Experimental protocol To be able to perform aortofemoral shunting and CSF drainage in a consistent and technically satisfactory manner, a canine model was chosen for these experiments. The dog model was found suitable for similar experiments in our previous studies, n,~2 Thirty-two mongrel dogs of either sex weighing between 25 and 30 kg were divided into four groups. All animals underwent cross-clamping of the proximal and distal descending thoracic aorta for 60 minutes. The control group consisted of eight animals that underwent thoracic aortic crossclamping without CSF drainage. The three treatment groups included eight dogs that had CSF drainage alone; eight dogs that had aortofemoral shunting alone; and eight dogs that had both CSF drainage and aortofemoral shunting. Operative procedure Anesthesia was induced with intravenous methohexital sodium (10 mg/kg), and the dogs had their tracheas intubated. Ventilation was controlled with a Bird Mark 7 (Bird Products Corp., Palm Springs, Calif.) ventilator, and anesthesia was maintained with halothane (1.0% to 1.5% concentration). Approximately 1500 ml of glucose-free isotonic crystalloid were administered intravenously. Operative proce- dures were done with sterile technique with use of a model described previously, n,12 In brief, indwelling catheters consisted of the following: (1) a right carotid and a right femoral arterial cannula; (2) a pulmonary artery catheter; (3) an intrathecal catheter placed by direct cutdown on the cisterna magna to record pressure and drain CSF (Fig. 1). Cerebrospinal fluid drainage, once instituted, was continued throughout the cross-clamp period. A left fourth intercostal space thoracotomy was performed with dissection of the thoracic aorta 1 cm distal to the origin of the left subclavian artery for proximal clamping. A left eleventh intercostal space thoracotomy was also made to clamp the distal thoracic aorta 1 cm proximal to the celiac artery. A left atrial line was inserted for injection of microspheres. Aortofemoral shunting was performed by use of a temporary shut from the transverse arch to the left common femoral artery (Fig. 1). Cannulation was performed with 16F catheters. Flow was continuously measured in the shunt with an ultrasonic flow probe (Transonic System, Inc., St. Paul, Minn.). Baseline data samples of temperature, blood glucose, arterial blood gases, cardiac output, pulmonary artery pressures, and proximal and distal aortic pressures were recorded. Spinal cord perfusion pressure was calculated as the difference between the mean distal aortic pressure and the CSF pressure. Baseline MEPs, SEPs, and measurement of spinal cord blood flow (SCBF) were done as detailed below. Intravenous heparin (100 units/kg) was given, and the proximal and distal descending thoracic aorta were cross-clamped for 60 minutes. Data sampling and microsphere injection were performed during cross-clamping at 5 minutes and 60 minutes. Acidosis was corrected during cross-clamping by the administration of intravenous sodium bicarbonate (50 to 75 meq/dog). Repeat measurements were made 5 minutes after reperfusion. After injection of the last microsphere sample, the thoracotomy was closed in layers, and air was aspirated from the pleural space. Animals were maintained in a postoperative recovery room and received butorphanol tartrate (10 mg/dose) intramuscularly as an analgesic. After 24 hours neurologic function was evaluated. The dogs were sacrificed with an intravenous overdose of sodium pentobarbital. The animal care complied with the "Principles of Laboratory Animal Care" (formulated by the National Society for Medical Research) and the "Guide for the Care and Use of Laboratory Animals" (NIH publication No , revised 1985).

