Spinal Cord Ischemia: Clinical and Imaging Patterns, Pathogenesis, and Outcomes in 27 Patients

Background  The natural history and pathogenesis of ischemic spinal cord infarction remain largely unknown because most clinical studies have included mostly patients with ischemic lesions associated with aortic surgery or prolonged arterial hypotension.

Objective  To assess the pathogenetic mechanisms and outcomes of these cord infarctions based on clinical findings and spinal vascular anatomy.

Design  Retrospective review.

Patients  We analyzed the clinical, laboratory, imaging, and outcome data for 27 patients with acute spinal cord infarction admitted between 1990 and 2003. There were 11 men and 16 women (age range, 19-80 years [mean age, 56 years]).

Results  Ten patients had anterior spinal artery patterns, 4 each had anterior and posterior unilateral patterns, 3 had central patterns, and 2 each had posterior spinal artery patterns, transverse syndromes, and unclassifiable clinical pictures. Twenty patients had no identifiable etiology. Patients with a central or transverse infarct showed a high frequency of peripheral vascular disease, and all transverse infarcts occurred following prolonged arterial hypotension. The onset of all other infarcts was associated with mechanical triggering movements (P = .02), and these patients frequently had diseases of the spine (P = .003) at the level of the spinal lesion, with the clinical data suggesting root involvement at the level of the spinal cord lesion and pointing to mechanical injury of a radicular artery. The outcomes were favorable, with only 13 patients showing significant gait impairment on leaving the hospital.

Conclusions  There are 2 main types of spinal cord ischemia: (1) radicular artery territory infarct (bilateral anterior or posterior spinal artery infarcts and unilateral infarcts) and (2) extensive spinal cord hypoperfusion (central and transverse infarcts). Each type has characteristic clinical, imaging, pathogenetic, and prognostic features.

Spinal cord infarction is much less frequent than cerebral infarction, accounting for only 1% of all strokes.¹ Because of this, few large clinical investigations have been carried out, and most have dealt with spinal cord ischemia due to aortic surgery; therefore, the pathogenesis and natural history of spontaneous or nonsurgical spinal cord infarctions remain largely unknown. The objective of this study was to assess the pathogenetic mechanisms and outcomes of these cord infarctions based on clinical findings and spinal vascular anatomy.

The spinal cord is supplied by 1 anterior and 2 posterior spinal arteries, which extend longitudinally in a variable fashion. Rostrally, these arteries originate from the V4 region of the vertebral arteries and anastomose at the level of the cone.² At many levels, they receive supply from the radicular arteries, which enter the canal along with the nerve roots. The anterior spinal artery gives rise to the central arteries, each of which enters the spinal cord to supply the anterior horn and the anterior part of the lateral column on the left or right side at each level. The spinal arteries are connected by a pial plexus that surrounds the spinal cord, and the posterior spinal arteries may be linked together^3,4 (Figure 1). Each radicular artery supplies a separate functional region of the spinal arteries, particularly the anterior spinal artery.² The first region extends from C1 until T3 and is supplied at the C3 level from the vertebral arteries⁵ and at the level from C6 until C7 from the cervical ascending arteries.^3,5 The second region extends from T3 until T7 and sometimes receives a branch from the intercostal artery at the T7 level.¹ The third region extends from T8 to the cone and receives a branch (Adamkiewicz artery) from the intercostal artery, most frequently between T9 and T12.⁶ There is sometimes a cone artery originating from the internal iliac artery (Desproges-Gotteron artery) at the L2⁷ or L5⁸ level (Figure 2).

We studied 72 patients admitted for investigation of acute myelopathy between 1990 and 2003. The patients underwent complete etiologic investigations.

The laboratory workup consisted of complete blood cell count; erythrocyte sedimentation rate; electrolyte, glucose, creatinine, and cholesterol levels; liver function test; protein electrophoresis; coagulation tests; electrocardiogram; and thoracic radiography. The infection workup consisted of screening serologically for syphilis (VDRL test), Lyme borreliosis, herpes viruses, human immunodeficiency virus, and human T-cell lymphotropic virus. Patients were screened for antinuclear and antineutrophil cytoplasmic antibodies. The cerebrospinal fluid was analyzed for cell count; glucose and protein levels; and oligoclonal bands and was screened for Lyme borreliosis, syphilis, and herpes virus antibodies.

Spinal magnetic resonance imaging (MR imaging) (Siemens Vision and Symphony 1.5-T imager; Siemens, Erlangen, Germany) was performed on all patients. The patients were screened for aortic disease (computed tomographic scan or ultrasonography) and vertebral dissection (angio–computed tomographic scan or angio–MR imaging) based on the level of involvement. Cardiac ultrasonography was used to screen the patients for a cardioembolic source.

