Influence of segmental arteries, extent, and atriofemoral bypass on postoperative paraplegia after thoracoabdominal aortic operations

Transcription

1 Influence of segmental arteries, extent, and atriofemoral bypass on postoperative paraplegia after thoracoabdominal aortic operations Lars G. Svensson, MD, Phi), Kenneth R. Hess, MS, Joseph S. Coselli, MD, and Hazim J. Sail, MD, Burlington, Mass., and Houston, Texas Purpose: The purpose of this article was to study the influence of either reattachment or oversewing of patent segmental intercostal or lumbar arteries, extent of aneurysm, and atriofemoral bypass on the incidence of postoperative paraplegia/paraparesis in patients at high risk with type I or II thoracoabdominal aneurysms. Methods: Data were prospectively collected on 99 patients undergoing type I or II thoracoabdominal aneurysm repairs, including exact extent of repair and whether atriofemoral bypass was used. Patency of intercostal arteries from T3 to T12 and lumbar arteries from L1 to L4 were checked by intraoperative inspection. If the arteries were patent, note was taken of whether they were reattached to the new aortic prosthesis. Postoperative neurologic motor function was graded daily for the first 5 days, and the worst score in the first 30 postoperative days (POD) was used for analysis. Results: Ninety-five of 99 (96%) patients were 30-day survivors. By POD 30, 31 of 98 (32%) patients had had a neurologic deficit. There was no difference in the incidence of deficits according to whether lumbar or intercostal arteries were reattached, ignoring the effect of patency of the arteries. Of greater importance, however, was whether patent segmental arteries were oversewn at specific levels. Thus, for patients who had one or more arteries at Tll, T12, or L1 oversewn (often because they could not be reattached), a deficit developed in 11 of 23 (48%) patients versus 20 of 75 (27%) patients who did not have patent arteries or had all patent arteries reattached (i0 = 0.05, odds ratio = 2.5). More specifically, if all arteries at this level were oversewn, a neurologic deficit developed in 63% of patients versus 23% if all their arteries were reattached (p = 0.01). Reattachment of patent arteries at individual levels from T7 to L4 showed a trend toward a lower risk of deficits but did not reach statistical significance. On multivariate analysis, atriofemoral bypass was associated with a lower risk of paralysis (p = 0.068), and significantly so when controlled for age (p = , odds ratio 0.287). Subgrouping of extent type I thoracoabdominal aneurysms resulted in an incidence of paralysis of 14% (3/22) for subgroup A and 23% (5 of 22) for subgroup B compared with 43% (23 of 55) for type II thoracoabdominal aneurysms (type I [8 of 44 18%], versus type II [p = ]). Conclusion: Patients with no or few patent segmental arteries in the aortic segment being replaced have a lower risk of neurologic deficits, compared with those with patent arteries. Every effort should be made to reattach all arteries at Tll, T12, and L1 and, when possible within the constraints of technical feasibility and time, also those from T7 to L4. Preoperative angiography or intraoperative hydrogen testing may better identify the arteries that need to be reattached. When feasible, atriofemoral bypass appears to be protective, particularly when sequential damping and segmental repairs can be performed. (J VASC SURG 1994;20: ) From Lahey Clinic, Burlington, and Baylor College of Medicine Houston. Sponsored by Edward R. Jewell, MD. Presented at the Twentieth Annual Meeting of the New England Society for Vascular Surgery, Cambridge, Mass., Oct , Reprint requests: Lars G. Svensson, MD, PhD, Director, Center for Aortic Surgery, Lahey Clinic, 41 Mall Rd., Burlington, MA Copyright 1994 by The Society for Vascular Surgery and International Society for Cardiovascular Surgery, North American Chapter /94/$ /6/

