Neurologic complications associated with endovascular repair of thoracic aortic pathology: Incidence and risk factors. A study from the European Collaborators on Stent/Graft Techniques for Aortic Aneurysm Repair (EUROSTAR) Registry

Methods 

The data on which the present report is based come from the EUROSTAR Thoracic Registry. This voluntary observational study was established in 2000. It is linked in terms of administration and protocol with the EUROSTAR Registry of Endovascular Repair of Abdominal Aortic Aneurysms. Aggregated data from the EUROSTAR and UK Thoracic Registry, which are compatible registries, were published in 2004.11

This analysis was based on 606 patients with aneurysm or dissection in the thoracic aorta who underwent endovascular repair between July 2000 and July 2006. Only patients who had been prospectively entered in the registry were considered. Ten patients with thoracic aortic abnormalities classified as penetrating ulcers without coexisting dissection or aneurysm were excluded from the assessment because this number was considered too small for meaningful conclusions.

Patients were recruited from 58 European institutions (Appendix 1, online only). Patients received the following commercially available devices approved for use in countries of the European Community (CE approved): Talent (Medtronic/AVE, Minneapolis, Minn) in 386, TAG-Excluder (W. L. Gore & Assoc, Flagstaff, Ariz) in 119, Zenith TX2 (William Cook Europe, Bjaeverskov, Denmark) in 39, Valiant (Medtronic/AVE, Minneapolis, Minn) in 28, Endofit (Endomed, Phoenix, Ariz) in 12, AneuRx (Medtronic/AVE, Minneapolis, Minn) in 4, Relay (Bolton Medical, Sunrise, Fla) in 2, and 16 other or unknown devices.

Emergency procedures were defined arbitrarily as those that require treatment ≤7 days of first presentation, except for dissections, in which the conventional period of 2 weeks for discrimination between acute or chronic condition was followed.

Collected baseline data included comorbidities, fitness for open surgery as classified according to the American Society of Anesthesiology (ASA), aneurysm anatomy, and operative details. Aneurysmatic lesions were categorized according to their primary anatomic location and extend in lesions within (1) the ascending thoracic aorta or arch, (2) proximal, (3) middle, (4) distal one-third of the descending thoracic aorta, and (5) lesions in three or more of these segments. Dissections were categorized similarly according to the site of the primary intimal tear.

Conduct of the thoracic endograft procedures, adjunct operations, and perioperative monitoring was at the discretion of the managing physicians. Covering of different levels of intercostal arteries was judged on intraoperative fluoroscopy combined with assessment of the preoperative computed tomography (CT) study.

Data were collected on case record forms before October 2004 and thereafter entered by the participants onto the database by using the EUROSTAR Web site. Outcome reporting adhered to the guidelines from the Ad Hoc Committee for Standardized Reporting Practices in Vascular Surgery of the Society for Vascular Surgery/American Association for Vascular Surgery.12 Groups that were compared consisted of patients with postoperative symptoms of spinal cord ischemia (SCI), those with stroke or neurologic event, and patients without neurologic event. Definitions and incidence rates of technical success, follow-up methods, and data on the entire study cohort and subgroups are presented in Appendix 2 (online only).13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24

Data were presented as means and ranges or standard deviations. Odds risk ratios (OR) and 95% confidence intervals (95% CI) were calculated to correlate patient characteristics and operative technical factors with the main outcome events in this study. Multivariate logistic regression was performed to assess the independent correlations of potential risk factors and outcomes were presented as P values, OR, and 95% CI.

Variables entered in the multivariate model were selected because of significant differences between subgroups at univariate comparison or because of clinical considerations. Death occurring ≤30 days of the initial procedure or during the hospital admission period was categorized as operative death and death occurring >30 days as late death (results presented in Appendix 2, online only).

Statistical significance of differences in proportions was determined with χ2 tests or Fisher tests and differences in continuous parameters with the Mann-Whitney test. The Kaplan-Meier method was used to determine cumulative survival. P < .05 was considered statistically significant. The data were analyzed with SAS 8.02 software (SAS Institute Inc, Cary, NC).

