Mechanism of Ischemic Infarct in Spontaneous Cervical Artery Dissection

Extracranial cervical artery dissection (CAD) accounts for nearly 20% of cases of ischemic stroke in young adults.¹ Some authors have suggested that artery-to-artery embolism is the main mechanism of stroke in CAD,^2–4 whereas others assume that reduced flow from the primary cervical lesion plays a crucial role,^5,6 yet determining whether most CADs leads to cerebral ischemia because of embolism or because of hemodynamic failure may influence the therapeutic approach. For several authors, reduced flow from the primary cervical lesion could be approached through endovascular revascularization,^7–13 whereas anticoagulation or antiplatelet regimens would be more appropriate to prevent secondary embolic events.^14,15

Imaging can be used as an end point in distinguishing hemodynamic from thromboembolic infarct. Acute thromboembolism may be demonstrated directly on brain MRI using a T2* sequence, because intraluminal acute thrombus appears as a signal loss along the course of occluded symptomatic cerebral arteries.¹⁶ A presumed embolic mechanism may also be evoked indirectly on diffusion-weighted imaging (DWI) in the case of a pial or perforating territory stroke pattern, whereas junctional or watershed infarcts are more likely to be of hemodynamic origin.^17–19 The aim of this study was to identify the most likely mechanism of stroke using cervical and cerebral imaging parameters in consecutive patients with CAD.

We identified 172 consecutive patients with CAD (198 dissected arteries, 131 carotid CADs, and 67 vertebral CADs) including 100 patients (100 of 172 [58%]) with acute stroke on DWI (117 dissected arteries: 80 carotid CADs and 37 vertebral CADs). The remaining 72 patients had no DWI (n=9) or normal DWI (n=63) and were excluded (Figure 4). The Table summarizes the clinical and DWI characteristics of the population. Median delay from onset to brain MRI was 3 days (interquartile range, 1–7; mean, 5.1; range, 0–26 days). None of the patients had an old stroke. Among the 100 patients with stroke, 50 patients had multiple infarct patterns. The total number of noncontiguous DWI lesions was 338 with multiple lesions in 56 of 100 patients and lesions in multiple territory in 13 of 100 patients. Interobserver agreements were high for SVS on T2* (κ=0.91; 95% CI, 0.82–1) for asymmetry of signal or size in intracranial vessel on 3-dimensional time-of-flight MR angiography (κ=0.88; 95% CI, 0.80–0.96), for hyperintense vessel sign on FLAIR (κ=0.86; 95% CI, 0.77–0.95), for stroke pattern definition on DWI (κ=0.84; 95% CI, 0.76–0.92), and for dissected artery patency on CE-MRA (κ=0.90; 95% CI, 0.83–0.96). Stroke was attributed to a thromboembolic mechanism in 85 of 100 patients with direct visualization of SVS in the symptomatic intracranial artery in 57 of 100 patients and stroke pattern₁ and/or pattern₂ in 85 of 100 patients. Stroke was attributed to a hemodynamic mechanism in 12 of 100 patients, including 4 patients with isolated watershed infarction and 8 patients with stroke pattern₁ and/or pattern₂ associated with pattern₃ and a hemodynamic score ≥2. Stroke was attributed to a mixed mechanism in the remaining 3 patients (3%). Overall, stroke involved pial artery territories (pattern₁, 83 of 100) or perforating artery territories (pattern₂, 17 of 100) in all except 4 patients (96%). When an SVS was present, we did not observe isolated watershed infarction. In 4 patients with SVS, a pattern₃ was associated with a pattern₁ and in 1 case with both pattern₁ and pattern₂. The proportion of embolic and hemodynamic stroke did not differ between anterior and posterior circulation stroke (embolic stroke: 52 of 70 and 23 of 30, respectively, P=1; hemodynamic stroke: 9 of 70 and 3 of 30, respectively, P=0.51).

