Operation on the descending thoracic and thoracoabdominai

Transcription

1 Pathogenesis of Spinal Cord Injury During Simulated Aneurysm Repair in a Chronic Animal Model Otto E. Dapunt, MD, Peter S. Midulla, MD, Ali M. Sadeghi, MD, PhD, Craig K. Mezrow, MS, David Wolfe, MD, Alejandro Gandsas, MD, Rosario A. Zappulla, MD, PhD, Carol A. Bodian, PhD, M. Arisan Ergin, MD, PhD, and Randall B. Griepp, MD Departments of Cardiothoracic Surgery and Biomathematics, The Mount Sinai Medical Center, New York, New York The pathogenesis of paraplegia after repair of thoracic aortic aneurysms is controversial. Using direct spinal cord evoked potential monitoring, critical intercostal arteries (CICA) were identified to evaluate the impact of backbleeding and ligation versus that of preservation during simulated aneurysm repair. Thirty pigs (40 kg) were randomly assigned to one of five groups. In groups 1 through 4, a thoracic segment containing CICA was cross-clamped for 60 minutes and distal aortic perfusion was provided by a centrifugal pump. In groups 1 and 2, the thoracic segment was vented, maintaining segment pressure at 0 mm Hg; CICA were ligated in group 1 and preserved in group 2. Thoracic segment was perfused at 70 mm Hg in groups 3 and 4; CICA were ligated in group 3 and preserved in group 4. Critical intercostal artery ligations were performed at the end of the cross-clamp period. In group 5 simple cross-clamping at the left subclavian artery was performed as a control. The combination of venting and ligation of CICA correlated with impaired neurologic outcome according to Tarlov's score (median, 1.5 in group 1 versus 3 in group 2; p = 0.015), indicated by a significant difference in median values of direct spinal cord evoked potential amplitude (expressed as a fraction of baseline values) at 120 minutes after cross-clamping (0.76 in group 1 versus 0.98 in group 2; p = ). Ligation of CICA without prior venting did not result in a significantly reduced Tarlov score (median, 3.5 in group 3 versus 4 in group 4; p = 0.182) and the difference in direct spinal cord evoked potential amplitude was less pronounced (0.85 in group 3 versus 1.0 in group 4, p = ). Simple cross-clamping caused paraplegia in all animals (group 5 versus group 1, p = 0.002). These results have prompted modifications of our operative strategy to include reattachment of CICA identified by monitoring of somatosensory evoked potentials and prevention of a steal phenomenon by serial intercostal ligation before aneurysm resection. (Ann Thorae Surg 1994;58:689-97) Operation on the descending thoracic and thoracoabdominai aorta that requires proximal cross-clamping results in spinal cord ischemia with a variable risk of paraplegia, dependent on the length of aorta replaced and the duration of cross-clamp time. The incidence of postoperative paraplegia is low in coarctation repair in children (1.5%) and after operation for aneurysms of the proximal descending thoracic aorta (3% to 5%). It is considerably higher after repair of traumatic aortic ruptures (25%) and can be as high as 38% after extensive operation for dissections of the thoracoabdominal aorta [1-4]. Several factors are reported to contribute to spinal cord ischemia, including simple aortic cross-clamping for periods longer than 30 minutes, reduction of spinal cord perfusion pressure by perioperative hypotension and severe hypertension proximal to the aortic cross-clamp, and elevation of cerebrospinal fluid pressure due to hemodynamic changes or administration of sodium nitroprusside. Presented at the Thirtieth Annual Meeting of The Society of Thoracic Surgeons, New Orleans, LA, Jan 31-Feb 2, Address reprint requests to Dr Dapunt, Division of Cardiac Surgery, Department of Surgery I, University of Innsbruck, Anichstrasse 35, A-6020 Innsbruck, Austria by The Society of Thoracic Surgeons The incidence of paraplegia is also determined by the location, extent, cause, and possible rupture of aortic aneurysms. Backbleeding of intercostal arteries during open aneurysm resection may substantially decrease spinal cord perfusion pressure [5] by creating a steal phenomenon. The failure to reanastomose critical intercostal arteries (CICA) is assumed to be responsible for immediate and delayed neurologic deficits [6]. Experimental and clinical studies have focused on the development of methods to prevent spinal cord ischemia, resulting in substantial improvements in surgical technique and perioperative management. Shunts or bypass pumps to control proximal hypertension and distal aortic pressure have emerged as strategies that are employed frequently during operations on the descending thoracic and thoracoabdominal aorta. Systemic or local hypothermia (or both), control of cerebrospinal fluid pressure by fluid drainage, and various pharmacologic maneuvers have been shown to protect the spinal cord or to ameliorate the effects of ischemia on neurologic outcome [7-9]. Due to the multifactorial etiology of postoperative paraplegia, it is unlikely that any single intervention will eliminate this complication. The failure to identify exactly /94/$7.00