Results 

The study cohort consisted of 606 patients, of whom 471 (77.7%) were men. Mean patient age was 63.2 years (range, 13 to 91 years, Table I). Degenerative aneurysm was present in 291 patients, aortic dissection in 215, anastomotic false aneurysm in 24, and traumatic aortic injury in 67. Two patients with mycotic aneurysm were treated by thoracic stent grafts, and in seven, the diagnosis was not clearly recorded. Penetrating ulcer was diagnosed in 30 patients (5%), usually in association with aortic dissection, but was a degenerative aneurysmatic aorta in 14. The maximum diameter in patients with aneurysmatic lesions (excluding patients with dissection without aneurysmatic dilatation) was 62.7 ± 15.3 mm (range, 29 to 120 mm). Disorders were chronic in 379 patients and acute in 205; whether the disease was chronic or acute was not clear in 22 patients.

Table I. Patient characteristics and comorbidity

		With paraplegia (n = 15)	With stroke (n = 19)	Other patients (n = 574)	P⁎	OR (95% CI)
	Patient factors†
	Age, mean(range)	66.0(37-83)	71.7(59-82)‡	62.9(13-91)	.008‡	5.89(1.35-25.75)‡
	Male	14(93.3)	11(57.9)§	448(78.1)	.0388	0.39(0.15-0.98)
	ASA IV	6(40.0)§	5(26.3)	96(16.7)	.0187	3.32(1.15-9.54)
	ASA ≥III	10(66.7)	9(47.4)	268(46.7)
	Previous MI/AP/HF	5(33.3)	5(26.3)	143(24.9)
	Previous PTCA/CABG	1(6.7)	0(0.0)	55(9.6)
	Diabetes ≥1#	2(13.3)	1(5.3)	57(9.9)
	Hypertension ≥1#	12(80.0)	8(42.1)§	399(69.5)	.0113	0.32(0.13-0.81)
	Renal function impairment (SVS/ISCVS score ≥1)#	7(46.7)§	4(21.1)	101(17.6)	.0041	4.10(1.45-11.56)
	Cardiac disease ≥1#	6(40.0)	7(36.8)	201(35.0)
	Previous thoracic aortic procedures	0(0.0)	0(0.0)	64(11.2)
	Unfit for open thoracic repair	6(40.0)	3(15.8)	185(32.2)

OR, Odds ratio; CI, confidence interval; ASA, American Society of Anesthesiologists; MI, myocardial infarction; AP, angina pectoris; HF, heart failure; PTCA, percutaneous transluminal coronary angioplasty; CABG, coronary artery bypass graft; SVS/ISCVS, Society for Vascular Surgery/American Association for Vascular Surgery.

⁎ P values for three groups. In text, P values are for a study group vs control group. † Data are presented as number(%) unless indicated otherwise. ‡ Age dichotomized at ≥65 years was significantly higher in stroke patients. § Statistically significant from control group(other patients). # Indicates Society for Vascular Surgery/International Society for Cardiovascular Surgery class ≥1.

Stent grafts were a mean length of 12.4 ± 3.7 cm, and 1.8 ± 0.97 stent grafts were used per patient. Data on length of overlap zones between devices and overall covered aortic length were not available. In 78 patients (13%), there was involvement of the aortic arch or ascending thoracic aorta. The proximal one-third segment of the descending thoracic aorta was involved in 393 patients (65%), the middle one-third in 240 (40%), the distal one-third in 145 (24%), and the entire length of the descending thoracic aorta in 69 (11%).

During the first month after the thoracic stent graft procedure, neurologic complications were observed in 32 patients. Of these, 15 patients had evidence of paraplegia or paraparesis and constituted the group with postoperative SCI. Two patients, one with temporary sensory loss in the lower extremities and one with minimal symptoms described as “possibly caused by spinal cord infarction,” were not included in the SCI group because of the diagnosis was uncertain. Symptoms of cerebrovascular accidents were present after the stent graft procedure in 19 patients, and they constituted the group with postoperative stroke. A combination of SCI and stroke symptoms was observed in two patients, and they were included in both study groups. Excluded from the group with stroke were six patients with temporary cerebral symptoms, including three with obvious transient ischemic attacks and three with atypical symptoms. During the first month, 574 patients had no symptoms of paraplegia/paraparesis or stroke. General outcome events in groups with and without neurologic events, including technical success rates of procedures, morbidity, early and late mortality (Appendix 2 Fig, online only), and a comparison with previous publications13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24 is presented in Appendix 2 (online only).