2 256 Svensson et al. JOURNAL OF VASCULAR SURGERY August 1994 Controversy exists about the influence of either reattachment to a new aortic graft or oversewing of segmental intercostal or lumbar arteries, or the use of atriofemoral bypass, on the postoperative incidence of paraplegia or paraparesis after thoracic and thoracoabdominal aortic aneurysm operations. 1~6 In 1979, on the basis of his initial experience with 27 thoracoabdominal aortic aneurysm repairs, Crawford ~ came to the conclusion that he was able to reduce the incidence of postoperative paraplegia from 38% to 16% in those patients whose intercostal arteries were reattached. The basis for reattaching intercostal arteries was if they were large or if there was apparent backbleeding from the arteries. In a subsequent publication, 2 however, with a larger cohort of patients and with more statistical power, he noted that in patients whose intercostal arteries were reattached, the incidence of paraplegia and paresis was increased. This was further supported by the stepwise logistic regression model that identified reattachment of intercostal arteries as a risk factor for postoperative paraplegia or paresis. What superficially appeared to be a paradox required further investigation. Also, in our experience with 832 descending thoracic aneurysm repairs, atriofemoral bypass was associated with a lower risk of paraplegia or paresis, raising the issue of whether atriofemoral bypass was protective for thoracoabdominal repairs. 6 This prospective study was therefore undertaken to evaluate the influence of intercostal and lumbar arteries on the postoperative incidence of paraplegia or paresis. For the purposes of the study we classified the segmental arteries into three types: first, those that were occluded by atheroma within the aorta or by thrombus and thus were not available for reattachment; second, those that were present, but could not be reattached for various technical reasons; and third, those that could be successfully reattached. Because nothing could be done about the arteries that were not present, the crux of the question was whether patent intercostal or lumbar arteries that were either oversewn or reattached influenced the postoperative incidence of paraplegia or paresis. In addition, we evaluated the use of atriofemoral bypass and the exact extent of the aneurysm repair. METHODS Ninety-nine patients were entered into a prospective study of the influence of the segmental arteries on postoperative paraplegia. These patients have been the subject of other reports) 7 Briefly, all were patients of the authors or the late E. Stanley Crawford, MD, undergoing Crawford type I or type II thoracoabdominal aortic aneurysm repairs. Aortic aneurysms extending from the proximal half of the descending aorta to the renal arteries were classified as type I, and those extending below the renal arteries were classified as type II. During the operation the standard technique for repairing thoracoabdominal aortic aneurysms was used. During the period of aortic cross-clamping, all the patent arteries between T3 to T12 and lumbar arteries L1 to L4 were noted. This was done by counting the intercostal space above or below the site of resection of the sixth rib. Because it can be difficult to precisely locate from which vertebral level the arteries arise, this was a convenient method of identifying the vertebral level. Also, an attempt was made, whenever possible, to reattach the vessels from T7 down to L1. Nevertheless, sometimes the aortic wall was not strong enough to be reattached, or calcification around the segmental artery did not allow for reattachrnent to the aortic graft because of a very fragile aortic wall, such as in patients with Marfan's syndrome, or because of acute dissection or rupture. Thus this group of patients formed the subset in whom patent segmental intercostal lumbar arteries were present but could not be reattached for technical reasons. The remaining operative technique was as described previously. After operation the patients were examined dally for the first 5 days after operation according to their motor function of the lower limbs, with the worst score in the first 30 postoperative days being used for analysis. Their muscular strength was scored as follows: 0, no motor function; 1, slight movement, usually only toes; 2, moved most of the lower limb, but not against gravity; 3, lift lower limb against gravity, but weak; 4, normal function. There were 38 women and 61 men with a median age of 63.5 years (range 18 to 78 years). Of the 99 patients, 55 (56%) had Crawford type II thoracoabdominal aortic aneurysms. The 44 (44%) patients with type I repairs were subgrouped into those with repairs ending at the celiac artery without a posterior extension (IA), and those that had a beveled graft extending below the celiac artery but ending before the renal arteries (IB). Patients who had reimplantation of the visceral arteries as an island, thus with extension below the renal arteries, were classified as type II. The site of the proximal anastomosis was also noted and whether the distal arch was clamped together with the left subclavian artery. Patients with aneurysm repairs below the midpoint of the descending thoracic aorta, which we defined as vertebral level T6, were classified as type III repairs and were not included in the study. T6 is a convenient level because