We included 27 patients with spinal cord ischemia. These patients displayed acute myelopathy (onset within minutes to a few hours, in some cases); the workup excluded other causes (compressive, traumatic, infectious, or inflammatory). In most cases, the MR imaging showed a well-defined lesion corresponding to arterial territory. The combined clinical and imaging data were used to determine the level and extent of the infarct. We analyzed the clinical features of the different stroke syndromes and correlated them with the imaging data.

The median follow-up time to discharge from the hospital was 12 days (range, 1-42 days). Fisher exact test was used to examine statistical correlations. Statistical significance was set at P<.05.

Our 27 patients (11 men and 16 women) had a mean age of 56 years (age range, 19-80 years). Patient details about sex, age, and cardiovascular risk factors are summarized in Table 1. Fewer than one third of the patients had cardiovascular risk factors.

Of the 27 patients, 10 (37%) had anterior spinal artery patterns, 4 (15%) had anterior and 4 (15%) had posterior unilateral patterns, 3 (11%) had central patterns, 2 (7%) had posterior spinal artery patterns, and 2 (7%) had transverse syndromes, while 2 (7%) had unclassifiable clinical pictures. The manifestations usually developed within about 2 minutes (within a few hours in a few cases). In 5 patients, they appeared in 2 acute stages. Two of these 5 patients had a deficit that was initially unilateral and then became bilateral, while in another patient the deficit increased secondarily. The 2 remaining patients experienced transient ischemic attacks (TIAs) before spinal cord infarction (1 patient experienced several TIAs per year for 6 years). The TIAs had the same motor features as the definitive stroke. In these cases of progressive stroke or TIAs, the infarct was frequently localized at the cervical level (4 of 5 cases).

Sixteen patients (59%) reported back or neck pain at the onset of symptoms, and this was always localized at the level of the spinal cord lesion. In 13 (81%) of these patients, the pain radiated in a radicular fashion (bilaterally in 11 and unilaterally in 2). Three patients with posterior bilateral or unilateral syndrome experienced paresthesia. The initial back pain and paresthesia resolved spontaneously within days.

The topography of the lesions was associated with specific clinical pictures. These are summarized in Table 2.

The laboratory workup did not show any abnormalities except for previously identified diabetes mellitus and hypercholesterolemia. Cerebrospinal fluid analysis revealed an increased protein concentration in 12 patients (44%), but none had pleocytosis or intrathecal synthesis of immunoglobulins. There were no cases of vertebral dissection or a cardioembolic source. One patient had an aortic endoprosthesis.

Magnetic resonance imaging showed an ischemic lesion defined as a well-demarcated T2-weighted hyperintensity matching an arterial territory of the cord (Figure 3) in 18 patients (67%). Pathologic MR imaging findings were more frequent in the patients with severe weakness (less than M4) than in those with less severe weakness (P = .01). Only 1 patient (with a posterior spinal artery infarct) had a vertebral body infarction, at the upper limit of his cord lesion. No patient displayed signs of arteriovenous fistula or malformation.

Except in patients with central and transverse infarcts, the infarcts were grouped in the following 4 distinct spinal segments: from C1 until C3, C4 until C7, T3 until T7, and T8 until L1. None of the patients had spinal cord infarction between T1 and T3.

No identifiable cause was found in 20 patients (74%). Three patients had experienced a prolonged episode of arterial hypotension just before the infarction. The other most common factor was disk prolapse or herniation (3 patients), diagnosed in all cases by MR imaging; in 1 patient there was a median disk prolapse at the level of the infarct, with no compression of the spinal cord (Figure 4A), and in 2 patients the disk herniation was lateral and compressed the root at the level of the arterial lesion (Figure 4B). Among the other potential causes, 1 patient had an aortic endoprosthesis, and the 2 patients with a transverse infarct had experienced prolonged arterial hypotension.

The cases in which the spinal arteries were clearly exposed to mechanical tension associated with spine mobility, spinal diseases, and infarct-triggering movements and those with prolonged hypotension and symptomatic arteriopathy (peripheral or coronary) are summarized as a function of stroke syndrome in Table 3. Spinal disease, seen in 12 patients, was associated with chronic back or neck pain with radicular radiation in 5 patients and with spine compression fractures in 4 patients, while the other 3 patients showed spondylolisthesis, chronic arachnoiditis, or chronic cervical disk protrusion. In 11 of these 12 patients, the spinal disease was at the level of the cord infarct.