Patient factors, procedural details, and correlation with neurologic complications 

Patients with postoperative stroke were significantly older, more frequently women, and less frequently known with preoperative hypertension. Paraplegic or paraparetic patients had more frequently a high ASA classification (class IV) and renal function impairment (Table I). In patients with SCI, degenerative aneurysm as the primary thoracic aortic pathology was more frequently observed than in other patients. No other significant differences were observed in the distribution of aortic disorders or the prevalence of acute or chronic presentation in the two groups with neurologic events compared with the other patients (Table II). Localization of the aortic pathology demonstrated a higher incidence of aortic arch or ascending thoracic aorta involvement in patients with stroke compared with patients without neurologic events: six patients (32%) vs 83 (14%), respectively (OR, 2.61; 95% CI, 1.02 to 6.70; P = .040).

Table II. Classification of thoracic aortic pathologies

	Pathology	With paraplegia (n = 15), No. (%)⁎	With stroke (n = 19), No. (%)⁎	Other patients (n = 574), No. (%)
	Degenerative aneurysm	11(73.3)†	11(57.9)	270(47.0)
	Chronic	8(72.7)	7(63.6)	209(77.4)
	Acute	3(27.3)	3(27.3)	48(17.8)
	Unknown	0(0)	1(9.1)	13(4.8)
	Dissection	3(20.0)	7(36.8)	206(35.9)
	Chronic	0(0)	3(42.9)	110(53.4)
	Acute	3(100)	4(57.1)	96(46.6)
	Traumatic rupture	1(6.7)	1(5.3)	65(11.3)
	Chronic	0(0)	1(100)	26(40.0)
	Acute	1(100)	0(0)	38(58.5)
	Unknown	0(0)	0(0)	1(1.5)
	False anastomotic aneurysm	0(0)	0(0)	24(4.2)
	Chronic	0(0)	0(0)	15(62.5)
	Acute	0(0)	0(0)	9(37.5)
	Infectious complications	0(0)	0(0)	2(0.4)
	Chronic	0(0)	0(0)	2(100)
	Acute	0(0)	0(0)	0(0)
	Unknown etiology	0(0)	0(0)	7(1.2)
	All etiologies‡
	Chronic	8(53.3)	11(57.9)	361(62.9)
	Acute	7(46.7)	7(36.8)	192(33.5)
	Unknown	0(0)	1(5.3)	21(3.7)

⁎ Percentage calculated in regard to the number of patients in each specific pathology group for each study group. † Statistically significant from control group(other patients): Odds ratio, 3.10; 95% confidence interval, 1.00 to 9.84; P = .0441. ‡ Total number of patients with chronic, acute or unknown etiology was 379, 205 and 22 respectively in the entire cohort(accounting for the two patients who had SCI and stroke combined).

No difference was observed between groups with and without neurologic events in type of anesthesia (regional or local anesthesia was used in 7.1% of patients in the entire cohort), induced hypotension or use of adenosine for cardiac cessation (43%), or use of iliac artery, abdominal aorta, or surgical implanted prostheses for access to the thoracic aorta (13%). The duration of the procedure was significantly longer in patients with stroke at 195.6 ± 152.6 vs 124.4 ± 78.6 minutes for patients with ≥160 minutes vs a shorter procedural time (OR, 3.67; 95% CI, 1.46 to 9.25; P = .003). From one to seven stent grafts were used, and significantly more stent grafts were used in patients with paraplegia or paraparesis, with three or more device components in 53% vs 19% in controls (OR, 4.94; 95% CI, 1.76 to 13.92; P = .0009).

The frequency of covering of intercostal arteries at the T9, T10, T11, and T12 levels was similar in patients with stroke and without neurologic events. However, intercostal arteries at the T10 level were more frequently occluded in patients with SCI compared with those without neurologic events (6 patients [40%] vs 102 [18%], respectively; P = .034, OR, 2.98; 95% CI, 1.04 to 8.55).