3 Volttme 15 Number 5 May 1992 Spinal cord ischemia 791 MEP and SEP measurements Spinal cord potentials were recorded with an evoked potential system (TECA TD20, Pleasantville, N.Y.) with use of a technique previously described, n This system delivered a square wave constant voltage stimulus of 300 volts (500 ~sec duration) at a rate of 10 impulses per second for SEPs and 20 impulses per second for MEPs. Recorded potentials were amplified, filtered (20 to 2000 Hz), and averaged for a minimum of 200 impulses per recording. For MEPs, two small burr holes 2 cm apart were made 1 cm lateral to the midline of the skull just posterior to the frontal sinus without entering the dura. The stimulating electrodes consisted of two stainless steel screws placed through the burr holes and onto the dura. The posterior screw served as the anode. For -SEPs the left sciatic nerve was exposed and encircled with a nerve stimulator oriented with the cathode proximally. Recordings of spinal cord potentials were made from monopolar electromyography needles (50 mm length) coated with Teflon placed percutaneously against the bony lamina at the T-2 and T-13 levels. Recordings were made at both levels, with results reported for MEPs at the low thoracic position and for SEPs at the high thoracic level. The T-13 bony landmark in dogs correlates approximately with the T-13 to L-1 spinal cord level. Needle electrodes placed in the subcutaneous tissue 4 cm lateral to each recording electrode served as the reference. A ground electrode was placed in the soft tissue of the neck. Complete neuromuscular blockade for recording of MEPs was achieved after thoracotomy with succinylcholine (1 mg/kg, intravenously) with repeated doses as necessary. Recorded data were analyzed for peak-to-peak amplitude, onset latency, and time until loss of the signal with ischemia and time until return with reperfusion. Recordings were made throughout the experiment every 60 to 90 seconds. Measurements are reported at baseline, 5 minutes, 60 minutes (or last measurable curve after 30 minutes), and again 5 minutes after reperfusion. Spinal cord blood flow Spinal cord blood flow measurements were made with isotope tagged microspheres with use of a method developed by Heymann et al. 13 and previously reported in work from our laboratory. 12 Microspheres ( ~m) were labeled either with 57Co (cobalt), H3Sn (tin), SSSr (strontium), or 46Sc (scandium) obtained from New England Nuclear (Du Pont Diagnostic Imaging Div., North Billerica, Mass.). Approximately microspheres were Carotid arterial pressure line /* ~ Femoral/, ~N~'~.)~,:~o, arteriaj //g" 11/ pressure '~ '/ ii'~ line S CSF drainage catheter and pressur C1 C2 Sagittal section Fig. 1. Experimental design of a canine model of double cross-clamping the thoracic aorta and use of aortofemoral shunting. Catheter is inserted into cisterna magna to monitor pressure and drain CSF. suspended in a well-mixed solution of 10% dextran, and polyoxyethylene sorbitan mono-oleate (Tween- 80; Sigma Chemical Co., St. Louis, Mo.). Before injection the microspheres were agitated in a sonicator and then mixed to a total volume of 10 ml with saline solution. The microspheres were injected in the left atrial line over 10 seconds. A simultaneous reference sample was withdrawn from the carotid arterial line by a withdrawal pump at a constant rate of 5.82 ml/min for 60 seconds. The injection of microspheres was random to prevent ordering bias. At necropsy the spinal cord was removed and divided into cervical, upper thoracic, middle thoracic, lower thoracic, and lumbar regions. Each region was divided into gray and white matter and weighed to the nearest 0.01 gm. Samples were placed in a gamma counter (Beckman 310; Beckman Instruments Inc., Fullerton, Calif.) and counts were made with reference to each of the four microspheres after setting the

4 792 Elmore et al Journal of VASCULAR SURGERY Table I. Neurologic outcome based on Tarlov's scoring system at 24 hours; significant improvement in all treatment groups as compared with controls Aortofemoral Tarlov score Control CSF drainage shunting CSF drainage and aortofemoral shunting 0 (paraplegia) (normal) appropriate window for the energy spectra of each microsphere. Neurologic assessment Neurologic injury was graded by the researchers who were not blinded with regard to the experimental protocol. Functional outcome was assessed at 24 hours and graded according to the method of Tarlovl4: (0)-no movement of hind limbs; (1)- perceptible movement of the joints of the hind limbs; (2)-good movement but unable to stand; (3)-able to stand and walk; (4)-complete recovery. Animals with scores of 0 are paraplegic and animals with scores of 1 to 3 are paraparetic. Histologic analysis Small sections of each segment of the spinal cord were submitted for hematoxylin and eosin staining. Sections were taken after the cord had been kept in formalin for a minimum of 48 hours. The pathologist reviewing the specimens was blinded to each animal's neurologic outcome. Statistical analysis All statistical calculations used two-tailed probability tests with statistical significance based on p _< The four groups of dogs were compared on categorical variables with chi-square analysis, ordinal variables with Kruskal-Wallis rank sum tests, and on continuous variables with analysis of variance tests. Where significant differences existed, twosample t tests were performed between groups, to determine which groups were different. In addition, paired t tests or paired signed rank tests were performed within groups to assess changes in measures at selected times in the experiment. Data are presented as the mean value + the standard