3 JOURNAL OF VASCULAR SURGERY Volume 20, Number 2 Svensson et al. 257 Table I. Univariate variables significantly associated with postoperative paraplegia/paraparesis (based on 98 survivors) Variable Group Patients PAR/PLG p Value* Age (yr.) (22%) (43%) Extent I 44 8 (18%) replaced II (43%) Aortic cross (7%) clamp time (rain.) (12%) (48%) (55%) Postoperative Yes 31 8 (26%) hypotension~- No 46 2 (4%) PAR/PLG, Paraparesis/paraplegia. *Pearson chi-square test. tpatients with immediate deficits excluded we perform the incision for thoracoabdominal repairs in the sixth intercostal space, and therefore it is easy to classify an aneurysm on the basis of the proximal extent of the aneurysm in relation to the incision. Types I and II aneurysms originate above T6. Thirty-four patients (34%) had aortic dissection, either DeBakey type I or DeBakey type III. Atriofemoral bypass was used in 39 patients (39%), mostly for performance of the proximal anastomosis, but sometimes also during the reattachment of the intercostal arteries, and thereafter stopped. Intercostal or lumbar arteries were reatrached in 78 patients (79%). Reoperation for postoperative bleeding was required in six patients (6 of 98, 6%), and postoperative hypotension occurred in 31 patients who were neurologically intact immediately after operation (31 of 77, 40%). Thirteen patients also had aneurysms or dissection involving the ascending aorta or transverse aortic arch. The median aortic cross-clamp time was 44.5 minutes (range 17 to 109 minutes). The data were entered in a computer and analyzed by BMDP Statistical Software package (BMDP Statistical Software, Inc., Los Angeles, Calif.). A Pearson chi-squared test and Fisher exact test were used to assess a univariate statistical influence of reattachment or failure to reatrach intercostal or lumbar arteries. Multiple stepwise logistic regression analysis was used to identify the variables associated with the neuromuscular deficit occurring within 30 days of surgery. For this model, the only variable concerning segmental arteries used was whether patent arteries were reattached at level Tll to L1. RESULTS The 30-day survival rate in the 99 patients monitored prospectively was 96% (95 of 99). The cause of death in the four patients who died within 30 days was myocardial infarction during operation in one patient, sudden death in one, respiratory failure associated with kidney failure in one patient, and multiple organ failure in one patient. Paraplegia or paraparesis was observed immediately after surgery in 21 patients (21%, 21 of 98, excluding one intraoperative death), and a deficit later developed in 10 additional patients between postoperative days 3 and 22, for a total of 31 patients with paraplegia or paraparesis (31 of 98, 32%). The deficit scores were either 0, 1, or 2 in 21 patients or 3 in 10 patients. Table I shows the influence of the factors significant by univariate analysis in postoperative paraplegia or paraparesis, which included the (p < 0.05): age, extent replaced, aortic cross-clamp time, oversewing of arteries Tll to L1, and postoperative hypotension. Of note was the use of atriofemoral bypass in 39 patients in whom there was a trend toward a lower incidence of paraplegia or paresis namely 26% (10 of 39) versus 36% (21 of 59) (p = 0.3). In examining the association between study factors it was noted that aortic dissection was more frequent in younger patients (p = ); atriofemoral bypass was more frequently used in younger patients (p = ) and for patients with aortic dissection (p < ); the aortic cross-clamp times were longer in patients with aortic dissection (p = ), type II extent aneurysms (p < ), and for atriofemoral bypass (p = ); and younger patients were less likely to have arteries at segment Tll to L1 oversewn (p =0.035). Aortic cross-clamp time increased with the number of patent segmental arteries (Rs = 0.49, p < ) and number of segmental arteries reattached (Rs = 0.58,p < ), including at level Tll to L1 (p < ), although there was no correlation with the number of arteries oversewn. The number of patent segmental arteries were increased with aortic dissection (p = ), and more were reattached with aortic dissection (p = 0.011) and type II extent (p = 0.024). There did not appear to be any correlation between the number of patent segmental arteries and age or extent, although patients with type H aortic dissection had the most patent segmental arteries. On stepwise logistic regression to reduce the effects of variable associations, the multivariate fac-. tors associated with a deficit were age (p = ), the aortic cross-clamp time (p < ), reattach-

