Although the overall mortality rate after surgical repair

Ascending Aorta Replacement and Local Repair of Tear Site in Type A Aortic Dissection With Arch Tear Kay-Hyun Park, MD, Kiick Sung, MD, Kwhanmien Kim, MD, Tae-Gook Jun, MD, Young Tak Lee, MD, and Pyo Won Park, MD Department of Thoracic and Cardiovascular Surgery, Samsung Medical Center, Sungkyunkwan University School of Medicine, Seoul, Korea Background. Transaortic repair of the tear was previously proposed as a compromise alternative to total arch replacement for acute type A aortic dissection with an arch tear. However, there are no data about long-term prognosis and radiologic findings after this procedure. Methods. We reviewed the postoperative course and computed tomographic findings of 13 patients who underwent replacement of the ascending aorta and transaortic repair of an arch tear for acute type A aortic dissection. Results. There were no early or late deaths during a mean follow-up period of 36.8 months. Computed tomographic follow-up showed complete thrombosis with or without later regression of the false lumen in the descending thoracic aorta in 9 patients (69.2%). Thrombosis of the false lumen usually occurred within 3 months postoperatively. Repair techniques incorporating the full thickness of the aortic wall in closure of the tear resulted in higher rate of success than approximation of the intima only (7 of 8 versus 2 of 5 patients, p < 0.05). Conclusions. Transaortic repair of the arch tear with replacement of the ascending aorta can be an option in selected patients who have a small intimal tear in the aortic arch. This option would be more viable for less experienced surgeons who would hesitate to replace the total arch. (Ann Thorac Surg 2003;75:1785 91) 2003 by The Society of Thoracic Surgeons Accepted for publication Jan 16, 2003. Address reprint requests to Dr Kay-Hyun Park, Department of Thoracic and Cardiovascular Surgery, Samsung Medical Center, #50 Ilwon-Dong, Kangnam-Ku, Seoul, Korea 135-710; e-mail: drkhpark@yahoo.co.kr. Although the overall mortality rate after surgical repair of acute aortic dissection is still higher than for other types of cardiac operations, the outcome has been improving in experienced centers [1]. Many authors attributed this advance to open distal anastomosis under hypothermic circulatory arrest which avoids aortic crossclamping, a patient-specific approach to aortic root pathology, resection of the aortic segment containing the intimal tear, and vigilance in preventing intraoperative malperfusion [1 3]. On the basis of these considerations, a more aggressive approach to acute aortic dissection has been gaining popularity. However, there remains a controversy with regard to the optimal strategy in cases with retrograde dissection from an entrance tear in the transverse arch or proximal descending aorta. Although replacement of the ascending aorta and entire arch is advocated in many of recent publications, simple replacement of only the ascending aorta can also be justified in some instances, especially for surgeons with less experience, because arch reconstruction carries a high risk of bleeding and prolonged circulatory arrest [2]. Previously, von Segesser and associates [4] and Baumgartner and colleagues [5] reported a technique of local closure of the entrance tear in the aortic arch by aortotomy and replacement of the ascending aorta. Although their reports were based on a small number of patients and lacked long-term results, their satisfactory early outcome inspired us. We applied this compromise option in selected patients who had suitable tear morphology. During the follow-up of these patients, we found that the clinical course and computed tomographic (CT) findings were better than expected. Therefore, we reviewed our experience with transaortic closure of arch tear in patients with acute type A dissection, focusing on the late clinical outcome and follow-up CT findings, which have not been mentioned in previous reports. Patients and Methods Between November 1995 and June 2002, a total of 103 patients underwent surgical repair of Stanford type A acute aortic dissection at our institution (Fig 1). In 19 patients of this group, intimal tear was not found in the ascending aorta or the arch by visual inspection during the operation; 9 patients had a diagnosis of intramural hematoma and the other 10 had a diagnosis of retrograde propagation of distal descending aortic dissection. Fortyfour patients had an entrance tear in the ascending aorta, and 40 patients had an entrance tear in the aortic arch. Twelve patients had the tear on the lesser-curvature side of the arch and underwent ascending aorta and hemiarch replacement after resection of the tear site. In 28 patients, 2003 by The Society of Thoracic Surgeons 0003-4975/03/$30.00 Published by Elsevier Inc PII S0003-4975(03)00170-X