deviation. RESULTS Neurologic outcome At 24 hours, seven of eight dogs in the control group were paraplegic. None of the dogs undergoing CSF drainage or aortofemoral shunting were paraplegic. A significantly improved neurologic outcome occurred in all treatment groups as compared with controls (p 0.001). No statistical difference could be identified between the treatment groups with regard to neurologic outcome, but this may reflect aiinsufficient sample size. Only in the group with both CSF drainage and aortofemoral shunting was there no postoperative neurologic injury, but this group did not have a significantly improved neurologic outcome as compared with aortofemoral shunting alone (Table I). Spinal cord blood flow Spinal cord blood flow was significantly improved in dogs undergoing CSF drainage alone as compared with controls. The animals with aortofemoral shunting with or without CSF drainage had significantly greater SCBF as compared with the group with CSF drainage alone or controls. No statistical difference was observed in SCBF between the two groups with aortofemoral shunts. The differences in SCBF were seen during aortic occlusion in the lumbar region of the spinal cord in both the white and gray matter (Fig. 2 displays the results for the lumbar gray matter). At baseline and after reperfusion no significant differences occurred between groups. The SCBFs measured after 5 minutes of reperfusion were not significantly different as compared with baseline within each group (Fig. 2). No statistically significant differences occurred in SCBF in the cervical, upper thoracic, middle thoracic, or lower thoracic SCBFs. Motor-evoked potentials The amplitude and latency of MEPs were not significantly different between groups despite the significant improvement in neurologic outcome and SCBF in the treatment groups as compared with the controls (Table II). The frequency with which the MEPs were lost was not significantly different between groups (Table III); only two dogs lost

5 m Volume 15 Number 5 May 1992 Spinal cord ischemia 793 t- 80 6o [] Control CSF drainage Distal aortic perfusion / :c / lib zoo 40 CG B-2 2O 0 ~ Aortic ~ ~ / i Aortic clamp i ~ cclusi n~l r ~ rem ved 1 Baseline 5-min 60-min 5-min occlusion occlusion repeffusion Fig. 2. Spinal cord blood flow in lumbar gray matter. Data expressed as median blood flow in ml/100 gm/min. Significant improvement in flows during occlusion in animals with CSF drainage or aortofemoral shunts as compared with controls. MEPs during the ischemic interval. In the control group, one dog lost MEPs and was paraplegic, whereas none of the remaining six paraplegic dogs lost MEPs. The second animal that lost MEPs underwent aortofemoral shunting alone and subsequently was able to walk with some paraparesis (Tarlov 3). In both of these dogs the MEPs returned immediately after reperfusion. The overall sensitivity, specificity, and accuracy of the loss of MEPs at predicting neurologic injury (Tarlov 0 to 3) were 16%, 100%, and 69%, respectively. Somatosensory-evoked potentials The amplitude and latencies of SEPs were significantly different between groups (Table II). At 5 minutes of occlusion, the dogs with CSF drainage alone had a significantly reduced amplitude and prolonged latency as compared with the dogs with distal aortic perfusion. No statistical difference could be calculated at 60 minutes of occlusion since all but one of the CSF drainage animals lost SEPs. Five minutes after reperfusion, the control animals had the lowest amplitude and longest latency (Table II). Cerebrospinal fluid drainage animals had significant improvement as compared with the controls in the amplitude at this point, but still had a very prolonged latency. The animals with aortofemoral shunting (+ CSF drainage) had the smallest alterations (p _< 0.05) in the amplitudes and latencies of SEPs. The loss of SEPs varied significantly between the groups (Table III). All eight control dogs and seven of the eight dogs that underwent CSF drainage alone lost SEPs after cross-clamping. In the two groups of dogs undergoing aortofemoral shunting, SEPs were lost in only one of eight dogs in each group. The overall sensitivity, specificity, and overall accuracy of the loss of SEPs at predicting neurologic injury (Tarlov 0 to 3) were 92%, 68%, and 77%, respectively. The time between clamping and the loss of SEPs was not significantly different between the control and CSF drainage groups (Table III). Distal aortic perfusion significantly improved the specificity of SEP monitoring as compared with the dogs with CSF drainage alone. The actual return of SEPs after reperfusion did not correlate with neurologic outcome because all dogs that lost SEPs had return of the signals except for two of the paraplegic animals (Table III). The time until return of the potentials was significantly prolonged in the control group as compared with the CSF drainage group (Table III). Metabolic/hemodynamic parameters Arterial blood gases were not significantly different between groups at baseline and 5 or 60 minutes of aortic occlusion. The Pco 2 of all dogs was 36.7_+3.5 mm Hg, 34.2_+ 3.8 mm Hg, mm Hg, and nun Hg at baseline, 5 minutes of occlusion, 60 minutes of occlusion, and 5 minutes ofreperfusion, respectively. After 5 minutes of reperfusion, the animals not undergoing aortofemoral shunting had a significantly lower ph: control, ph 7.28 _ and CSF drainage alone, ph ; as compared with