2 690 DAPUNT ET AL Ann Thorae Surg and to reanastomose important intercostal arteries that supply the spinal cord is believed to be one of the major causes of postoperative neuromuscular deficits [61, despite optimal protection of the spinal cord during aortic crossclamping. We have conducted an experimental study to evaluate the effects of backbleeding and interruption of important intercostal arteries on neurologic outcome. We developed a method to identify critical intercostal arteries (CICA) employing direct spinal cord evoked potential (DSEP) monitoring. Time 2 f Latency (rns) _I Amplitude (/lv) Material and Methods Pilot Study Anesthesia was induced in 5 female Yorkshire pigs (40 kg) by intramuscular injection of ketamine 00 mg/kg), endotracheal intubation was performed after intravenous administration of pentobarbital mg/kg), and anesthesia was maintained with isoflurane 0 %-3%). Muscle relaxation was achieved with pancuronium bromide (0.1 mg/kg). The animals were ventilated throughout the operative procedure with an inspired oxygen fraction of 0.45 to 0.75 and positive end-expiratory pressure of 5 cm H 20 according to oxygen tension more than 100 mm Hg and carbon dioxide tension of 30 to 40 mm Hg. Electrocardiographic measurements, esophageal and rectal temperatures, and distal arterial blood pressure through the left femoral artery were recorded continuously. The left femoral vein was cannulated for central venous access and a Swan-Ganz catheter was inserted into the right pulmonary artery through the right femoral vein. The posterior laminae and spinous processes of two vertebrae at the cervicothoracal junction and at L4, L5 were removed, exposing the spinal dura mater in these areas. Monitoring of DSEP was performed by stimulating the spinal cord at the cervical laminectomy and recording at the level of L5, Sl using standard electroencephalogram needle electrodes 0 em, length), which were placed into the epidural space. The Cadwell Spectrum 32 neurophysiological monitoring station (Caldwell Laboratories, Kennewick, WA) was used to monitor spinal cord function. The spinal cord was stimulated with 15 to 50 rna current. Stimulus duration was 50 microseconds, and repetition rate was 5.13/s. Electrical activity was filtered between 10 and 300 Hz with a 60 Hz notch filter at maximum gain setting and scaling with a magnification factor of less than 5. Evoked potential signals were analyzed after 100 averages and stored on optical discs. Data measured included latencies of the positive (Tl) and negative (T2) peaks of the biphasic signal, and peak to peak amplitude (Fig 1). Muscle relaxation was maintained throughout the experiment to avoid misinterpretation of DSEP signals due to muscle artifacts. A left posterolateral thoracotomy through the fifth intercostal space was made and extended to a left paramedian incision. Dividing the diaphragm down to the aortic hiatus permitted excellent exposure of the entire descending aorta. The aortic arch was cannulated for proximal aortic pressure monitoring. The animal was given heparin, 100 U/kg, and the right femoral artery was cannulated using a 14F or 16F Bard cannula. Perfusion of the caudal f Time 1 Fig 1. Biphasic direct spinal cord evoked potential signal with occurance of first peak (Tl) at 6 to 8 ms and second peak (T2) at 11 to 14 ms. Amplitude in microvolts (/-LV) was measured from peak to peak. body was accomplished by instituting a left heart bypass with a centrifugal pump (Biornedicus, Medtronic, Inc, Minneapolis, MN) after cannulation of the left atrium through the left atrial appendage with a 22 or 24F Pacifico cannula. The descending aorta was divided into two thoracic and two abdominal segments (numbered I to IV) of approximately equal size by applying five crossclamps as shown in Figure 2. Before cross-clamping, each aortic segment was cannulated with 14F Bard cannulas, which were connected to the inflow as well as to the outflow line of the centrifugal pump circuit. With this arrangement, each of the segments could be vented or perfused individually while distal body perfusion was maintained at a mean pressure of 60 to 70 mm Hg (flow 1.5 to 2.0 Lz'min). Spinal cord ischemia was induced by simultaneous venting of all four aortic segments until loss of DSEPs was recorded (average, 8 minutes; range, 3 to 14 minutes). Recurrence of the evoked potentials signal after reperfusing one of the four segments in random sequence while keeping the others vented permitted identification of the aortic segment containing CICA. Experimental Study Thirty female Yorkshire pigs weighing an average of 37.3 kg (range, 21 to 62 kg) were anesthetized as described above and monitoring of DSEP was carried out in the same manner as in the pilot study group. Hemodynamic parameters (heart rate, aortic arch pressure, left femoral artery pressure, pulmonary artery pressure) and both esophageal and rectal temperatures were monitored continuously. A left posterolateral thoracotomy through the seventh intercostal space was performed and the aortic arch was cannulated for proximal blood pressure monitoring. The caudal segment of the descending thoracic aorta (numbered 11), which had been identified to carry important intercostal arteries in all of the pilot experiments, was cannulated using a standard three-lumen aortic root venting-infusion cannula, which enabled continuous pressure monitoring of this segment.