The correlation of additional procedures and the occurrence of postoperative neurologic deficits were assessed in detail. The left subclavian artery (LSA) was covered by the device in 159 patients (26%), of whom 40 were additionally treated with a LSA revascularization by transposition or bypass, and 119 of these patients did not undergo LSA revascularization (Table III). The incidence of LSA occlusion without revascularization was significantly higher in the study group with paraplegia/paraparesis than in other patients at 40% vs 19% (OR, 2.82; 95% CI, 1.00 to 8.08). The frequency of LSA covering was comparable in stroke patients and others. However, the rate of combined neurologic complications (paraplegia or stroke) in the group with LSA covering without revascularization was 8.4% (10 of 119), which compared unfavorably with the 0% rate in 40 patients with revascularized LSA covering (P = .049). Simultaneous open repair of the abdominal aorta was significantly more frequently performed in the SCI group (n = 3, 20%) compared with other patients (n = 18, 3.1%; OR, 7.96; 95% CI, 2.06 to 30.68; P = .0003).

Table III. Left subclavian artery covering and correlation with neurologic event

		Paraplegia (n = 15), No. (%)	Other (n = 591), No. (%)	P
	With LSA artery covering	6(40.0)	153(25.9)	.2199
	Without transposition/bypass	6(40.0)⁎	113(19.1)	.0444
	With transposition/bypass	0	40(6.8)	.2971
		Stroke(n=19), N(%)	Other(n = 587) N(%)
	With LSA artery covering	5(26.3)	154(26.2)	.9937
	Without transposition/bypass	5(26.3)⁎	114(19.4)	.4565
	With transposition/bypass	0(0)	40(6.8)	.2954

LSA, Left subclavian artery; NS, not significant.

⁎ One patient with left subclavian artery covering without revascularization had combined paraplegia and stroke.

Multivariate analysis of risk factors for spinal cord ischemia and stroke 

Multivariate regression analysis demonstrated an independent correlation of SCI and the variables of the number of stent grafts used, covering of the LSA without revascularization, renal function impairment, and simultaneous abdominal aortic repair (Table IV). Of the four variables with significant correlation in this model, two or more were present in 11 of the patients with SCI and three or more in eight.

Table IV. Multivariate regression analysis for paraplegia

	Risk factor	P value, paraplegia (n = 15)	OR (95% CI)
	Degenerative aneurysm	.1406	2.75(.72-10.6)
	Localization disease at proximal, middle, and descending thoracic aorta	.4866	.54(.10-3.03)
	Number of stent grafts ≥3	.0428	3.49(1.04-11.7)
	Left subclavian artery covering without transposition/bypass	.0274	3.94(1.17-13.3)
	Occlusion by device T10	.2474	2.07(.60-7.11)
	Renal failure(score ≥1)⁎	.0215	3.63(1.21-10.9)
	Concomitant open abdominal surgery	.0371	5.52(1.11-27.5)

OR, Odds ratio; CI, confidence interval; NS, not significant.

⁎ Society for Vascular Surgery/American Association for Vascular Surgery risk classification.

In patients with stroke, two factors had an independent significant correlation: the duration of the stent graft procedure and female sex (Table V). None of the other tested variables had an independent correlation with stroke.

Table V. Multivariate regression analysis for stroke

	Risk factor	P value, stroke (n = 19)	OR (95% CI)
	Degenerative aneurysm	.4626	.65(.20-2.04)
	Proximal into arch or ascending aorta	.7760	1.20(.34-4.26)
	Duration of procedure ≥2 h 40 min	.0134	5.25(1.41-19.6)
	Left subclavian artery covering without transposition/bypass	.4782	1.58(.44-5.65)
	Female sex	.0395	2.95(1.05-8.26)
	Maximum diameter aneurysm	.7924	1.00(.98-9.02)
	Age	.0937	1.05(.99-1.11)

OR, Odds ratio; CI, confidence interval; NS, not significant.

Discussion 

The incidence of paraplegia after open surgical repair varies from 2% to 21%, depending on the extent of the descending thoracic aorta replacement.25, 26, 27, 28, 29 In conventional surgery, much effort is directed to preserving intercostal arterial inflow by reattachment of these vessels, in particular those between the T8 and T12 level. A variety of diagnostic and therapeutic adjunctive methods have been used, including cerebrospinal fluid (CSF) drainage, to reduce the arterial-cerebrospinal fluid gradient.27, 30, 31, 32, 33 Despite the large choice in available options to prevent SCI, neurologic deficit remains a considerable threat in open repairs.29

The introduction of endovascular techniques as an alternative option for the treatment of thoracic aortic disease gave rise to some optimism in that the risk of paraplegia seemed reduced somewhere in the range of 0% to 6%.5, 7, 8, 10, 13, 14, 34 Avoidance of thoracic aortic clamping and prolonged episodes of hypotension most likely account for a lower incidence of spinal ischemic symptoms.