4 258 Svensson et al. JOURNAL OF VASCULAR SURGERY August 1994 Table II. Effect of reattachment or oversewing intercostal or lumbar arteries Variable Paraplegia No. % p Value OR None oversewn to 8 oversewn None reattached to 7 reattached No arteries present Arteries present * % or less re attached % or more * reattached Tll-L1 None oversewn to 3 oversewn T11-L1 None present All reattached Not all reattached None reattached TII-LI None present None reattached * 5.55 Tll-L1 All reattached None reattached * 4.16 *Fisher exact test. ment of arteries at Tll to L1 (p = , reduced risk), and atriofemoral bypass (p = 0.068, reduced risk). There was a highly significant correlation between age and the use of atriofemoral bypass (p = ). Thus, when age was controlled for and entered into the stepwise logistic regression model after keeping the variable atriofemoral bypass in the model, the variable atriofemoral bypass was significant at ap value of , with a reduction of risk at an odds ratio (OR) of Paradoxically, the incidence of paraplegia or paresis according to whether intercostal arteries were reattached was 35% (27/78) versus 20% (4/20), (p = 0.2), respectively, not taking into consideration whether arteries were patent. Table II shows the effect of reattaching or oversewing intercostal or lumbar arteries in the patients. Of note, in the patients in whom no intercostal or lumbar arteries were oversewn, postoperative paraplegia or paraparesis developed in only one patient (1 of 17, 6%) (p = 0.013, OR = 0.1). Similarly, if no arteries were present, thus none needed to be oversewn or reattached, then paraplegia or paraparesis did not develop in any of the eight patients (0/8) (p = 0.05). Table III. Paraplegia or paraparesis according to number of actual arteries reattached or oversewn at Tll to L1 for the entire group of 98 patients Variable PLG N p Value OR Total oversewn * arteries 0.4 Total reattached * arteries *Fisher exact test. Furthermore, if 75 % or more of the patent intercostal or lumbar arteries were reattached, then paraplegia or paraparesis developed in only 5% of patients (1 of 20) (p = , OR was for not having them reattached, [p = Fisher's exact test]). If, in the aortic segment between Tll and L1, none of the intercostal or lumbar arteries were oversewn, the incidence of paraplegia or paraparesis was 27% (20 of 75) as compared with those patients who had one (7 of 15, 47%) or two and three (4 of 8, 50%) arteries in this segment oversewn (p--0.05, OR = ). If i to 3 arteries were oversewn, a neurologic deficit developed in 11 of 23 patients (68%) (p = 0.05, OR = 2.52). Further detailed analysis at the level of the aortic segment Tll to L1 showed that the risk for development of postoperative paraplegia or paresis increased according to whether no arteries were present (6 of 26, 23%) as compared with whether all arteries were reattached (14 of 49, 29%), whether not all arteries were reattached (6 of 15, 40%), and whether none of the present arteries were reattached (5 of 8, 63%); the patients who had no arteries reattached were at highest risk (p = , contingency table randomization test). Furthermore, comparison within this group showed that if no vessels were present, and thus if none needed reattachment, the incidence of paraplegia or paresis was 23% (6 of 26) versus 63% if none of the present arteries were reattached (5 of 8) (p = 0.05, OR = 5.56, Fisher's exact test). This was further supported by the finding that if all the vessels were reattached, 29% developed paraplegia or paraparesis (14 of 49) versus 63% (5 of 8) in those patients who had no arteries reattached (p = 0.07, odds ratio 4.17, Fisher's exact test). Table III shows the overall results of oversewing the arteries in the segments between Tll and L1 and reattaching the vessels in this segment and the postoperative incidence of paraplegia or paresis. This shows a significant protective effect for vessels reattached in this segment (p = , OR = 0.4).