6 794 Elmore et al. Journal of VASCULAR SURGERY Table II. Amplitude and latency data for MEPs and SEPs MEP No. in Amplitude Latency No. in Amplitudes MEP group (% of baseline) (msec over baseline) SEP group (% of baseline) Occlusion 5 min Control CSF AFS _ CSF + AFS Occlusion 60 rain Control _ CSF AFS CSF + AFS Repeffusion 5 min Control CSF 8 56 _ AFS CSF + AFS 8 62 _ SEP Latency (msec over baseline) Loss of SEPs _ _ ~ Loss of SEPs Loss of SEPs _ AFS, Aortofemoral shunting. Data represent mean _+ standard deviation. No significant difference between groups for MEPs. 2Significant difference in amplitude or latency of SEPs between groups as compared with controls and/or CSF drainage. Table III. Loss and return of MEPs and SEPs Aortofemoral CSF drainage and Control CSF drainage shunting aortofemoral shunting MEP Loss of potentials (no.) Return of potentials (no.) 1 N/A 1 N/A SEP Loss of potentials (no.) Time till loss (min) N/A N/A Return of potentials (no.) Time till return (min) 28.2 _ _ N/A N/A 2Significant difference as compared with control or CSF treatment groups or both. aortofemoral shunting alone, ph and aortofemoral shunting and CSF drainage, ph Blood glucose levels were not significantly different between groups at baseline, with a value for glucose of all dogs of mg/dl. A significant increase occurred in all dogs to a value of mg/dl during the reperfusion period. The glucose during reperfusion was significantly higher in the control ( mg/dl) and CSF drainage alone groups ( mg/dl) than in the groups with aortofemoral shunting without CSF drainage ( mg/dl) or in combination with CSF drainage ( mg/dl). Cardiac output and index were not significantly different between any group at any time interval. Baseline cardiac output and index for all dogs were l/min and l/min/m 2, respectively. Shunt flows were not significantly different between the two groups of dogs with aortofemoral shunting and averaged _ ml/min. Pulmonary artery pressures increased in all groups with cross-clamping. The mean pulmonary artery diastolic pressure of all dogs was 9.9 _+ 5.1 mm Hg at baseline, which significantly increased to 12.2 _+ 6.5 mm Hg after 5 minutes of occlusion. This increase persisted throughout the cross-clamp period and remained stable 5 minutes after reperfusion, with a mean value of 13.1 _+ 6.3 mm Hg. No correlation occurred between pulmonary artery distal pressures or central venous pressures and neurologic injury. Proximal (carotid) arterial pressure was similar in all groups at baseline. With aortic occlusion, a significant increase was observed in proximal pressures in all groups of dogs compared with their respective baselines. The dogs without aortofemoral shunts had significantly higher proximal aortic pres-

7 Volume 15 Number 5 May 1992 Spinal cord ischemia Proximal mean 12o aortic pressure, 8o mm Hg O Distal mean 6o aortic pressure, mm Hg 40 2o Cerebrospinal 9 fluid pressure, 6 mm Hg 3 o 100 Spinal cord 8o perfusion 60 pressure, 40 mm Hg 20 o AodJc OCClus=on Aotbc clamp removea I ~ I I I I Aofb occlusl ~ ~rern~ved Aortic o~:clusion t~cno~ Aothc clamp Aortic clamp Baseline 5-min 60-min 5-min occlusion occlusion repeffusion f'l Conlrot CSF clrmnage O D~slal ao,llc Derfusion CSF drainage and d=slal aorhc CG B-1A Fig. 3. Hemodynamic pressure measurements before, during, and after thoracic aortic occlusion. Animals with aortofemoral shunts had the lowest proximal mean aortic pressures and the highest distal mean aortic pressures and spinal cord perfusion pressures. No difference in CSF pressure, in animals without CSF drainage, either with or without aortofemoral shunts. sures during aortic occlusion. Distal aortic pressures measured from the femoral artery were not significantly different between groups at baseline. After aortic occlusion, the groups with aortofemoral shunting had significantly greater distal pressures (Fig. 3). Cerebrospinal fluid pressure significantly increased with cross-clamping in the controls and in the group with aortofemoral shunts alone. This increase persisted throughout the cross-clamp period and returned to baseline after reperfusion. No statistical difference occurred in CSF pressure between the controls and the dogs with aortofemoral shunts alone (Fig. 3). The groups of dogs with CSF drainage were kept at a pressure of 0 mm Hg by continuous drainage. ' The calculated spinal cord perfusion pressures were not significantly different between groups at baseline or during reperfusion. During aortic occlusion, the spinal cord perfusion pressure significantly increased in each treatment group starting with the controls (the lowest) and increasing in those with CSF drainage; further increases were seen in those with aortofemoral shunting; and the greatest spinal cord perfusion pressure was seen in those with CSF drainage and aortofemoral shunting (Fig. 3). Histology Histologic examination revealed anterior horn cell degeneration with degeneration of surrounding neural tissue in the gray matter of paraplegic animals (Fig. 4). These changes were seen in five of the seven paraplegic dogs. DISCUSSION Paraplegia was first observed in 1910 by Carrel is after thoracic aortic occlusion in dogs. Since that rime numerous methods have been used both experimentally and clinically in an effort to reduce the incidence of paraplegia after reconstruction of the thoracoabdominal aorta, but none have been consistently successful. Current techniques used for spinal cord protection primarily involve CSF drainage, distal aortic perfusion, reimplantation of intercostal arteries, expeditious operation, and control of hemodynamics. Distal perfusion can be achieved with either partial bypass or with shunts from the left ventricle or proximal thoracic aorta to the distal aorta or femoral vessels. 3'16 Proponents of bypass or shunts argue that not only is distal perfusion improved, but that proximal aortic hypertension and cardiac workload are decreased.