1786 PARK ET AL Ann Thorac Surg LOCAL ARCH REPAIR IN AORTIC DISSECTION 2003;75:1785 91 Fig 1. Location of intimal tear and surgical procedures in 103 patients with acute type A aortic dissection. the intimal tear was located on the greater-curvature side of the arch such that excision of the tear site would necessitate total arch replacement with reattachment of brachiocephalic branches. Among these, 13 patients had total arch replacement and another 13 patients had replacement of the ascending aorta with the entry tear closed through the aortotomy opening. We reviewed the hospital records and serial CT scan findings of these 13 patients who had transaortic closure of the tear in the greater-curvature side of the aortic arch with replacement of the ascending aorta. Patient Profile There were 10 male and 3 female patients, and the mean of age of the patients was 54.6 years (range, 34 to 68 years). One patient had Marfan syndrome. All patients were operated on within 2 weeks of symptom onset. Five patients had bloody pericardial effusion, and 3 manifested cardiogenic shock resulting from pericardial tamponade. In 1 of those patients cardiac arrest developed on entrance into the operating room, and cardiopulmonary resuscitation was continued until the start of cardiopulmonary bypass. Diagnosis of aortic dissection was made by CT scan in 11 patients, whereas only transthoracic echocardiography was possible in 2 patients. Presence of an intimal tear in the aortic arch was suspected by preoperative CT scan findings in only 2 patients. In all patients, the dissection was classified as DeBakey type I, as dissecting flap and false lumen in the descending aorta were visualized by CT scan or intraoperative transesophageal echocardiography. The locations of the intimal tear in the arch found by Table 1. Methods and Computed Tomographic Results of Arch Tear Repair Patient No. Sex/Age (y) Location of Intimal Tear Method of Repair Finding of Descending Thoracic Aorta in the Last Follow-up CT Scan (interval from operation [mo]) 1 M/62 Between RBCA & LCCA Intimal approximation with mattress 2 M/34 Anterior to LCCA & LSCA Intimal approximation with mattress 3 M/59 Between LCCA & LSCA Intimal approximation with mattress 4 M/49 Anterior to LCCA & LSCA Intimal approximation with continuous suture 5 F/66 Just distal to LSCA Intimal approximation with mattress 6 M/40 Between RBCA & LCCA Full-layer transfixion with mattress 7 M/56 Just distal to LSCA Full-layer transfixion with Dacron graft patch covering the intimal defect 8 M/63 Just distal to LSCA Full-layer transfixion with mattress 9 M/60 Between RBCA & LSCA b Full-layer transfixion with mattress 10 M/47 Anterior to LSCA Full-layer transfixion with mattress 11 F/68 Just distal to LSCA Full-layer transfixion with pericardial patch covering the intimal defect 12 F/56 Distal to LSCA Full-layer transfixion with mattress 13 M/53 Just distal to LSCA Full-layer transfixion with mattress Disappearance of false lumen (37) Disappearance of false lumen (9) Patent false lumen (40) Partial thrombosis of false lumen (22) Patent false lumen (9) Complete thrombosis of false lumen (3) a Disappearance of false lumen (24) Complete thrombosis of false lumen (41) Patent false lumen (4) Disappearance of false lumen (7) Disappearance of false lumen (22) Disappearance of false lumen (4) Complete thrombosis of false lumen (4) a This patient underwent second aortic surgery after this CT. b The patient had another multiple small tears in the descending aorta. CT computed tomography; LCCA left common carotid artery; LSCA left subclavian artery; RBCA right brachiocephalic artery.