5 JOURNAL OF VASCULAR SURGERY Volume 20, Number 2 Svensson et al. 259 Table IV. Segmental arteries and postoperative paraplegia/paraparesis Present Reattached Reattached with PLG Artery No. (%) No. (%) No. (%) PLG Operscwn No. No PLG % PLG T3 2 (2%) 0 (0% T4 36 (37%) 3 (9% T5 57 (58%) 4 (7% T6 ~ 66 (67%) 7 (11% T7 67 (68%) 17 (25% T8 69 (70%) 35 (51% T9 61 (62%) 46 (75% T10 65 (66%) 56 (86% Tll 66 (67%) 55 (83% T12 49 (50%) 39 (80% L1 26 (27%) 21 (81% L2 12 (12%) 6 (50% L3 14 (14%) 6 (43% L4 5 (5%) 3 (60% (33%) (50%) (71%) (24%) (37%) (35%) (38%) (31%) (28%) (24%) (33%) (50%) (0%) *3-way chi-square, p = (two-tailed). PLG, paraplegia/paraparesis. Table IV shows the percentage of intercostal or lumbar arteries present at each level. Of note, the percentages of vessels that were reattached reflect our policy of reattaching arteries between T8 and L1 whenever possible. Also shown is the paraplegia or paraparesis incidence, according to vessels reattached at each level and the paraplegia or paraparesis rate according to vessels oversewn at each level. Because of the small number of arteries at each level, no statistically significant difference was found at any particular level between reattached and oversewn vessels. Nevertheless, examination of the data at Tll to L1 clearly shows the tendency for higher paraplegia or paraparesis rates for vessels oversewn in these segments as opposed to those that were reattached. At T6, there was a trend toward a higher incidence of postoperative paraplegia or paraparesis in the patients who had vessels at T6 reattached (71%, 5 of 7) versus in those who had T6 oversewn (34%, 20 of 59) (p = 0.07, Fisher's exact test). Of further interest was a finding that chi-squared contingency table comparison at this level showed an incidence of paraplegia/paraparesis of 19% in patients with no vessels present versus 34% in those who had oversewn vessels and 71% for those who had reattached vessels. This reached statistical significance at p = for the two-tailed chi-square test. Of note, the thoracotomy incision was made in the sixth intercostal space (T6 level) and then spread widely apart. This may have damaged the T6 intercostal artery, accounting for the higher incidence of deficits when this level was reattached. The incidence of neuromuscular deficits was 18% (8 of 44) for type I and 43% (23 of 54) for type II aneurysms (p = , OR = 3.3). For type IA it was 14% (3 of 21) and for type IB 23% (5 of 22) (p = ns). When the proximal extent of the aneurysms was analyzed, either beginning immediately at the left subclavian artery or at a lower level but before T6, the aneurysms originating at the left subclavian artery tended to be at greater risk, but this did not reach statistical significance. The incidence of deficits did not differ according to type I and type III DeBakey aortic dissection. DISCUSSION Intuitively, it appears logical that reattachment of the blood supply to an organ should protect the organ supplied by the arteries, in this case, the spinal cord. Nonetheless, this has to be balanced with the fact that with an increased time of aortic crossclamping, the risk of postoperative paraplegia or paresis increases for both thoracoabdominal aortic aneurysm repairs and descending thoracic aortic aneurysm repairs. 4"7 Therefore the surgeon is left with the quandry of either not reattaching the intercostal and lumbar vessels with a shorter aortic cross-clamp time or reattaching intercostal and lumbar vessels with a longer aortic cross-clamp time as shown in this study. Whereas Crawford advocated the reattachment of intercostal arteries on the basis of the operative findings, 1'2'12 Williams 8 and Kieffer 9 were proponents of preoperative angiography for the localization of vessels that supply the spinal cord and their reattachment at the time of surgery. On the basis of the spinal cord anatomy studies in eight human cadavers and the previous studies reported, we modified our technique of thoracoabdominal aneu-

6 260 Svensson et al. JOURNAL OF VASCULAR SURGERY August 1994 rysm surgery repairs in 1988 by reattaching all patent vessels between T7 and L1 whenever possible. 4,s,13,~4 The reason for reattaching this segment was that, in 90% of patients, this would revascularize the artery of Adamkiewicz, which is the largest of the radicular arteries supplying the spinal cord. As found in this study, the problem with reattaching segmental arteries is that it entails a prolonged aortic cross-clamp time to reattach the vessels. Furthermore, other thoracic radicular arteries arising at a higher level that supply the spinal cord may not be reattached with this technique. ~,ls Addressing this problem, we performed porcine experiments using a technique with hydrogen in solution and a platinum electrode alongside the spinal cord, which enabled the intraoperative identification of the intercostal or lumbar arteries that supplied the spinal cord.14 This technique was then used to either preserve or divide lumbar arteries that were shown to supply the spinal cord with blood. 