8 796 Elmore et al. lournal of VASCULAR SURGERY Fig. 4. Top, Light micrograph of normal lumbar spinal cord gray matter from a dog without neurologic injury after aortic occlusion. Bottom, Light micrograph of lumbar spinal cord gray matter of paraplegic control animal. Histologically there is anterior horn cell degeneration with ischemia of surrounding neural tissue. (Hematoxylin-eosin stain; original magnification x 400.) Blaisdell and Cooley 17 in 1962 showed that CSF drainage significantly reduced paraplegia in dogs undergoing aortic cross-clamping. Subsequent experimental work has correlated this improvement in neurologic outcome after CSF drainage with increased SCBF. 12 A recent randomized prospective study by Crawford et al. 8 failed to document efficacy of CSF drainage at reducing ischemic cord injury in patients undergoing repair of extensive thoracoabdominal aneurysms. Cerebrospinal fluid drainage may, however, provide some degree of protection by increasing the time during the cross-clamp period before irreversible ischemia occurs38 In a similar fashion, the use of distal aortic perfusion in an attempt to decrease ischemic cord injury is controversial. Both experimental studies by Laschinger et al39 and clinical studies by Verdant et al.20 and Donahoo et al.~6 demonstrate a beneficial effect. Crawford et al, 3 in a prospective study, however, found no benefit with the use of distal aortic perfusion. Once again, it may be possible that distal aortic perfusion may provide some degree of protection and that in combination with CSF drainage there could be further improvement in neurologic outcome. In this study it has been shown that aortic cross-clamping can cause significant elevations in CSF pressure resulting in critical reductions in spinal cord perfusion. Drainage of CSF before crossclamping of the thoracic aorta improved neurologic outcome, improved spinal cord perfusion pressure, and significantly improved perfusion to the gray and white matter of the lumbar spinal cord, the areas most susceptible to injury. These findings are consistent with previous animal models in which CSF drainage was used. 12'17'21 Aortofemoral shunting in this study was also found to result in significantly improved neurologic outcome as compared with controls. Dogs undergoing aortofemoral shunting had the greatest increase in flows as compared with control animals or animals having CSF drainage alone, but the addition of CSF drainage to aortofemoral shunting did not further increase SCBF. Likewise, the addition of CSF drainage to aortofemoral shunting did not significantly improve neurologic outcome, although this was the only group with complete protection from neurologic injury. Further refinements in experimental methods and monitoring may identify subtle differences between groups with combined methods of spinal cord protection. An additional advantage of aortofemoral shunting in this animal model is that it prevented a severe acidosis that was seen after reperfusion in the control and CSF drainage groups. In clinical practice, control of acidosis may help maintain cardiovascular stability and influence reperfusion syndromes. In a similar fashion, aortofemoral shunting was associated with significantly less hyperglycemia. Because hyperglycemia may influence neurologic outcome, 22 control of hyperglycemia with aortofemoral shunts may be beneficial. Further development of adjuncts to prevent ncurologic injury have been hampered, in part, by the lack of a reliable method to monitor the adequacy of spinal cord perfusion during the operation. Our study has shown a relatively high overall accuracy of SEPs in predicting ischemic cord injury in the animals with an aortofemoral shunt. This is in agreement with data of Laschinger ct al.,19,2s who used SEPs in similar animal models of spinal cord ischemia. However, the clinical experience of Crawford et al.a demonstrated a high incidence of false-

9 Volume 15 Number 5 May 1992 Spinal cord ischemia 797 positive results with SEPs. The main limitations of SEP monitoring are the following: (1) impulses are not traveling in the motor pathways and may, therefore, not reflect ischemic injury to the motor system; (2) peripheral nerve ischemia may block the transmission of impulses into the central nervous system and give the false impression of spinal cord injury. Peripheral nerve ischemia likely accounts for the observation that seven of eight animals in the group having CSF drainage lost SEPs compared with only one of eight animals in the distal aortic perfusion group. This occurred despite the fact that no significant difference occurred in neurologic outcome between these groups of animals. Motor-evoked potentials have been proposed as an alternative method to directly monitor the motor tracks of the spinal cord and avoid the problem of peripheral nerve ischemia. 