3 Ann Thorae Surg DAPUNT ET AL 691 The animals were randomly assigned to one of five groups consisting of 6 pigs each. In all animals in groups 1 through 4, segment II was double cross-clamped for 1 hour while distal body perfusion with a left atrium to right femoral artery centrifugal pump bypass (left femoral artery pressure 60 to 70 mm Hg, average flow 1900 ml/min) was provided (Fig 3). In groups 1 and 2, segment II was vented during the I-hour cross-clamp period, maintaining segment pressure at 0 mm Hg; all intercostal arteries originating in this aortic segment (usually five pairs) were ligated in group 1 and preserved in group 2. Segment II was perfused at 70 mm Hg in groups 3 and 4; intercostal arteries were ligated in group 3 and preserved in group 4. Ligation of intercostal arteries was performed at the end of the cross-clamp period. In group 5, as a control, simple aortic cross-clamping at the left subclavian artery without distal perfusion was carried out. Measurements of latency (Tl, T2) and peak to peak amplitude of DSEP signals were taken at baseline (prior to establishment of left heart bypass), at 5-minute intervals during cross-clamping until 30 minutes after crossclamping, and at 60 and 120 minutes after cross-clamping. Hemodynamic parameters (as indicated previously) and \ \ Blood pressure \ \ \ I I / I / ( I \\ \ '-'-<,--- Fig 2. Identification of critical intercostal arteries originating in segment II by venting of the entire aorta and subsequent reperfusion of segments in random sequence. Fig 3. Venting or perfusion of caudal thoracic aortic segment (11) with ligation or preservation of critical intercostal arteries. cardiac output, hematocrit, blood gas analysis, and esophageal and rectal temperatures were measured at 30 minute intervals from baseline until 120 minutes after crossclamping. Care was taken to keep hemodynamic conditions and metabolic status stable by administering crystalloid solutions, human albumin, donor blood, and sodium bicarbonate as necessary. Catecholamines or vasodilators (except isoflurane) were strictly avoided. A heating/ cooling blanket and warming of the Biomedicus pump head with a heating lamp were used to keep the animal's body temperature at baseline values. Comparison of parameters of hemodynamics, metabolic status, and body temperatures revealed no statistically significant differences within or between groups that could have been responsible for changes in evoked potential signals. Average hematocrit at baseline was 28.3%. Due to fluid administration during and after the cross-clamp period, mean hematocrit dropped to 25.2% at 60 minutes after crossclamping and to 25.7 at 120 minutes after cross-clamping. No significant differences in hematocrit values were seen among the five groups. After the last measurement, cannulas were removed and thoracotomy, laminectomy, and groin incisions were closed. Anesthesia was discontinued and the animals were

4 692 DAPUNT ET AL Ann Thorac Surg 1994;58: Table 1. Number of Animals With Loss of Positive Peak (TV of Direct Spinal Cord Evoked Potential Signal at Different Time Intervals 5 min 10 min 30 min 60 min 5 min 30 min 120 min Group XC XC XC XC AXe Axe Axe AXC = weaned from the ventilator and placed in an incubator with oxygen supply. Pigs that were unable to change body position by themselves were turned at 6-hour intervals. Acetaminophen (300 mg/day orally) was given for control of postoperative pain. Neurologic evaluation according to Tarlov's score was performed daily by a veterinarian who was blinded to the study design (Tarlov 0 = spastic paraplegia; 1 = paraparesis, slight movement of hindlimbs; 2 = paraparesis, strong movement of hindlimbs; 3 = paraparesis, animal able to stand up and walk with unsteady gait; 4 = full recovery). To ensure complete recovery from anesthesia and extensive operation, Tarlov scores from postoperative day 3 were taken for statistical analysis. HISTOLOGIC EXAMINATION. Sixteen pigs that survived until postoperative day 7 were anesthetized as described previously, a median sternotomy was performed, and the ascending aorta was cannulated. The animals then were sacrificed by in vivo, in situ fixation with paraformaldehyde to optimize pathologic assessment of the spinal cord. Two liters of paraformaldehyde was infused into the aortic root using a roller pump at a pressure of 30 mm Hg. Two hours after completion of perfusion the thoracic and lumbosacral spinal cord was excised for histologic examination. The pathologist performing histologic examinations was blinded to the study design. All experimental animals received humane care in compliance with the "Guide for the Care and Use of Laboratory Animals" published by the National Institutes of Health (NIH publication 85-23, revised 1985). STATISTICAL ANALYSIS. Wilcoxon tests were used to evaluate differences in amplitude between the groups at certain time points. The value of different measurement parameters of evoked potentials in predicting neurologic outcome was assessed by Fisher's exact test (two-tailed). Results after cross-clamping; XC ~during cross-clamping. or IV did not cause recovery of the signal, independent of the sequence of segments reperfused. Experimental Study Postoperative mortality (within 7 days after the operation), mostly due to respiratory or renal failure, was highest in group 5 (50%). There were two deaths in group 1 (33%), one in group 2 (17%), and no deaths in groups 3 or 4. Latency of DSEP peaks occurence did not change significantly during the different steps of the experiment in any of the animals. Ischemia was indicated by a progressive loss in amplitude. Reduction in amplitude affecting the second hump of the signal (T2) was seen as an early sign of spinal cord ischemia. Loss of amplitude at T2 was encountered as early as 5 minutes after aortic cross-clamping in 2 animals in group 5, whereas loss of amplitude at T1 always occurred in conjunction with complete loss (flat line) of the signal. Loss of amplitude at T1 without prior loss of amplitude at T2 was not seen in any of the experiments (Tables 1, 2). Baseline values of evoked potentials amplitude differed among the animals, depending on animal size and positioning of the stimulating and recording electrodes. Mean amplitude height was /LV, ranging from 2.12 to /LV. For this reason, changes in amplitude had to be expressed as a percentage fraction of baseline values. Means of percentage change in evoked potentials amplitude of the five groups in comparison are shown in Figure 4. Table 3 depicts mean values, standard deviations, and median values at three different time points. Induction of spinal cord ischemia by either venting of segment II (groups 1 and 2) or simple aortic cross-clamping (group 5) correlated with a progressive diminishing of amplitude, which occurred earliest in group 5. Complete loss of the signal was seen in 3 animals in group 1, 4 in group 2, and 5 in group 5. As expected, signal amplitude remained stable around 100% of baseline value in groups 3 and 4, where segment II was perfused during the cross-clamp period. Ligation of intercostal arteries in group 3 caused a reduction in amplitude which was most significant at 120 minutes after cross-clamping compared with no changes in group 4, where arteries were not interrupted. Complete signal recovery occurred in group 2 after 1 hour of venting with preservation of important arteries. In contrast, recovery of the signal was incomplete in group 1 after venting and ligation of arteries. In group 5 (the simple clamping group), signal amplitude at 120 minutes after cross- Table 2. Number of Animals With Loss of Negative Peak T2 of Direct Spinal Cord Evoked Potential Signal at Different Time Intervals 5 min 10 min 30 min 60 min 5 min 30 min 120 min Group XC XC XC XC AXe Axe Axe Pilot Study 2 a a 4 5 a 0 0 The lower thoracic aortic segment (segment II) was found to contain important intercostal arteries, as shown by 4 a 0 0 a a 0 a consistent rapid recurrence of the evoked potential signal after 5 minutes' reperfusion of segment II while all other segments were still vented. Reperfusion of segments I, III, AXC = after cross-clamping; XC = during cross-clamping.