4 In the animal experiments, those animals randomized to division of the vessels supplying the spinal cord developed paraplegia or paraparesis, whereas those that had them preserved and had the others divided remained neurologically normal, confirming the influence of the lumbar and intercostal vessels on the incidence of postoperative paraplegia or paresis. 4 This technique of use of hydrogen and a platinum electrode was then used in patients. 4,~6 Results showed that the technique was both accurate and safe. This study confirms the conclusion of Crawford, 1 published in 1979 on a series of 27 patients, that reimplantation of patent intercostal arteries reduces the postoperative risk of paraplegia or paraparesis. It should be noted, however, that if a patient has no or few patent intercostal or lumbar arteries, as is often the case with thoracoabdominal degenerative (atherosclerotic) aneurysms, the risk of postoperative paraplegia or paraparesis is lower than for a patient who has patent arteries. This explains the apparent paradox of reimplantation of intercostal arteries not reducing the risk of postoperative neurologic injury reported in previous studies. 2'~7 The finding that patency of segmental arteries is the critical factor partly explains the higher risk of neurologic injury in patients who have aortic dissection and extensive thoracoabdominal aortic aneurysms because these patients have more patent arteries that are affected by aortic repairs. Moreover, if the patent intercostal or lumbar arteries are not reimplanted, the risk of neurologic injury is significantly further increased. Nevertheless, as has been emphasized previously, 2,47,1s,17-19 the failure of reimplantation of patent intercostal or lumbar arteries is not the sole cause of postoperative paraple- gia and paraparesis. As the stepwise logistic regression model shows (Table I), the aortic cross-clamp time is also a significant predictor of postoperative paraplegia or paraparesis. Although the extent of aneurysm resection was not a significant factor on stepwise logistic regression analysis in this study of only Crawford type I and II aneurysms, the extent of aneurysmal replacement was a significant factor in a larger study 2 also including types HI and IV aneurysms. This was probably related to the number of intercostal and lumbar arteries that have to be preserved in more extensive aneurysms, particularly type II aneurysms, to maintain adequate spinal cord perfusion. Because the segmental intercostal and lumbar arteries are an important factor associated with postoperative paraplegia or paraparesis, and reattachment of the arteries reduces the risk on multivariate analysis, several questions arise. First, which patent intercostal and lumbar arteries need to be reattached? Our research suggests that Tll, T12, and L1 segmental arteries must be reattached at the time of surgery if they are present and patent. Also, the study shows that failure to reattach these vessels is statistically associated with a significantly higher risk of paraplegia or paresis. Furthermore, the tendency for higher postoperative incidence of paraplegia or paresis is present if the intercostal arteries from T7 to T10 are not reattached at the time of' surgery. Therefore these vessels should also be attached when feasible. These findings are in keeping with our understanding of the anatomic blood supply to the spinal cord} 3-1s The intercostal or lumbar arteries give off small branches known as the radicular arteries, which enter the foramcn of the vertcbral bodies and accompany the nerve roots to the spinal cord. The largest of these radicular arteries, the artery of Adamkiewicz, has been found from anatomic dissections of the human spinal cord to arise in 90% of patients between T7 and L1} 3'1~ Thus both the findings of this study and the anatomic blood supply of the spinal cord add further credence to anastomosing the vessels in the segment from T7 to L1. Second, how should the vessels be reattached? The technique developed by Crawford 1'12 was to incise the graft longitudinally and excise an oval from the graft, and then reattach the aortic wall containing the intercostal arteries to the aortic tubular graft. Reattaching all the intercostal arterics from T7 to L1, however, does take some time. Therefore a balance has to be achieved between reattaching intercostal arteries and lumbar arteries and the time spent reattaching them. The average time required for the intraoperative identification of the vessels supplying