9,24,25 These electrical impulses are generated at the level of the cerebral cortex and travel down to the lower motor neurons of the spinal gray matter. Motor-evoked potentials recorded from the spinal cord would not be hampered by peripheral nerve ischemia and in theory would be more representative of ischemic injury to the motor tracts. In this study and in our prior experimental work, 11 changes in amplitude and latency or loss of MEPs did not significantly correlate with neurologic outcome. Motor-evoked potentials recorded over the spinal cord after cerebral electrical stimulation were lost in only 16% of the animals with neurologic injury, giving the MEP a very low sensitivity. The low sensitivity of MEPs is probably related to the method used to stimulate and record these potentials. We hypothesize, from this and our prior study, n that the MEPs recorded over the low thoracic and proximal lumbar spine are potentials that are transmitted in the axons of the descending corticospinal tracts. The cell bodies of these neurons lie within the cerebral cortex making their fibers in the spinal white matter tracts more resistant to ischemia than anterior horn cells and interneurons whose cell bodies lie within the ischemic cord. Our data suggest that to improve the sensitivity of MEPs, recordings must be made distal to the anterior horn cell. This is supported by Svensson et al.10 who recently reported that MEPs had a high sensitivity at predicting paraplegia when stimulations were elicited with an intrathecal electrode and recordings were made from lower extremity muscles. The mechanisms producing paraplegia are complex, and must certainly involve alterations in regional SCBF, CSF dynamics, and reperfusion. Al- though neither the minimal level of flow nor the duration of ischemia that can be tolerated by the spinal cord is known, adjuncts such as CSF drainage and distal aortic perfusion may allow for a longer duration of cross-clamping before irreversible ischemia occurs. Since paraplegia is multifactorial, and in patients none of the adjuncts alone provide complete spinal cord protection, a combination of multiple adjuncts may be necessary to reduce the incidence of neurologic complications after thoracoabdominal aortic reconstructions. Although in our model a clear benefit from combined methods of protection could not be demonstrated, further refinements in monitoring techniques may help evaluate different methods of spinal cord protection in the future. The authors thank Haruo Okazaki, MD, for reviewing the histology, Peter C. O'Brien, PhD, for advice in the statistical calculations, O. Arian Hildestad for technical expertise, and Marcia A. Simonson and Sharon K. Miller for editorial assistance. REFERENCES 1. Crawford ES, Rubio PA. Reappraisal of adjuncts to avoid ischemia in the treatment of aneurysms of the descending thoracic aorta. ~ Thorac Cardiovasc Surg 1973;66: Crawford ES, Walker HSJ III, Saleh SA, Normann NA. Graft replacement of aneurysm in descending thoracic aorta: results without bypass or shunting. Surgery 1981;89: Crawford ES, Mizrahi EM, Hess KR, Coselli JS, Sail HS, Patel VM. The impact of distal aortic perfusion and somatosensory evoked potential monitoring on prevention of paraplegia after aortic aneurysm operation. J Thorac Cardiovasc Surg 1988;95: Livesay JJ, Cooley DA, Ventemiglia RA, et al. Surgical experience in descending thoracic aneurysmectomy with and without adjuncts to avoid ischemia. Ann Thorac Surg 1985;39: Jex RK, Schaff HV, Piehler JM, et al. Early and late results following repair of dissections of the descending thoracic aorta. J VAsc SURG 1986;3: Homer LH, Symmonds JB, Pairolero PC, Cherry KJ, HaUett JW, Gloviczki P. Thoracoabdominal aortic aneurysm repair. Arch Surg 1988;123: Laschinger JC, Izumoto H, Kouchoukos NT. Evolving concepts in prevention of spinal cord injury during operations on the descending thoracic and thoracoabdominal aorta. Ann Thorac Surg 1987;44: Crawford ES, Svensson LG, Hess KR, et al. A prospective randomized study of cerebrospinal fluid drainage to prevent paraplegia after high-risk surgery on the thoracoabdominal aorta. J VAsc SURG 1991i13: Laschinger JC, 0wen J, Rosenbloom M, Cox JL, Kouchoukos NT. Direct noninvasive monitoring of spinal cord motor fimction during thoracic aortic occlusion: use of motorevoked potentials, l VAsc SUttG 1988;7: Svensson LG, Patel V, Robinson MF, Ueda T, Roehn JOF, Crawford ES. Influence of preservation or perfusion of intraoperatively identified spinal cord blood supply on spinal

10 798 Elmor et al. Journal of VASCULAR SURGERY motor evoked potentials and paraplegia after aortic surgery. J VASC SURG 1991;i3: Elmore JR, Gloviczki P, Harper CM, et al. Failure of motor-evoked potentials to predict neurologic outcome in experimental thoracic aortic occlusion. J VAsc SURG 1991; 14: Bower TC, Murray MJ, Gloviczki P, Yaksh TL, HoNer LH, Pairolero PC. Effects of thoracic aortic occlusion and cerebrospinal fluid drainage on regional spinal cord blood flow in dogs: correlation with neurologic outcome. J VAse SUV, G i989;9: i3. Heymann MA, Payne BD, Hoffman JiE, Rudolph AM. Blood flow measurements with radionuclide-labeled particles. Prog Cardiovasc Dis 1977;20: Tarlov IM. Spinal cord compression. Mechanisms of paralysis and treatment. Springfield, Illinois: Charles C Thomas, Carrel A. On the experimental surgery of the thoracic aorta and heart. Ann Surg. 16. Donahoo JS, Brawley ILK, Gort VL. The heparin coated