5 Ann Thorac Surg DAPUNT ET AL 693 % of baseline amplitude Fig 4. Comparison of mean values of change in direct spinal cord evoked potential amplitude (expressed as a fraction of baseline values) of 5 different groups at different intervals during and after cross-clamping. Symbols from left to right correspond with groups 1 through 5. (Preserv. = Preservation.) Time (minutes) Cross Clamping Post Cross Clamping... Venting Ligation... Venting PresslY... Perfusion ligalioltie- Perfusion P r Simple~ Clamping clamping remained less than 50% in all but one of the cases. The negative impact of ligation of intercostal arteries on evoked potential signal amplitude was documented by comparing groups 2 and 4 (arteries preserved) with groups 1 and 3 (arteries ligated) using the Wilcoxon test. Venting of segment II was correlated with a significant reduction of amplitude during cross-clamping. The adverse influence of interrupting intercostal arteries was confirmed by a significant difference between group 1 and group 2, as well as between group 3 and group 4, at 120 minutes after crossclamping (Table 4). Neurologic examination and scaling according to Tarlov's score was possible in all animals in groups 1,3, and 4. Data from 3 group 5 pigs that died 24 hours after operation were included in the study because paraplegia was confirmed at this time and improvement within 2 days was considered very unlikely. One animal in group 2 died in the early postoperative period while still under influence of anesthesia, which made accurate neurologic evaluation impossible. Median values of Tarlov scores at postoperative day 3 are shown in Table 5. The combination of venting and ligation of CICA (group 1) was correlated with impaired neurologic outcome. Group 1 differed significantly from group 2 (p = 0.015, Fisher's exact test, Table 3. Change in Spinal Cord Evoked Potential Amplitude in Groups 1 Through 5 at Different Time Intervals" Time Interval 30 min XC 60 min XC 120 min AXC Group Mean SD Median Mean SD Median Mean SD Median O.B LOS O.D Data are expressed as mean values, standard deviation, and median values of fractions of baseline measurements. Axe = after cross-clamping; SD = standard deviation; XC ~during cross-clamping. two-tailed) if neurologic status was compared at the level of Tarlov score greater than 2 (able to stand or walk). Ligation of CICA without prior venting did not result in a significantly reduced Tarlov score. Defining postoperative success as a Tarlov score greater than 3 (full recovery), groups 3 and 4 did not differ significantly (p = 0.182). Simple aortic cross-clamping caused paraplegia in all animals. However, no cases of paraplegia were seen in groups 1 through 4, where distal body perfusion was provided (group 1 versus group 5: Tarlov > 0, P = 0.002) (Table 6). The predictive value of different measurement parameters of evoked potential monitoring on neurologic outcome was calculated using Fisher's exact test. Data from animals in groups 1 and 3 were excluded from analysis of crossclamp period measurements to control for the effect of artery ligation. Parameters that correlated significantly with postoperative paraplegia included (1) loss of amplitude at Time 2 at 120 minutes after cross-clamping (p = 0.003), (2) amplitude of less than 50% of baseline value for more than 40 minutes during cross-clamping (p = ), and (3) amplitude of less than 50% of baseline value at 120 minutes after cross-clamping (p = 0.001). Measurement characteristics indicative of a significantly higher likelihood to develop at least some kind of neurologic sequela (indicated by a Tarlov score of 3 or less) included (1) loss of amplitude at T1 at any time during cross-clamping (p = ), (2) loss of amplitude at T2 at any time during cross-clamping tp = ), (3) amplitude of less than 90% of baseline value for more than 40 minutes during cross-clamping (p = ), and (4) amplitude of less than 90% of baseline value at 120 minutes after cross-clamping (p = 0.017). Histologic Examination Perfusion fixation of the spinal cord was possible in 2 animals each from groups 1 and 5; in 3 from group 2; in 4 from group 4; and in 5 from group 3. Three or four sections from thoracic and lumbar levels were analyzed. Complete symmetrical (bilateral) infarction of spinal grey (both ventral and dorsal horns) with sparing of white funiculi was diagnosed in the sections from all levels in both animals from group 5. Crescentic subacute infarction involving

6 694 DAPUNT ET AL Ann Thorac Surg Table 4. Impact of Ligation of Critical Intercostal Arteries on Direct Spinal Cord Evoked Potential Amplitude" Comparisons Segment Segment Groups Vented Perfused Time of 2 and 4 1 and 3 3 and 4 1 and 2 CICA CICA CICA CICA Measurement (CICA (CICA p (Seg II (Seg II p Preserved Ligated p Preserved Ligated v (min) preserved) ligated) Value perfused) vented) Value (group 2) (group 1) Value (group 4) (group 3) Value 30 min XC min XC min AXC ,1 Data are expressed as median values in microvolts. AXC = after cross-clamping; CICA = critical intercostal arteries; Seg = segment; XC = during cross-clamping. intermediate grey and adjacent midportion of lateral funiculus was seen in a high lumbar section in 1 animal from group 1. The other specimen from this group showed minute equivocal infarction in one mid-lumbar section, involving superficial lateral white funiculus just below entering dorsal root fibers. Two ipsilateral subacute infarcts in the mid-lumbar section were found in 1 pig from group 3, whereas there were no pathologic findings in all other animals. Comment Prolonged aortic cross-clamping without distal perfusion is the single most important factor determining the probability of paraplegia. Ischemic periods with loss of somatosensory evoked potentials for more than 30 minutes have been correlated with a 71% incidence of postoperative paraplegia [10]. After more than 30 minutes of aortic crossclamping, the risk of