7 JOURNAL OF VASCULAR SURGERY Volume 20, Number 2 Svensson et al. 261 a spinal cord with hydrogen has been 5.8 minutes. This technique is still being evaluated, but in the initial patients in whom the method was used, it has been found to be accurate in determining the vessels that supply the spinal cord and may reduce the time required for reattachment of multiple vessels. 4 Third, how successful is the reattachrnent of segmental intercostal and lumbar arteries as a patch in maintaining blood flow to the spinal cord? The risk is that the insertion of catheters, such as Fogarty catheters, into intercostal or lumbar arteries to reduce backflow from these vessels during reattachment might damage these arteries, resulting in either rupture or thrombosis of these vessels from tears in the arteries. Furthermore, there is the risk that during the reattachment of the patch to the aortic graft, deep sutures placed in the aortic wall or too close to the ostia of the arteries might result in occlusion of the arteries. Moreover, because the aortic wall has been manipulated and often traumatized, there is the risk that when blood flow is reestablished to the small intercostal and lumbar arteries, thrombosis of the vessels will occur. Our postoperative, highly selective angiograms obtained in patients who had intraoperative hydrogen studies suggest that a number of reattached intercostal and lumbar arteries did not remain patent after reattachment. 4 This is a technical problem that must be overcome to achieve successful reestablishment of blood flow to the spinal cord. Factors that appeared to be important in maintaining patency of these vessels after reattachment appeared to be to avoid the use of Fogarty catheters in the intercostal and lumbar arteries during aortic crossclamping, to allow backbleeding from the vessels; to attach intercostal and lumbar arteries with at least a 0.5 to i cm margin from the ostia; and to avoid deep "bites" in the aortic wall, particularly in the posterior left lateral part so that intercostal and lumbar arteries are not occluded in the anastomosis. It should further be noted that the artery of Adamkiewicz usually arises from the left of the pair of segmental arteries at any particular level. 13 We have also observed that the risk of thrombosis of the vessels appears to be higher if the proximal aorta is undamped and if a clamp is then placed below the reattached intercostal arteries before reestablishing blood flow to the visceral vessels. The possible reason for this is that a large clot will often form in the aortic graft because of the inadequate blood outflow through the intercostal patch alone. Although it would be tempting to advocate the use of intravenous heparin to reduce the risk of thrombosis of the small intercostal and lumbar arteries, the complications of heparinization, including bleeding into the left lung and coagulopathy, mitigate against its routine use. Therefore we do not use heparin for descending and thoracoabdominal aneurysm repairs even when attiofemoral bypass is used. The fourth question is whether an alternative method of reattachment of intercostal and lumbar arteries should be considered. Connolly and colleagues 2 have been strong advocates of extensive beveling the aortic graft and even performing long anastomoses on an aortic patch to include the intercostal and lumbar arteries. Although this technique is attractive, one has to be careful in patients with aortic dissection and particularly in those patients with Mat-fan syndrome, that an excessive amount of aortic wall is not left behind. On late follow-up, aneurysmal involvement of the residual aorta will develop in some of these patients. Clearly, when the extent of aorta containing the segmental arteries is short, a beveled anastomosis is both easy and expeditious to perform for the distal anastomosis. However, when the aortic graft has to be taken to below the visceral arteries, this is not advisable. Other alternatives that can be used are the placement of small lumen tube grafts from the aortic graft to individual intercostal or lumbar arteries. The concern of using such grafts is that the size of the Gortex or Dacron graft, usually 6 to 8 mm in size, is a mismatch to the intercostal or lumbar arteries which are usually less than 2 mm in size, and more often than not, 1 mm in size. Thus graft thrombosis poses a significant risk factor when using individual tube grafts to the vessels because of the low flow rate in these grafts. Another alternative technique is one that has been advocated by Williams 21 for patients with aortic dissection. Using this approach, the septum is excised and the aorta is repaired by narrowing it down without replacing the aortic wall with a tube graft. This is an attractive technique in selected patients and long-term follow-up wilt determine its role in the management of patients with aortic dissection. In patients with inflammatory aneurysms or patients with heavily scarred aortic walls, the aneurysm wall can be partially excised and the aorta repaired with a technique similar to that advocated by Matas.22 We have described using this technique in an infant who had a thoracoabdominal aneurysm that was repaired by preservation of part of the aortic wall. 23 We have found that in patients undergoing descending thoracic aortic repairs 6 or thoracoabdominal repairs, 7 the incidence of postoperative kidney failure is significantly reduced by the use of atriofemoral bypass. For patients having descending aortic repairs requiring more than 40 minutes of aortic clamping, 6 we have also found that atriofem-