Only two studies have compared the incidence of SCI directly with a control group undergoing open surgery. The first study observed an insignificantly lower rate of perioperative paraplegia of 6.7% in the endovascular and 8.6% in the open surgical group.18 The second study demonstrated a statistically lower SCI incidence in patients treated by endografts (3% vs 14%).35

A number of studies have analyzed which factors increase the risk of SCI after endograft treatment.7, 8, 18, 36, 37 Simultaneous or previous open infrarenal aortic replacement has been recognized as a factor associated with a higher incidence of paraplegia.5, 8, 9, 36 Our multivariate analysis demonstrated this variable as an independent risk factor for SCI. The EUROSTAR registry only recorded data on concomitant infrarenal aortic replacement, and no conclusions could be drawn about whether previous abdominal aortic repair is associated with similar risks.

It has recently been documented that a compromised hypogastric artery inflow constituted a significant risk factor for SCI.38 Some authors suggest that patients with combined aneurysmatic disease of the thoracic and abdominal aorta be treated with staged procedures to better allow a gradual development of collateral spinal cord blood flow from lumbar and hypogastric arteries.8, 37

Extensive covering of the thoracic aorta by stent grafts also presents an increased risk for SCI, as was first documented by Greenberg et al7 and confirmed by several other studies.8, 20, 21 The present series demonstrated that the number of stent grafts used per patient correlated with development of paraplegia at univariate and multivariate analysis. The number of stent grafts per patient clearly corresponds with the covered aortic length; therefore, our observation seems in agreement with the previous observations on the importance of the covered aortic length.

Clinical studies of surgical thoracic and thoracoabdominal aneurysm repairs have documented a significant correlation of renal function and SCI injury.26 The underlying metabolic mechanism is not exactly known; however, renal failure may reflect more advanced peripheral atherosclerosis, including intercostal arteries and collaterals to the spinal cord. The present analysis also indicates renal failure (Society for Vascular Surgery/International Society for Cardiovascular Surgery class ≥1) as an independent risk factor for SCI. Other previously observed risk factors for paraplegia after open thoracic aorta surgery included symptomatic disease or rupture of the aorta and perioperative hypotensive episodes. Perioperative blood pressure data were not available in the present series; however, Chiesa et al37 reported in 2004 that a lowest mean arterial blood pressure of ≤70 mm Hg presented a significant predictor of SCI.37

Although many of the risk factors discussed here are difficult to modify, some precautions can be taken. Caution is warranted with the continuation of antihypertensive medication, which is used by many patients with thoracic aortic disease. Postoperatively, these medications may lead to episodes of hypotension that sometimes is difficult to treat. In practice, many physicians involved in the management of thoracic aortic disease now frequently (particularly in patients presenting with risk factors for SCI) reduce or withhold any antihypertensive medication before endovascular treatment, with the exception of β-blockers.

The role of CSF drainage is less well documented in endovascular than in open repair of thoracic aorta disease. Many experts, however, advise the selective use of prophylactic CSF drainage in cases requiring coverage of long segments of thoracic aorta, in particular when the territory distally of T10 is involved.8, 18, 37 These patients may be more vulnerable to postoperative hypotension and so-called delayed paraplegia. Our data set had no information on whether paraplegia presented immediately or delayed after the procedure and whether CSF drainage was used.

The role of unimpeded flow into the left subclavian and its first branch, the vertebral artery as supplier of the anterior spinal artery, appeared of considerable importance, because occlusion of the LSA was associated with an almost fourfold increased incidence of paraplegia. Another less well recognized collateral source of spinal cord blood supply is the internal mammary artery and its anterior intercostal branches. This pathway becomes also compromised by LSA overlapping. When thoracic aortic disease is localized near the LSA, the surgeon has to decide whether the origin of this vessel can be covered by the stent graft fabric or will need a revascularization by subclavian-carotid transposition or bypass.39, 40, 41, 42, 43

Most authors seem to agree that the LSA can be overstented in most cases, and suggest that preoperative computed tomography (CT), magnetic resonance imaging (MRI), or angiography may be helpful in selecting candidates for adjunct LSA revascularization.40, 41, 43 These assessments should provide information about whether the left vertebral artery is a dominant source of collateral flow to the anterior spinal and basilar arteries and identify any acquired disease or congenital anomalies of the supra-aortic, vertebral, and internal mammary vessels that preclude safe occlusion of the LSA.