paraplegia rises significantly [11]. These findings are confirmed by the 100% incidence of paraplegia after 1 hour of simple aortic cross-clamping (group 5) in our study. Another important factor that can playa decisive role in the development of spinal cord ischemia is the failure to identify exactly and reanastomose those intercostal or lumbar arteries that supply the spinal cord [6, 12, 13]. In humans, blood supply to the spinal cord is provided mainly by the anterior spinal artery, which is a branch of the circle of Willis. The cervical and upper thoracic parts of the spinal cord receive additional blood from branches of vertebral and subclavian arteries, and therefore are wellprotected against ischemic injury. The lower thoracic and lumbar portions of the spinal cord are dependent on a Table 5. Comparison of Tarlov Scores at Postoperative Day 3 varying number of radicular arteries originating from intercostal and lumbar arteries. The largest and most developed radicular artery, the arteria radicularis magna, or so-called artery of Adamkiewicz, is highly variable [14]. In humans the artery arises at the level of T5 to T8 in 12% to 15% of all cases, T9 to TI2 in 60%, L1 in 14%, L2 in 10%, L3 in 1.4%, and L4 to L5 in 0.2% [15]. In the pig the artery originates at the level L1 to Sl, usually at L4 [16]. The spinal cord blood supply in pigs is plurisegmental, and the Adamkiewicz artery plays a relatively less important role in blood supply of the spinal cord. In view of this plurisegmental blood supply, aortic segments rather than individual intercostal arteries were examined in this study. One of the most commonly used adjuncts to ameliorate the effects of spinal cord ischemia is the reimplantation of islands of aorta containing the origins of critical intercostal or lumbar arteries. Due to the lack of available methods to define exactly the origin of the Adamkiewicz or other important radicular arteries, this is mostly a blind procedure. This could be one of the reasons why Crawford and associates [4] could not demonstrate any beneficial effect of intercostal artery reimplantation on postoperative neurologic outcome in their study. Svensson and associates [6] recently developed a method for identifying CICA using an intrathecal electrode to detect hydrogen in solution injected into intercostal arteries. The results of this experimental study in pigs have not been widely confirmed in the clinical setting. An open aorta with freely bleeding intercostal or lumbar arteries can cause a reduction in spinal cord perfusion pressure by stealing blood from the high-resistance collaterais of the anterior spinal artery. Adjuncts for rapid occlusion of intercostal artery ostia during open aneurysm resection have been described [17]. There has been a debate about the impact of backbleeding of intercostal arteries on No. with % Group No. in Tarlov with Median Table 6. Neurologic Outcome on Postoperative Day 3 Group Group score 4 Tarlov 4 Score Tarlov % % % % % Score Group 1 Group 2 Group 3 Group 4 Group : : : :

7 Ann Thorac Surg DAI'UNT ET AL 695 neurologic outcome. As reported in a recent clinical study [18], spinal cord injury occurred less frequently in patients with descending aortic aneurysms who underwent operation employing the principle of open aneurysm resection with delayed oversewing of intercostal arteries and open distal anastomosis. Monitoring of somatosensory evoked potentials is a frequently employed adjunct in descending aortic aneurysm operation [19]. The method enables early determination of spinal cord ischemia and permits the surgeon to react in a timely manner [20]. Disadvantages of somatosensory evoked potentials include: exclusive monitoring of sensory pathways (paraplegia, however, is a result of damage to the anterior motor column); the need to employ distal aortic perfusion techniques to avoid misinterpretation due to peripheral nerve ischemia; and somatosensory evoked potentials' sensitivity to factors other than ischemia (such as anesthesia, temperature, and drugs). Motor evoked potentials have been described to overcome some of these drawbacks [21]. Motor evoked potential monitoring ensures assessment of anterior and lateral motor column function and is not affected by peripheral nerve ischemia. However, motor evoked potentials have been shown to have a low sensitivity and low accuracy in predicting neurologic injury [22]. Another disadvantage is the possibility of triggering seizures by stimulating the motor cortex. On the contrary, other experimental work [6] suggests that motor evoked potentials are highly sensitive when recorded from lower extremity muscle electrodes. We have developed an animal model for monitoring directly the thoracic and lumbar portions of the spinal cord which can be exposed to ischemia during descending aortic operation. To avoid difficulties in recording signals transmitted over multiple synapses (somatosensory, motor evoked potentials), we have found it suitable and possibly more accurate to monitor just that part of the spinal cord which in fact is of interest by directly stimulating and recording form epidural electrodes. We have termed this method, which was described in a similar way before, DSEP monitoring [21]. The accuracy of DSEP in indicating spinal cord ischemia and the predictive value regarding neurologic outcome was documented by our experimental study. Signal loss for a longer period of time (more than 40 minutes) and incomplete recovery of DSEPs are correlated highly with impaired neurologic outcome. Considering ideal monitoring conditions (laminectomies, no hemodynamic or temperature changes) in our animal experiments, which of course can hardly be achieved in the clinical situation, even minor differences of DSEP amplitude (less than 90% from baseline) were indicative of at least some degree of neurologic impairment. Based on our findings it seems worthwhile to focus interest on the development