Ann Thorac Surg PARK ET AL 2003;75:1785 91 LOCAL ARCH REPAIR IN AORTIC DISSECTION 1787 visual inspection after aortotomy are given in Table 1. In 2 patients, a larger primary tear was found in the ascending aorta, and 1 of those patients had other multiple small tears in the proximal descending aorta. The morphology of the tear in the arch was linear in 10 patients, whereas 3 patients had an ulcer-like defect in the middle of a severely atherosclerotic intima. Indications for Transaortic Tear Closure In our initial experience, choosing the treatment method for a tear in the greater-curvature side of the arch was largely determined by the operating surgeons judgment of the risk of total arch replacement. Because felt reinforcement and anastomosis of the distal stump and arch branches were considered the most time-consuming and prone to bleeding, closure of the tear was favored when the following factors were present: (a) poor visualization of the distal arch through median sternotomy, (b) extremely thin aortic wall in the distal arch, and (c) severe arteriosclerosis in the distal arch with muddy atheroma or calcification. In two patients, the decision was based on poor preoperative condition, namely, preoperative shock and cardiac arrest, in order to simplify and shorten the operation. However, as we gained experience with aortic dissection and arch surgery during the study period, the decision was made more by the feasibility of expeditious (within 10 minutes) and tension-free closure of the tear. It was considered feasible and safe if the tear was linear, not longer than half the aortic circumference (2 to 3 cm), and if the margins could be approximated easily without tension. A less-than-1-cm large ulcer-like defect of intima in the atherosclerotic bed was also considered feasible for closure with a patch. The narrowing of indication for transaortic tear closure is reflected in the change in proportion of patients who had total arch replacement. Whereas only 3 of the first 14 patients who had a tear in the greater-curvature side of the arch had total arch replacement, 10 of the later 14 patients had total arch replacement. Operative Procedures In all patients the chest was opened by median sternotomy. Arterial cannulation for cardiopulmonary bypass was done into the femoral artery in 7 patients and into the right subclavian artery through an interposing 8-mm Dacron graft in the last 6 patients. Venous drainage was established by direct cannulation of both vena cavas separately. After systemic cooling to achieve rectal temperature below 18 C, total circulatory arrest was induced to perform the aortic procedures. After resection of the ascending aorta, the intimal tear in the arch or proximal descending aorta was repaired with the techniques described below. The proximal and distal stumps of the aorta were reinforced by securing strips of Teflon felt to the inside and outside of the lumen with 4-0 polypropylene continuous mattress. During distal anastomosis, air and small debris were flushed out of the aorta and brachiocephalic branches by retrograde cerebral perfusion through the superior vena caval cannula for 10 to 15 minutes. Duration of circulatory arrest including the time for retrograde cerebral perfusion was 54.3 15.8 minutes on average (range, 33 to 84 minutes). Duration of circulatory arrest decreased as we accumulated experience; the mean duration was 65.9 minutes in the first 7 patients, whereas it was 42.7 minutes in the last 6 patients. Proximal anastomosis was performed during the rewarming period. Along with the aortic procedures, resuspension of detached aortic valve commissure(s) was needed in 3 patients, coronary artery bypass grafting in 1, and Bentall procedure in 1 patient. Techniques for Local Repair of Arch Tear To close the dissection entry in the first 5 patients, only the aortic intima was approximated with 5-0 polypropylene in an interrupted horizontal mattress manner in 4 patients and in a continuous manner in 1 patient. However, speculating that inclusion of the whole layer of the aorta would be advantageous for secure closure, we changed the technique to whole-layer transfixion suture methods (Fig 2). After careful dissection to expose the outer surface of the aorta corresponding to the tear site, 4-0 or 5-0 polypropylene with a Teflon pledget buttress were passed through the aortic wall. For a linear tear, three or four were placed along the length of the tear (Fig 2A). If a tear was located between two closely located ostia of the brachiocephalic branches, the tear site was sandwiched with the wall of those branches (Fig 2B). If the tear had an irregular margin or ulcer-like feature, a small patch was used to cover the intimal defect (Fig 2C). After completion of the local repair, fibrin glue was applied to the outer wall of the repair site. No glue was injected into the false lumen. Follow-Up In addition to echocardiographic examination, which was performed in all patients, early follow-up CT scan before discharge was done in 6 patients. After discharge, all patients were followed up by the operating surgeons through the outpatient service at 3- to 6-month intervals. As long as the patient agreed, follow-up CT scan was done once a year, and all patients had at least one follow-up CT examinations. Results Postoperative Course There were no in-hospital deaths or significant complications, such as permanent or transient neurologic deficit, renal failure, or reoperation due to bleeding. All patients recovered consciousness within 24 hours postoperatively, and 3 patients experienced delirium which resolved after transfer to the general ward. Two patients required mechanical ventilatory care for more than 48 hours. The mean duration of postoperative hospital stay was 17.2 6.9 days (range, 8 to 31 days). There were no late deaths during the follow-up period, of which the mean duration was 36.8 19.9 months