The ischemic symptoms developed immediately after a movement in 13 patients (48%). In 7 patients, this was a movement of the back (an extension in 4 of them), while in 3 patients it was an arm movement, and in the remaining 3 patients it was a Valsalva maneuver or a gait initiation. The level of the spinal cord lesion corresponded with the level of mechanical stress in the spine.

Anterior and posterior spinal artery infarcts and unilateral infarcts were associated with mechanical triggering movements (P = .02) and with acute and chronic spinal disease (P = .003). None of the patients with central or transverse infarcts reported a triggering movement or spinal disease, but these patients had a higher frequency of underlying arteriopathy (2 [40%] of 5 patients) than patients with other stroke syndromes (1 [5%] of 20 patients).

The outcome was generally favorable, with complete or incomplete recovery in 19 (70%) of 27 patients. Only 13 patients (48%) showed significant gait impairment on leaving the hospital. The clinical outcomes of the patients are summarized in Table 4. Motor deficits showed a higher frequency of recovery (17 of 19 patients who showed recovery) than sensory or sphincter deficits.

The 3 patients with an anterior spinal artery infarct at the cervical level and bibrachial paresis regained normal function in both arms. One of these patients developed unilateral amyotrophy localized in the territory of the C8 nerve root.

The patients received various treatments. Eight patients had no specific treatment, while 9 patients were treated with antiplatelet therapy (300 mg/d of aspirin), 5 patients with corticosteroids, 3 patients with antiplatelet therapy plus corticosteroids, and the remaining 2 patients with anticoagulation therapy.

The MR imaging showed a cord lesion in 18 (67%) of our patients. Sagittal and axial T2-weighted images were considered most useful. The radiological patterns for the different stroke syndromes are shown in Figure 3. These patterns correlated well with data from autopsy and imaging studies.^9-13 Infarction of a vertebral body has been described as a useful confirmatory sign of cord infarction,^14,15 but in our series it was rare (1 of 18 patients). Vertebral lesions are probably more frequent in surgical series in which the patients undergo resection of multiple branches of the aorta.

Diffusion-weighted imaging offers promising techniques to detect early cord ischemia,^16,17 but the quality of echo-planar diffusion-weighted images is reduced because of susceptibility artifacts due to the heterogeneous magnetic environment of the spine that can create false-positive results. Line scan diffusion imaging is less sensitive to artifacts,^18,19 but this capability is available from only a single main manufacturer of MR imaging machines.

In our series of patients, 2 different pathogenetic mechanisms were distinguished. These included (1) those occurring after a mechanical triggering factor (bilateral anterior or posterior spinal artery infarcts and unilateral infarcts) and (2) infarcts occurring after prolonged hypotension or in arterial insufficiency (central and transverse infarcts).

The most probable cause of infarcts occurring after hypotension or arterial insufficiency seems to be global hypoperfusion of the spinal cord. Zülch and Kurth-Schumacher⁷ and Nagashima and Shimamine²⁰ showed that a transverse ischemic cord lesion is reduced to a central lesion at its superior and inferior extremities, so central and transverse lesions probably represent 2 poles of a continuum. The infarcts involve the territories of several spinal arteries in the thoracolumbar region. This vulnerability to hypoperfusion is explained by the local high density of motoneurons²¹ and by the frequency of atherosclerosis in the aorta and iliac arteries. In an autopsy series, aortic atherosclerosis showed a correlation with lacunar infarction in the central region.²²

The most probable cause of infarcts occurring after mechanical stress in our series was a lesion of the radicular arteries, as proposed by other authors.^11,23 In contrast to central and transverse infarcts, the localization of the infarction corresponded to the territories of the radicular artery and spinal arteries. Unilateral forms are explained by the incomplete linking of posterior systems²⁴ and by duplication of the anterior system.^3,4 The T1 to T4 region (which is not supplied by the radicular arteries) was spared, as in a previous study.¹³ The initial pain showed a radicular distribution in many patients, sometimes with clear signs of a radicular lesion at follow-up (selective unilateral C8 amyotrophy). Vertebral body infarctions (which can be associated with cord infarction, as in a patient in our series) are another sign of radicular artery occlusion.^14,15

Unlike the cerebral vessels, the spinal arteries run along a mobile structure, which makes them prone to mechanical damage.¹⁰ The relevance of spinal diseases in our patients was difficult to assess because of the great variety of the pathologic conditions. In an asymptomatic population, Jensen et al²⁵ found that 27% had disk protrusion. However, if one includes chronic spinal diseases and infarcts due to disk lesions, mechanical factors were present in 15 (75%) of 20 patients with anterior and posterior spinal artery or unilateral stroke syndrome. This frequency is significantly higher (P = .003) than that in transverse and central infarcts. Moreover, the infarction coincided with the level of the spinal disease. Chronic compressive diseases induce foraminal and subarachnoid fibrosis, which reduces the collateral blood flow, endangering the mechanisms of compensation in the event of ischemia.²⁶ Although it is described as a classic mechanism of spinal cord damage in compressive myelopathy,²⁷ dysfunction of venous drainage leads to a subacute progressive course.²⁸ Therefore, it seems unlikely that this mechanism played a significant role in our acute cases.