vascular shunt fo~" thoracic aortic and great vessel procedures: a ten-year experience. Ann Thorac Surg I977;23: Blaisdell FW, Cooley DA. The mechanism of paraplegia after temporary thoracic aortic occlusion in its relationship to spinal fluid pressure. Surgery 196I;5I: Svensson LG, Grum DF, Bednarski M, Cosgrove DM, Loop FD. Appraisal of cerebrospinal fluid alterations during aortic surgery with intrathecal papaverine administration and cerebrospinal fluid drainage. ]" VAse SURG 1990;11: Laschinger JC, Cunningham JN, Baumann FG, Isom OW, Spencer FL. Monitoring of somatosensory evoked potentials during surgical procedures of the thoracoabdominal aorta. II. Use of somatosensory evoked potentials to assess adequacy of distal aortic bypass and perfusion after thoracic aortic crossclamping. J Thorac Cardiovasc Surg 1987;94: Verdant A, Page A, Cossette R, Dontigny L, Page P, BaiUot R. Surgery of the descending thoracic aorta: spinal cord protection with the Gott shunt. Ann Thorac Surg 1988;46: McCullough JL, Hollier LH, Nugent M. Paraplegia after thoracic aortic occlusion: influence of cerebrospinal fluid drainage, l VASC SURG 1988;7: Lanier WL, Stangland KJ, Scheithauer BW, Milde JH, Michenfelder JD. The effects of dextrose infusion and head position on neurologic outcome after complete cerebral ischemia in primates: examination of a mode. Anesthesiology 1987;66: Laschinger JC, Cunningharn JN, Cooper MM, Baumann FG, Spencer FC. Monitoring of somatosensory evoked potentials during surgical procedures on the thoracoabdominal aorta. I. Relationship of aortic cross-clamp duration, changes iv somatosensory evoked potentials, and incidence of neurologic dysfunction. J Thorac Cardiovasc Surg 1987;94: Merton PA, MortonMB. Stimulationofthe cerebral cortex in the intact human subject. Nature 1980;205: Konrad PE, Tacker WA, Levy WJ, Reedy DF, Cook JR, Geddes ME. Motor evoked potentials in the dog: effects of global ischemia on spinal cord and peripheral nerve signals. Neurosurgery 1987;20: Submitted Oct. 15, 1991; accepted Jan. 22, DISCUSSION Dr. Richard Fowl (Cincinnati, Ohio). This study is a superb example of very precise, detailed, scientific methodology. Its main purpose is to use multiple treatments aimed at the various etiologic factors believed to be responsible for injury to the spinal cord after ischemia. Such factors include decreased spinal cord perfusion, increased CSF pressure, and the more controversial role of reperfusion injury. The only aspect of the methodology with which I would take issue was that you allowed the animals to survive for only 24 hours. It may have been useful to allow all animals who were initially neurologically normal to survive for up to 7 days to determine if delayed neurologic deficits would occur. As you know, in humans and other animal models of spinal cord ischemia, neurologic deficits may occur from 12 hours to 21 days after operation. My first question is the following: Have you ever observed delayed paraplegia in your previous work with this canine model? I found it interesting that the glucose levels were significantly elevated in those animals who did not get shunts. Since elevated glucose levels have been associated with a higher incidence of ischemic spinal cord injury in other animal models, I wonder if hyperglycemia played an even more important role than elevated CSF pressures in causing injury. I ask this question because seven of eight animals subjected to shunting without CSF drainage had nearly normal glucose levels and were neurologically normal despite having elevated CSF pressures. Although you have achieved great success in preventing neurologic injury in this study and in your previous work with the canine model, unfbrmnately the fact remains that what is therapeutic in animals is not necessarily effective in humans. Both CSF drainage and shunting have not been shown to reduce the incidence of ischemic neurologic injury during repair of thoracic or thoracoabdominal aneurysms, although there has not been a well-conducted clinical study of both of these modalities used concomitantly. However, your data did not show any significant benefit of using combined shunting and CSF drainage compared with either modality alone. To date, the only treatment that has been shown to have any clinical effectiveness in reducing the extent of injury in the human spinal cord is massive doses of methylprednisolone sodium succinate. In one recently published study, this agent was used in victims of acute