of a less invasive form of DSEP monitoring, possibly by using bipolar epidural catheters placed percutaneously, which could be applied clinically. Monitoring of DSEP permitted identification of important sources of blood supply of the spinal cord in our pilot experiments. Direct spinal cord evoked potential signal loss after creation of ischemia by aortic venting and rapid and consistent restoration (l to 3 minutes) after reperfusion permitted identification of CICA originating in the caudal half of the descending thoracic aorta. Our observations are compatible with the contention that spinal cord blood supply in pigs is plurisegmental and that most important arteries arise from the caudal thoracic aorta. We are doubtful of the existence of a true equivalent of the artery of Adamkiewicz in pigs [6,23]. The results of our study clearly demonstrate the negative effect of backbleeding of important intercostal arteriessimulated by venting segment II at a pressure of 0 mm Hg -on spinal cord perfusion. Due to the excellent collateralization of spinal cord blood supply in pigs, segment II had to be vented for longer time periods to cause complete loss of DSEP signal, which kept the intervals of total ischemia shorter than 40 minutes in most of the venting experiments (groups 1 and 2). Isolating segment II and neither venting nor perfusing it in two pilot experiments resulted in only a modest drop in mean segment pressure to 40 mm Hg and no changes in DSEP signals could be observed, a finding that supports the contention of excellent collateralization. However, perfusion of segment II with pressure control (70 mm Hg) was considered more accurate in simulating a situation of optimal spinal cord protection during cross-clamping (groups 3 and 4). If sufficient reperfusion was provided after venting (group 2), only minor impairment of neurologic outcome (median Tarlov score, 3) was seen. On the other hand, venting followed by a second insult to the spinal cord by ligation of important intercostal arteries correlated with significantly lower Tarlov scores (median, 1.5). All but one animal in group 1 were unable to stand up or walk on postoperative day 3. The detrimental effect of interruption of important intercostal arteries becomes evident. However, we were unable to demonstrate a significant impact of ligation of intercostal arteries without prior venting (group 3, median Tarlov score, 3.5) which is possibly due to adequate perfusion through collaterals in the absence of prior damage to the cord. Interestingly, ligation of CICA was followed by delayed recovery of 1 animal in group 3 (Tarlov 2 on postoperative day 2, Tarlov 3 on postoperative day 3) indicated by incomplete recurrence of DSEP amplitude (31%) at 120 minutes after cross-clamping, which differed significantly from all the other animals in this group. Spinal cord damage by isolated ligation of CICA in this particular animal was additionally confirmed by histologic examination (two ipsilateral infarcts in the midlumbar section). The importance of distal aortic perfusion in prevention of paraplegia was also demonstrated by this study. Simple aortic cross-clamping for 1 hour resulted in paraplegia in all animals (group 5), whereas various degrees of paraparesis but no cases of paraplegia were found in the distal perfusion groups (l through 4). Neurologic examination findings were supported by histologic examination. Both animals in group 5 showed complete infarction of spinal grey matter in sections from all levels. Morphologic sequelae of ischemia were also found in both animals from group 1 after venting and ligation, which correlated well with a high percentage (83.3%) of paraparesis with inability to stand up or walk (Tarlov 1 and 2) in this group. The results of this experimental study have prompted modifications of our operative strategy in descending

8 696 DAPUNT ET AL Ann Thorac Surg thoracic and thoracoabdominal aneurysm surgery. A system of distal aortic perfusion with a left atrium or left ventricle to femoral artery centrifugal pump bypass, including a reservoir that can be used with low-dose heparinization or even without heparin has been described by our group [24]. A number of physiologic factors suggest that the provision of distal aortic perfusion is essential for preservation of kidney and spinal cord function and improvement of postoperative cardiac function by unloading the left ventricle. Ischemia to the spinal cord is kept as short as possible by employing the principle of staged clamping, performing the proximal anastomosis while the aneurysm is still perfused by the Biomedicus pump. All intercostal or lumbar arteries are prepared and soft vascular clips are applied from cranial to caudal as spinal cord monitoring with somatosensory evoked potentials is performed. Loss of somatosensory evoked potential signal indicates spinal cord ischemia due to obstruction of intercostal arteries. Critical intercostal arteries are identified by recurrence of somatosensory evoked potential signal after serial removal of clips. Intercostal arteries found not to contribute to spinal cord blood supply are ligated and CICA are temporarily clipped before the aneurysm is opened. Thus the development of a steal phenomenon can be avoided and spinal cord perfusion can be rapidly restored by anastomosing CICA to the graft before the distal aortic anastomosis is performed. The impact of these technical modifications on neurologic outcome remains to be seen with increasing clinical application. We thank Howard H. Shiang, DVM, for conducting the neurologic examinations of animals, William F. Brett Ill, David A. 0'Allessandro, and Richard 1. Smith for their technical assistance, and Wendy [ackelow for preparation of the artwork. References 1. Chang JHT, Burrington JD. Coarctation of the aorta in infants and children. J Pediatr Surg 1972;7: Griepp RB, Ergin MA, Lansman SL, Galla JD, Pogo G. The natural history