Movement of the spine can lead to acute vascular compression, as demonstrated in 3 of our patients. In our series, the target of this compression appeared to be a radicular artery, but involvement of the anterior spinal artery²⁹ or lumbar artery³⁰ has also been demonstrated. New techniques of positional imaging should more easily demonstrate dynamic vascular compressions in the future.³¹ In patients with damaged facets of the cervical or dorsal column leading to hyperlaxity, the spinal cord might be injured directly from mechanical compression. This usually occurs in traumatic settings or subacutely in predisposing conditions (eg, familial dystonia or spondylodiskitis sequela). In such cases, the imaging can be nonspecific, showing only spinal cord swelling.³² Although none of our patients had a predisposing condition or a recent substantial trauma, we cannot exclude that such a mechanism might have played some role because we did not formally rule it out with dynamic imaging.³³

Occlusion of spinal arteries and veins by fibrocartilaginous emboli has been reported after spine movement.³⁴ In this situation, one typically observes a free interval between the painful movement and the onset of neurological manifestations. The evolution is characteristically progressive without improvement and almost always lethal.

Transient ischemic attacks occurred at the cervical level in our patients. This region is more resistant to ischemia because of its intense vascularization. Transient ischemic attacks could result if vascular compromise resolves in this region. Spinal claudication, first described by Dejerine³⁵ in 1911, occurs if TIAs become repetitive, as in a patient in our series who experienced TIAs every time he worked with his arms (gardening), creating tension in his cervical spine. In another patient who experienced TIAs, the MR imaging showed a particular pattern with several small lesions surrounding the main infarction (Figure 3J). Transient ischemic attacks can also be observed in the progressive course of dural fistula.³⁶ In this case, these phenomena are attributed to venous hypertension³⁶ during a Valsalva maneuver or to arterial steal.³⁷ Although there were no clinical indications of dural fistula in our series (preexisting chronic myelopathy or worsening during the follow-up period) or any suggestive MR imaging sign, we cannot exclude that this particular mechanism may occasionally play a role.

Previous reports have shown that the prognosis of spinal cord infarction following aortic surgery is worse than that for other infarcts.^38-40 During hospitalization in our series, we generally observed a partial recovery, and gait was infrequently compromised. Unilateral infarcts had a particularly favorable prognosis.

There are no clear guidelines for the treatment of spinal strokes. Animal investigations have shown some benefit with certain agents, such as prostaglandins, nimodipine, naloxone hydrochloride, adenosine, thiopental sodium, magnesium, N-methyl-D-aspartate antagonists, or corticosteroids,⁴¹ but there have not been any prospective clinical studies, to our knowledge. de Sèze et al⁴² did not note any difference in the clinical course in a retrospective study of patients treated with corticosteroids or anticoagulation in which all patients also received antiplatelet therapy for secondary prevention.

The long-term functional prognosis seems to be determined by the degree of spinal cone sparing.⁴³ Another important problem is the neurogenic pain that generally develops in the hypesthetic area after a few months. This phenomenon is seen most frequently with anterior spinal artery infarcts.⁴⁴

When imaging results are negative, spinal cord infarction remains largely an exclusion diagnosis. In the future, new techniques should more clearly demonstrate the developing infarction in the acute phase. Early diagnosis would allow the consideration of treatments used in brain strokes, such as thrombolysis.

Correspondence: Jan Novy, MD, Department of Neurology, Centre Hospitalier Universitaire Vaudois BH10, 1011 Lausanne, Switzerland (Jan.Novy@chuv.ch).

Accepted for Publication: March 8, 2006.

Author Contributions:Study concept and design: Novy, Carruzzo, and Bogousslavsky. Acquisition of data: Novy, Carruzzo, and Maeder. Analysis and interpretation of data: Novy, Carruzzo, Maeder, and Bogousslavsky. Drafting of the manuscript: Novy and Carruzzo. Critical revision of the manuscript for important intellectual content: Novy, Carruzzo, Maeder, and Bogousslavsky. Administrative, technical, and material support: Novy and Bogousslavsky. Study supervision: Carruzzo, Maeder, and Bogousslavsky.