11 Volume 15 Number 5 May 1992 Spinal cord ischemia 799 spinal cord trauma who were given a loading dose of 2 to 3 gm followed by a continuous infusion of 300 to 500 mg per hour for 23 hours. Since many investigators believe that ischemia is an important pathologic process in traumatic spinal cord injury, I would be interested in your thoughts regarding use of this agent in a similar protocol during repair of thoracoabdominal aneurysms. Finally, since your laboratory has made many important contributions in this field, I think we all would'like to know what you currently do at the Mayo Clinic to protect the spinal cord during reconstructive procedures of the thoracoabdominal aorta. Dr. James Elmore; We have not seen any delayed paraplegia in our experiments, given the constraints of this 24-hour follow-up. We are well aware in humans as well as some animal models, such as the rabbit, that delayed paraplegia is a significant problem.., Glucose monitoring did identify hyperglycemia in the control group and in those dogs undergoing CSF drainage. However, the animals undergoing CSF drainage had a significant improvement in neurologic outcome as compared with the control animals, and therefore, hyperglycemia in this model is not as important as is the CSF pressure. Studies in stroke victims and experimental models of spinal cord injury have demonstrated a correlation between neurologic injury and hyperglycemia, and therefore it seems reasonable to prevent hyperglycemia. Your comments regarding the clinical studies of Crawford that have failed to support the use of shunting alone or CSF drainage alone prompted this animal study. I can just reiterate your comments that we did not find a beneficial effect from the combination therapy in this animal model. Steroids have been found to be useful in the traumatic setting of spinal cord injury as reported last year in the New England Journal of Medicine. It required a large series of patients to find statistical significance with the use of steroids in that study. Steroids appear to be a reasonable treatment in these patients, and we have started using this approach in our patients. Other adjuncts that we have routinely used include expedient operation and hemodynamic stability throughout the operative and postoperative period to try to avoid delayed paraplegia. We continue to reimplant intercostals and perform CSF drainage in all of our patients. With extensive thoraeoabdominal reconstructions we tend not to use distal aortic perfusion, because it is our feeling that perfusing just the lower extremities is not helpful. In thoracoabdominal reconstructions, only approximately 12% of the patients have undergone distal aortic perfusion. I think the future will undoubtedly bring additional studies on multiple adjuncts including multiple drugs, topical cooling, and ways to better visualize or localize critical intercostals that are necessary for reimplantation. Dr. Mark Adams (Milwaukee, Wis.). I am curious to know if you have considered using any of a variety of free radical scavengers. Quite a bit of material has been published over the last few years on reperfusion injury and the role of free radical scavengers in moderating it. I did actually hear an interesting presentation on a recent trip to British Columbia of someone there who had been doing this and showed quite a significant effect in a pig model like yours. Dr. Elmore. More experimental data must be obtained on the use of free radical scavengers and spinal cord ischemic injury. The results from experimental models are very contradictory. For instance, superoxide dismutase (SOD) has been one of the prime free radical scavengers that has been used in evaluating cardiac reperfusion injury. Problems with SOD include determining the ideal dose and analyzing the entry of SOD into the spinal cord. In patients we do use mannitol, and I suspect that mannitol, besides having the osmotic effect on renal function, is having an effect as a free radical scavenger. I believe that free radical scavengers merit further investigation. Dr. Charles Acher (Madison, Wis.). As you know, we have had an interest in preventing paraplegia at the University of Wisconsin for some years as any surgeon would who works on these aneurysms. There are a couple of things that are clear. Every experimental study that has ever been done with spinal fluid drainage has shown a benefit. The experimental evidence is strong, starting with Cooley and discovered by Lars Svenson who reported in 1985 or 1986 studies with baboons showing marked benefit with spinal fluid drainage and no benefit with free radical scavengers or laminectomy. The randomized study by Stanley Crawford in our opinion failed to really test CSF drainage because he limited the amount of spinal fluid removed to only 35 ml on average, and he clearly demonstrated in their pressurc data that this did not control spinal fluid pressures. The amount of spinal fluid that we had to remove in patients was 85 to 150 ml to keep spinal fluid pressures below 10 mm Hg. Could you comment on why the experimental evidence is so strong and whether you think any of the studies that have been done in humans really address the issue of spinal fluid drainage or not. Dr. Elmore. In our clinical practice we drain approximately 50 ml on average of CSF fluid in these patients and that is to keep the pressure less than 10 mm Fig. In our hands, we are still seeing paraplegia in these patients despite the CSF drainage, so it is not the sole answer. In the dog model, CSF drainage totally prevents paraplegia, and therefore in this model, CSF pressure plays a primary role. These animal models are invaluable and suggest treatment protocols that need to be evaluated in humans. These animal models will be especially important in testing different pharmaceutical agents in the future.