of thoracic aortic aneurysms. Semin Thorac Cardiovasc Surg 1991;3: Turney SZ, Attar S, Ayella R, et a1. Traumatic rupture of the aorta. A five year experience. J Thorac Cardiovasc Surg 1976;72: Crawford ES, Crawford JL, Safi HJ. Thoracoabdominal aortic aneurysms: Preoperative and intraoperative factors determining immediate and long-term results of operations in 605 patients. J Vase Surg 1986;3: Wadouh F, Arndt CF, Oppermann E, et a1. The mechanism of spinal cord injury after simple and double aortic crossclamping. J Thorac Cardiovasc Surg 1986;92: Svensson LG, Vasishta P, Robinson MF, Toshihiko U, Roehm JOF, Crawford ES. Influence of preservation or perfusion of intraoperatively identified spinal cord blood supply on spinal motor evoked potentials and paraplegia after aortic surgery. J Vase Surg 1991;13: Kouchoukos NT, Wareing TH, Izumoto H, Klausing W, Abboud N. Elective hypothermic cardiopulmonary bypass and circulatory arrest for spinal cord protection during operations on the thoracoabdominal aorta. J Thorac Cardiovasc Surg 1990;99: McCullough JL, Hollier LH, Nugent M. Paraplegia after thoracic aortic occlusion: Influence of cerebrospinal fluid drainage. J Vase Surg 1988;7: Coles JC, Ahmed SN, Mehta HU, Kaufmann JCE. Role of free radical scavenger in protection of spinal cord during ischemia. Ann Thorac Surg 1986;41: Cunningham IN, Laschinger JC, Spencer FC. Monitoring of somatosensory evoked potentials during surgical procedures on the thoracoabdominal aorta. IV. Clinical observations and results. J Thorac Cardiovasc Surg 1987;94: Livesay JJ, Cooley DA, Ventimiglia RA, et a1. Surgical experience in descending thoracic aneurysmectomy with and without adjuncts to avoid ischemia. Ann Thorac Surg 1985;39: Spencer FC, Zimmerman JM. The influence of ligation of intercostal arteries on paraplegia in dogs. Surg Forum 1958;9: Crawford ES, Palamara AE, Saleh SA, Roehm J0P. Aortic aneurysms: current status of surgical treatment. Surg Clin North Am 1979;59: Adamkiewicz A. Die Blutgefaesse des menschlichen Rueckenmarkes. Sitzungsberichte der Kaiserlichen Akademie der Wissenschaften. Wien 1882;85: Jellinger K. Zur Orthologie und Pathologie der Rueckenmarksdurchblutung. Wien: Springer; 1966:8-41, Wissdorf H. Die Ccfassversorgung der Wirbelsaeule und des Rueckenmarkes vom Hausschwein. In: Tieraerztliche Hochschule Hannover, Med vet Habilschrift. Berlin: Parey, 1970: Borst HG. Occlusion of intercostal and lumbar aortic branches using plastic plugs. Ann Thorac Surg 1987;44: Cooley DA, Baldwin RT. Technique of open distal anastomosis for repair of descending thoracic aortic aneurysms. Ann Thorac Surg 1992;54: Cunningham IN, Laschinger JC, Merkin BA, et a1. Measurement of spinal cord ischemia during operations upon the thoracic aorta: Initial experience. Ann Surg 1982;196: Laschinger JC, Cunningham IN, Cooper MM, et a1. Monitoring of somatosensory evoked potentials during surgical procedures on the thoracoabdominal aorta. 1. Relationship of aortic cross-clamp duration, changes in somatosensory evoked potentials, and incidence of neurologic dysfunction. J Thorac Cardiovasc Surg 1987;94: Laschinger JC, Owen J, Rosenbloom M, Cox JL, Kouchoukos NT. Direct noninvasive monitoring of spinal cord motor function during thoracic aortic occlusion: Use of motor evoked potentials. J Vase Surg 1988;7: Elmore JR, Gloviczki P, Harper M, et a1. Failure of motor evoked potentials to predict neurologic outcome in experimental thoracic aortic occlusion. J Vase Surg 1991;14: Wadouh F, Lindemann EM, Arndt CF, Hetzer R, Borst HG. The arteria radicularis magna as a decisive factor influencing spinal cord damage during aortic occlusion. J Thorac Cardiovase Surg 1984;88: Ergin MA, Galla JD, Lansman SL, Taylor M, Griepp RB. Distal perfusion methods for surgery on the descending aorta. Semin Thorac Cardiovasc Surg 1991;3: DISCUSSION DR CHRIS K. ROKKAS (Syracuse, NY): I enjoyed your presentation very much. The problem of identifying and reimplanting the critical intercostal or lumbar artery is a very important one, but unfortunately there is no solution on the horizon despite Dr Svensson's elegant work with hydrogen-detecting platinum electrodes. I would like to ask you a few questions. First, did you perform autopsies on these animals to actually identify the critical inter-

9 Ann Thorac Surg DAPUNT ET AL 697 costal or lumbar artery that gives origin to the arteria radicularis magna? In our experience with a similar model, the critical artery usually originates between L1 and L3, which sharply contrasts with your statement that the critical artery in your animals originated from the lower thoracic aorta. I would like you to comment on this. Second, you stated that in the animals that underwent ligation of the intercostal arteries, this was done at the end of the 60 minute period of aortic cross-clamping. I wonder why this was done, since in the clinical situation, ligation of intercostal or lumbar arteries is usually done in the beginning of the operation, not in the end. In any case, if the only difference between group 1 and group 2 animals is that intercostal ligation was performed at the end of the 60-minute period of cross-clamping in group 1 animals, there should not be much difference between these two groups up until the 60th minute, and I wonder why you found a difference in the evoked potentials at 30, 45, and 60 minutes during cross-clamping. Third, it is unclear to me how these data prompted you to modify your operative technique. I believe that the NYU group, as well as the Baylor group, have concluded that sensory evoked potential monitoring, in its present form, does not help in identifying the critical artery. I would like to know what is new in the meantime and how you actually modified your operative technique. DR DAPUNT: In our experience, we have not been able to demonstrate a porcine equivalent to the artery of Adamkiewicz. We believe that the spinal cord blood supply in the