Few data, however, are available on the association of occlusion of the LSA and the rate of paraplegia. In a 2006 publication, Peterson et al44 reported a review of the literature on the outcome of subclavian artery covering during thoracic stent grafting in 218 patients, of whom 114 were treated by a subclavian artery transposition. The mean incidence of paraplegia in patients with covered LSA in this review was 1%, which was lower than the 5% in our study. It must be noted that in patients with LSA overlap, this vessel was revascularized in 52% in the compilation of series from the literature by Peterson et al compared with only 25% in the current EUROSTAR series. This may reflect that many surgeons considered the covering of the LSA origin as an innocuous detail.

Most of the patients with SCI had several risk factors for a compromised spinal cord perfusion, and loss of collateral flow from the subclavian vertebral system contributed to the event. The best treatment in the patient with a short proximal landing zone to the LSA should be selected on the basis of high-quality imaging of the supra-aortic vessels to ascertain a functional collateral connection between the right-sided vertebral and mammary arteries and the spinal cord circulation. In our experience, many preoperative CT or MRI studies do not provide this kind of detailed information.

It should be recognized that LSA transposition is a low-risk procedure. Simply ignoring the LSA take off and occluding it by the stent graft seems no longer appropriate in most patients considering the data reported here. Options to maintain the perfusion of supra-aortic branches with fenestrated devices or retrograde stent deployment through the device are presently being developed.45, 46

The incidence of stroke, similar to SCI, varies widely. In a review of the literature reported in 2006, an intracranial stroke developed in average of 2.2% of patients undergoing thoracic aortic endograft treatment,10 which is in agreement with the 3% observed in the present study. The location of the cerebral infarctions after thoracic endografting suggests multiple emboli to the anterior or posterior circulation. These emboli are most likely caused by catheter, guidewire, or endoprosthesis manipulation in a diseased arch. Other potential etiologic factors for stroke include air embolism, cervical carotid, and vertebral-basilar atherosclerotic disease.

A significant risk factor for stroke in our study was a duration of the procedure ≥2.6 hours. A longer endovascular procedure will inevitably be associated with more manipulation of catheters, guidewires, and introducer systems. The second factor correlating with an increased stroke risk was female sex. Evidence is growing that female sex has an adverse influence on treatment outcomes of aortic aneurysms.47 The higher complication rate in women has been attributed to more advanced atherosclerosis as well as to smaller diameter peripheral arteries.

Precautions to prevent cerebral embolism include careful preinterventional planning to reduce procedural time. A poor quality arch with mural thrombus can be identified best by transesophageal echography. Five of 19 stroke patients (26%) in this EUROSTAR series had LSA occlusion without revascularization, which did not present a significant correlation. In other studies,19 however, this variable correlated statistically with the risk of stroke; therefore, similar vascular anatomic criteria to perform a LSA transposition to prevent SCI may also apply in perioperative stroke prevention.

Conclusions 

Severe neurologic complications are not rare after endovascular thoracic aortic procedures. The combined incidence of stroke and paraplegia was 5% in the current series. Risk factors for both complications were identified, and strategies for modification of these variables were discussed. The perhaps the most important precaution that emerged from this study involved the revascularization of the LSA in patients who need incorporation of the origin of this vessel in the proximal landing zone of the stent graft. In this series, neurologic complications were only observed in patients in whom prophylactic rerouting of the blood flow had not been performed, and 8.4% of this category had a severe neurologic event (stroke or paraplegia) compared with 0% in patients with revascularization of the LSA. To properly select patients for subclavian revascularization considerable expertise with imaging of this vascular territory is required. If after careful CT, MRI, or angiography, or a combination of these, there still is doubt about the collateral supply of the spinal cord, a left subclavian–carotid transposition may be the safest option.