1788 PARK ET AL Ann Thorac Surg LOCAL ARCH REPAIR IN AORTIC DISSECTION 2003;75:1785 91 preoperative CT scan showed severely compressed true lumen at the distal thoracic and supraceliac abdominal aorta. Although the follow-up CT 3 months after the initial operation showed complete thrombosis of the false lumen in the descending aorta which had normal diameter, the true lumen remained severely narrowed because of compression by the false lumen thrombus. Computed Tomographic Findings Early postoperative CT was done in 6 patients before discharge from the hospital. Three of the 6 patients had complete thrombosis of the entire false lumen that was present down to the common iliac arteries. The false lumens of the other 3 patients were partially thrombosed, ie, the entire or proximal two thirds of the thoracic aorta was completely thrombosed and the remaining distal aorta had patent false lumen. The first follow-up CT scan was done in the outpatient clinic 3 to 15 months postoperatively, and 4 patients had later CT examinations, with the final study done 24 to 41 months postoperatively. Complete thrombosis or resolution of the false lumen hematoma, resulting in normallooking aorta, occurred in the descending thoracic aorta of 9 patients (69.2%), and closure of the intimal tear was considered successful in these patients (Table 1). In 6 patients of this group, the entire descending aorta had complete thrombosis or resolution of the false lumen. The other 3 patients had patent false lumen in the Fig 2. Techniques of local closure of an intimal tear in the aortic arch. Usually, after passing the whole layer of the aortic wall, repairing could be tied from outside of the aorta (A). If the tear is located between two closely oriented branches, the were tied from inside the aorta (B). In cases of an irregularly shaped tear or intimal ulcer, a small patch was used to cover the defect (C). (range, 4 to 65 months). The accumulated follow-up period was 479 patient-months. One patient had a second aortic operation: the descending thoracic aorta was replaced because of lower limb claudication caused by so-called pseudocoarctation. This patient had pulse deficit in the lower extremities before the first operation, and Fig 3. Summary of status of descending thoracic aorta in serial computed tomographic scans postoperatively. In patients 3, 4, 5, and 9, local repair of the arch tear failed to obliterate the false lumen. In the remaining 9 patients, thrombosis of the false lumen occurred early, ie, 3 to 4 months postoperatively, followed by absorption of the hematoma and resolution of dissection.

Ann Thorac Surg PARK ET AL 2003;75:1785 91 LOCAL ARCH REPAIR IN AORTIC DISSECTION 1789 abdominal aorta below the origin of the celiac or renal arteries, suggesting the presence of reentry tears in that location. These changes occurred within 3 to 4 months postoperatively (Fig 3). Conversely, closure of the intimal tear was considered unsuccessful in 4 patients because the false lumen was enhanced by contrast agent throughout the descending aorta. A later study in 1 patient in this group showed false lumen thrombosis in the proximal one third of the descending thoracic aorta. However, because this patient rejected further CT examination, it was not possible to determine whether thrombosis progressed distally. One patient in the failure group (patient 3 in Table 1) and another in the success group (patient 7) had enlargement of the descending thoracic aorta requiring close follow-up with annual CT examinations. There seemed to be a difference in the rate of successful closure of the intimal tear according to the technique of repair. Among the 5 patients in whom only the intimal tissue was approximated, 3 had patent false lumen. Conversely, the false lumen remained patent in only 1 of 8 patients in whom the whole layer of the aortic wall was incorporated to close the tear. Because this patient had additional small tears in the descending aorta, it was not definite whether the closure failed. Comment In 20% to 30% of patients with acute type A aortic dissection, the intimal tear is not found in the ascending aorta [2, 6 8]. For patients whose tear site is not resectable by replacement of the ascending aorta or hemiarch, ie, in the greater-curvature side of the aortic arch, there is a controversy regarding the optimal strategy. Several authors advocate total arch replacement, with the rationale that resection of the entry site can decrease the incidence of false lumen patency and subsequent longterm complications [7, 9 11]. It is also believed that early postoperative complications, such as anastomotic bleeding, arch rupture, and malperfusion, can be decreased by resection of the arch tear. Conversely, others claim that limited operation, ie, replacement of only the ascending aorta can also be justified because the principal object of emergency operation for acute dissection is the immediate survival of the patient. Their argument is that total arch replacement carries a higher risk of bleeding and prolonged cerebral ischemia, which outweighs the longterm benefit if done by surgeons with inadequate experience with aortic dissection [2, 12]. Transaortic closure of the tear site and replacement of the ascending aorta is supposed to be a compromise between the preceding two options [4, 5]. Although it has been our belief that total arch replacement is the standard treatment for those patients, we adopted this compromise option in selected patients who had suitable tear morphology. As a group of surgeons who had had little previous experience with aortic dissection, we think that this strategy helped us to keep the results of our aortic dissections in the acceptable range. During the study period, the overall early mortality rate after 103 aortic dissection repairs was 8.7%; it was 9.1% (7 of 77) for replacement of the ascending aorta with or without the hemiarch and 15.4% (2 of 13) for replacement of the entire arch. Before we undertook this study, we speculated that our strategy could provide better long-term results than replacement of the ascending aorta only. We thought that the incidence of false lumen patency and subsequent complications would be lower if the entry is successfully obliterated. However, this speculation could not be substantiated because previous reports of this technique did not mention the long-term clinical course or radiologic follow-up findings. We believe that this study supports our initial speculation because we observed that the false lumen in the descending thoracic aorta was completely thrombosed or regressed in two thirds of the patients. The success rate was higher if the intimal tear was repaired with secure incorporating the full thickness of the aortic wall. Serial CT findings showed that this benign change usually occurred within 6 months. In some patients, the false lumen was completely thrombosed already at the time of discharge from the hospital. There are several arguments against our speculation. First, it can be argued that the false lumen could have been thrombosed by the natural healing process. However, we think that such changes could not have occurred if the arch tear was left unrepaired because previous studies reported the false lumen patency rate as high as 80% after surgery for ascending dissection [13 15]. It is also debatable whether our approach is really simpler than total arch replacement because the mean duration of circulatory arrest approached 60 minutes. In the early part of our experience, circulatory arrest was frequently prolonged because of our limited experience and our hesitation about total arch replacement. Recently, however, the duration of circulatory arrest does not exceed 45 minutes, which is much shorter than that for arch replacement, which requires about 60 minutes of circulatory arrest on average. In addition to our accumulation of experience, we believe that several modifications of our technique contributed to the shortening of circulatory arrest. First, we found that exposure of the outer wall of the aortic arch before circulatory arrest was helpful not only in local repair cases but also in the cases requiring total arch replacement. Currently, we perform meticulous dissection to expose the superior and isthmic portion of the arch and proximal portion of brachiocephalic branches during the cooling period in every patient with acute type A dissection. Second, because we try to replace as much of the ascending aorta as possible in all type A dissection cases, most of our patients undergo aortic replacement more like hemiarch replacement than ascending aortic replacement. With the aortotomy reaching the origin of the innominate artery and the lessercurvature side of the arch, the lumen of the distal arch is well visualized and the repair is facilitated.

1790 PARK ET AL Ann Thorac Surg LOCAL ARCH REPAIR IN AORTIC DISSECTION 2003;75:1785 91 The preceding arguments would have been more properly answered by a study that compares the outcome between simple ascending aortic replacement, total arch replacement, and the procedure we described. However, such a comparative study would require more patients and longer follow-up. Although we report short-term results that are better than expected, we did not intend nor do we conclude that transaortic closure of arch tear is generally preferable to total arch replacement. We think it should be limited to patients with a short linear tear or small ulcer-like defect in the intima. Even in patients with a linear tear, secure repair cannot be anticipated if there is a large gap between two edges of the intima. In addition, if the transverse arch has aneurysmal change, the high wall tension can make the repair site vulnerable to repeated tear. Finally, we