pig is plurisegmental and that the distal thoracic segment contains important segmental arteries. In all of our pilot experiments, only perfusion of the distal thoracic aorta (segment II) was sufficient to restore the evoked potential signal, after segmentally cross-clamping with distal perfusion. If there is an Adamkiewicz artery in the pig arising in the lumbar or sacral region, isolated perfusion of this artery was not sufficient to maintain spinal cord blood flow and restore the evoked potential signal in our model. Therefore, we concluded that important arteries that contribute to the spinal cord blood supply originate from the lower thoracic aorta (segment 11). The ligation of the intercostal arteries was done at the end of the cross-clamp period in an attempt to simulate the clinically encountered events of backbleeding or steal (venting), reanastomosis (nonligation), and total interruption of intercostal arteries (ligation). The observed differences in median values of spinal cord evoked potential amplitude between groups 1 and 2 that Dr Rokkas has referred to seem to be due to individual variability of the spinal cord blood supply among the relatively small number of experimental animals in each group. Regarding the question of clinical application, we certainly recognize the problems and limitations of using somatosensory evoked potentials in the intraoperative identification of critical arteries. We believe that direct epidural spinal cord recording as used in these experiments eliminates most of these problems and is a sensitive indicator of early spinal cord ischemia. If a similar technique can be adapted for consistent clinical application, it will facilitate identification of the critical intercostal arteries. That is why we continue to work with this model. The data presented also point to the importance of the intercostal steal phenomenon in spinal cord ischemia. Currently we use intraoperative somatosensory evoked potentials monitoring during sequential interruption of the intercostal arteries before opening the aortic segment to be resected to determine whether important intercostal arteries are present in the segment and at the same time to prevent intercostal steal. DR JOHN A. MACOVIAK (Rancho Mirage, CAl: I have two questions. The first is: were you able to establish a correlation between the appearance of the intercostal arteries, namely the size and the amount of backbleeding, with the location of the change in spinal cord evoked potentials? Second, would there be a difference, based on your model, in how you would handle an acutely dissected aorta that has ruptured versus a chronic atherosclerotic aneurysm? DR DAPUNT: Because the aorta was not opened in these experiments, it was not possible to establish a direct correlation between the amount of backbleeding and the size of the intercostals. The operative strategy outlined applies to elective cases. We are all familiar with the difficulties involved in application of this sort of spinal cord monitoring in emergency situations, especially in the presence of a ruptured aneurysm, where immediate control of exsanguination becomes the primary objective. DR VALAVANUR A. SUBRAMANIAN (New York, NY): I just have a clinical question. I know that there is a lot of interest in the steal phenomenon and venting of the aortic segment at the places supplied by the noncritical coronary intercostal arteries. Can you tell us if you have observed a clinical correlation between the cerebrospinal fluid pressure monitoring and where you do this venting procedure? DR DAPUNT: Cerebrospinal fluid pressures were not measured during these experiments and we do not use cerebrospinal fluid pressure monitoring clinically. DR LARS G. SVENSSON (Burlington, MA): This fine study you performed confirms the belief from clinical and previous animal work that distal aortic perfusion and the segmental arteries influence spinal cord injury after aortic operations. Concerning atriofemoral bypass, we recently reported at the New England Vascular Society meeting on 100 patients undergoing thoracoabdominal aneurysm repairs and found that atriofemoral bypass does reduce the incidence of paralysis on multivariate analysis when age is controlled for. We have also found that atriofemoral bypass is protective for descending thoracic repairs. We found that for both descending thoracic and thoracoabdominal repairs, if hydrogen is injected into the atriofemoral bypass machine during staged repairs, the hydrogen gets to the spinal cord, confirming that blood circulation is preserved to the distal spinal cord. Concerning the issue raised about steal, I think that your data confirm the findings of a previous study in the Japanese literature, which reported that by using partial bypass and keeping intercostals open, somatosensory evoked potentials disappeared. If the intercostals were occluded, however, either with catheters, which was Dr Crawford's technique, or by occluding the ones that were not going to be preserved by oversewing them, then perfusion was maintained to the cord and somatosensory evoked potentials were preserved. Your study has confirmed this elegantly. As for your comments about the critical vessels of the spinal cord in the pig, from my dissections in the pig, if one defines the artery of Adamkiewicz as the biggest radicular artery with a hairpin bend that tends to perfuse the distal part of the spinal cord and not the thoracic spinal cord, then in the pig the artery of Adamkiewicz usually arises from the lumbar segments. In humans, it usually arises from the lower thoracic intercostal arteries and sometimes from L1 or L2. If you did not remove the lower vertebrae, I would encourage you to examine the lumbar radicular arteries. Certainly, if you can stain the radicular arteries and the anterior spinal artery with a dye, I think that you will find the artery of Adamkiewicz located lower on the spinal cord in pigs.