8 262 Svensson et al. JOURNAL OF VASCULAR SURGERY August 1994 oral bypass is significantly protective (p < 0.05) against neurologic injury. This study in patients having thoracoabdominal repairs has now also shown that atriofemoral bypass is associated with a lower risk of neurologic injury, even when not used for the entire period of aortic cross-clamping. With modification of operative technique, namely with segmental repair and sequential clamping when feasible, and with atriofemoral bypass for the entire period of clamping, the risk of neurologic injury after thoracoabdominal repairs will probably be even more reduced. 7 These findings showing that atriofemoral bypass is protective are supported by our previous animal work showing that distal aortic perfusion improves spinal cord blood flow measured by radioactive microspheres and reduces the risk of neurologic injury, including spinal cord dysfunction as shown by spinal cord motor-evoked responses. 4,19 In conclusion, although the intercostal and lumbar arteries play a critical role in maintaining spinal cord blood supply after thoracoabdominal aneurysm repair, technical problems remain about the identification of the vessels that need to be reattached at the time of surgery and also how these vessels should be reattached. Further research is required to establish which techniques will be best for identifying and reattaching the vessels that supply the spinal cord. For type I and II thoracoabdominal aneurysms, atriofemoral bypass appears to be protective. REFERENCES 1. Crawford ES, Palamara AE, Saleh SA, Roehm JOF. Aortic aneurysm: current status of surgical treatment. Surg Clin N Am 1979;59: Crawford ES, Crawford JL, Sail HJ, et al. Thoracoabdominal aortic aneurysms: preoperative and intraoperative factors determining immediate and long-term results of operation in 605 patients. J VASC SURG 1986;3: Cambria RP, Brewster DC, Moncure AC, et al. Recent experience with thoracoabdominal aneurysm repair. Arch Surg 1989;124: Svensson LG, Patel V, Robinson MF, Ueda T, Roehm JOJ, 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 VASC SURG 1991;13: Svensson LG, Crawford ES, Hess KR, Coselli JS, Sail HJ, Dissection of the aorta and dissecting aortic aneurysms: improving early and long-term surgical results. Circulation 1990;82(5 Suppl):iV Svensson LG, Crawford ES, Hess KR, Coselli JS, Sail HJ. Variables predictive of outcome in 832 patients undergoing repairs of the descending thoracic aorta. Chest 1993;104: Svensson LG, Crawford ES, Hess KR, Coselli JS, Sail HJ. Experience with 1509 patients undergoing thoracoabdominal aortic operations. J VAsc SURG 1993;17: Williams GM, Perler BA, Burdick JF, et al. Angiographic localization of spinal cord blood supply and its relationship to postoperative paraplegia. J VASC SURG 1991;13: Kieffer E, Richard T, Chiras J, Godet G, Cormier E. Preoperative spinal cord arteriography in aneurysmal disease of the descending thoracic and thoracoabdominal aorta: preliminary results in 45 patients. Ann Vasc Surg 1989;3: Acher CW, Wynn MM, Hoch JR, Popic PM, Archibald J, Turnipseed WD. Combined use of cerebral spinal fluid drainage and naloxone reduces the risk of paraplegia in thoracoabdominal aneurysm repair. J VAse SURG 1994;19: Hollier LH, Moore WM. Avoidance of renal and neurologic complications following thoracoabdominal aortic aneurysm repair. Acta Chir Scand Suppl 1990;555: Crawford ES, Crawford JL. Diseases of the aorta including an atlas of angiographic pathology and surgical technique. Baltimore: Williams and Wilkins, Svensson LG, Klepp P, Hinder RA. Spinal cord anatomy of the baboon: comparison with man and implications on spinal cord blood flow during thoracic aortic cross-clamping. S Afr J Surg 1986;24: Svensson LG, Patel V, Coselli JS, Crawford ES. Preliminary report of localization of spinal cord blood supply by hydrogen during aortic operations. Ann Thorac Surg 1990;49: Svensson LG, Crawford ES. Aortic dissection and aortic aneurysm surgery: clinical observations, experimental investigations and statistical analyses: part I. Curr Probl Surg 1992;11: Svensson LG, Crawford ES, Patel V, Jones JW, DeBakey ME. Spinal cord oxygenation, intraoperative blood supply localization, cooling and function with aortic clamping. Ann Thorac Surg 1992;54:74-9, 17. 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 VAse SURG 1991;13: 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: Svensson LG, Von Ritter CM, Groenveld HT, et al. Cross-clamping of the thoracic aorta: influence of aortic shunts, laminectomy, papaverine, calcium channel blockers, allopurinol, and 'superoxide dismutase on spinal cord blood flow and paraplegia in baboons. Ann Surg 1986;204: Connolly JE. Prevention of paraplegia secondary, to operations on the aorta. J Cardiovasc Surg 1986;27: Williams GM. Treatment of chronic expanding dissecting aneurysms of the descending thoracic and upper abdominal aorta by extended aortotomy, removal of the dissected intima, and closure. J VASC SURG 1993;18: Matas R. An operation for the radical cure ofaneurysms based on atteriorrhaphy. Ann Surg 1903;37: Cribari C, Meadors FA, Crawford ES, Coselli JS, Sail HJ, Svensson LG. Thoracoabdominal aneurysm associated with umbilical artery catheterization: case report and review of the literature. J VAsc SURG 1992;16: Submitted Nov. 8, 1993; accepted March 7, 1994.