regard a tear coursing circumferentially around the ostium of one of the brachiocephalic branches as a contraindication to local repair, for fear of luminal compromise. Total or partial arch replacement with separate grafting of brachiocephalic branches is preferred in that situation. In conclusion, transaortic repair of the intimal tear with replacement of the ascending aorta resulted in satisfactory early outcomes in selected patients with acute type A dissection who had a small tear in the greater-curvature side of the aortic arch. Results of serial CT scans showed that thrombosis and even regression of the false lumen in the descending aorta progressed early postoperatively, especially if the repair was done with incorporating the full thickness of the aortic wall. Based on our experience, we think that it can be a viable option in properly selected patients. It could be more viable for surgeons who otherwise would leave the arch untouched to avoid the risk of total arch replacement. References 1. Westaby S, Saito S, Katsumata T. Acute type A dissection: conservative methods provide consistently low mortality. Ann Thorac Surg 2002;73:707 13. 2. Ehrlich MP, Ergin MA, McCullough JN, et al. Results of immediate surgical treatment of all acute type A dissections. Circulation 2000;102(Suppl III):III-248 52. 3. David TE, Armstrong S, Ivanov J, Barnard S. Surgery for acute type A aortic dissection. Ann Thorac Surg 1999;67: 1999 2001. 4. Von Segesser LK, Killer I, Ziswiler M, et al. Dissection of the descending thoracic aorta extending into the ascending aorta a therapeutic challenge. J Thorac Cardiovasc Surg 1994;108:755 61. 5. Baumgartner FJ, Omari BO, Pandya A, Pandya A, Bethencourt DM. Local transverse arch repair for type A aortic dissection. Ann Thorac Surg 1997;64:1331 2. 6. Bachet JE, Termingnon JL, Dreyfus G, et al. Aortic dissection. Prevalence, cause, and results of late reoperations. J Thorac Cardiovasc Surg 1994;108:199 206. 7. Okita Y, Takamoto S, Ando M, et al. Surgery for aortic dissection with intimal tear in the transverse aortic arch. Eur J Cardiothorac Surg 1996;10:784 90. 8. Carrel T, Pasic M, Vogt P, et al. Retrograde ascending aortic dissection: a diagnostic and therapeutic challenge. Eur J Cardiothorac Surg 1993;7:146 52. 9. Ando M, Nakajima N, Adachi S, Nakaya M, Kawashima Y. Simultaneous graft replacement of the ascending aorta and total aortic arch for type A aortic dissection. Ann Thoracic Surg 1994;57:669 76. 10. Bernard Y, Zimmermann H, Chocron S, et al. False lumen patency as a predictor of late outcome in aortic dissection. Am J Cardiol 2001;87:1378 82. 11. Pugliese P, Pessotto R, Santini F, Montalbano G, Luciani GB, Mazzucco A, Risk of late reoperations in patients with acute type A aortic dissection: impact of a more radial surgical approach. Eur J Cardiothorac Surg 1998:576 81. 12. Miller DC. Concomitant arch repair in acute type A aortic dissection [invited letter]. J Thorac Cardiovasc Surg 1992;104: 206. 13. Yamaguchi T, Guthaner DF, Wexler L. Natural history of the false channel of type A aortic dissection after surgical repair: CT study. Radiology 1989;170:743 7. 14. Moore NR, Parry AJ, Trottman-Dickenson B, Pillai R, Westaby S. Fate of the native aorta after repair of acute type A dissection: a magnetic resonance imaging study. Heart 1996;75:62 6. 15. Mathiu D, Keita K, Loisance D, Cachera JP, Rousseau M, Vasile N. Postoperative CT follow-up of aortic dissection. J Comput Assist Tomogr 1986;10:216 8. INVITED COMMENTARY Ours is a field in which aggressiveness is applauded and bold action rewarded. Accordingly, one is more likely to see articles in our literature advocating radical approaches to problems than conservative ones. This manuscript is an exception to that rule. Dr Park and his associates have revisited a reparative technique generally thought inadequate because of perceived poor durability. In my training I was taught to resect and replace the region of the tear in cases of acute dissection. This is easily said and generally not too complex to perform if the tear is on the underside of the arch, or even in the front or back wall. But what about tears between the brachiocephalic vessels or on the far side of the subclavian artery? Total arch replacement would likely be the safe answer for the board examinations, and there is certainly a growing literature supporting the safety of this approach. But as Dr Park has implied, these reports come from large centers with enormous experience in arch surgery. What about the surgeon in a smaller center who sees perhaps only a handful of dissections in a year? There is no doubt that total arch replacement will be a more durable repair than local repair. If the patient survives hospitalization, the likelihood of reoperation for arch disease will be less if the arch is made of dacron. However, if we accept Dr Park s data, the notion that the late results of local repair are poor appears to be in error. Although the number of patients undergoing local repair was small, the authors have provided just the data we require serial imaging studies demonstrating satisfac- 2003 by The Society of Thoracic Surgeons 0003-4975/03/$30.00 Published by Elsevier Inc PII S0003-4975(03)00357-6