This exploratory study of cervical and brain MRI in 100 carotid and patients with vertebral CAD stroke yielded the following results: (1) intracranial thrombus was seen in 57 of 100 patients; (2) pial and perforating artery territory stroke was present in 96 of 100 patients; (3) multiple DWI lesions were seen in 56 of 100 patients; and (4) isolated watershed infarction was seen in 4 of 100 patients. These results suggest that stroke in cervical artery dissection was most frequently associated with artery-to-artery embolic events both in anterior and posterior circulation.

Determining whether most dissections lead to cerebral ischemia because of artery-to-artery embolism or because of hemodynamic failure is important, because it may influence the therapeutic approach. Current therapeutic options in CAD include antiplatelet (eg, aspirin), anticoagulation, endovascular treatment with stent deployment, and often a combination when medical treatment fails to prevent ischemic stroke.^8–10 The natural history of CAD indicates a risk of recurrent stroke, mainly within the first weeks or months.^40,41 By analogy to cardioembolic stroke (eg, atrial fibrillation), in which anticoagulation is superior to antiplatelets for secondary stroke prevention,⁴² many physicians use anticoagulants on the assumption that this prevents embolism from thrombus at the dissection site more effectively than antiplatelet agents.^15,43 However, a large study⁴⁴ and a systematic review of nonrandomized studies showed no evidence of a therapeutic benefit favoring either antiplatelet or anticoagulant treatment in preventing stroke, transient ischemic attack, or death in CAD.¹⁴ The presumed benefit of the deployment of a stent all along the dissection includes dissected artery flow restoration. Although many patients who have failed conservative medical therapy are referred for endovascular treatment with angioplasty and stent placement, there have been no well-designed studies to support this practice.^7,8 Although valid therapeutic trials should address these questions, a trial remains difficult.^15,45 In the absence of comparative studies, the presumed mechanism of cerebral infarction is unlikely to impact on clinical decisions but may be useful to define selection criteria for randomized controlled trials.

The literature provides support for several concepts regarding the primary cervical lesion and the secondary intracranial lesion in CAD. An autopsy case has provided the only direct demonstration of secondary intracranial emboli.⁴⁶ Three studies described indirect signs of emboli on transcranial Doppler monitoring studies,^47–49 but the sample sizes of these studies limit the significance of these results. According to current concepts relating mechanism and stroke pattern, strokes in pial or perforating artery territories are more likely to be embolic,⁵⁰ whereas border-zone infarcts are more likely to be hemodynamic.⁵¹ This scheme has been used, in DWI- and CT-based studies, to shed light on the stroke mechanism in patients with CAD with conflicting results.^{3,5,6,46,52,53} The apparent discrepancy between DWI- and CT-based study results may be due to the better identification of acute ischemic pattern when using DWI as compared with CT scan. Thus, CT might underestimate multiple punctuate lesions and perforating artery stroke.⁵ Like others,⁴ we observed hemodynamic infarct patterns in <10% of patients. Our results challenge the findings of Lanczik et al⁵³ and Koch et al⁵ who demonstrated border-zone infarcts in 7 of 24 and 9 of 14 patients, respectively. In addition to the small number of patients of these studies, these results were most likely overestimated because of variability in the vascular anatomy and the misclassification of branch artery occlusion as watershed infarcts. Benninger et al,⁴ based on stroke pattern on CT scan and/or MRI, suggested that embolism is the essential stroke mechanism in CAD. As a novel observation, we directly demonstrated the occlusive intracranial thrombus in vivo with a T2* sequence in 57% of patients. The T2* SVS has been used in patients with acute stroke for >10 years.¹⁶ The susceptibility effect has been ascribed to local ferromagnetic field distortion associated with deoxyhemoglobin components.⁵⁴ Most successive studies reported detection rates of approximately 50%, in cohorts of patients with hyperacute stroke,^{16,23,54–56} with a 100% specificity.⁵⁷ Direct comparison with DSA⁵⁸ showed that SVS distinguishes embolic occlusion from stasis due to low flow. Other findings that substantiate the concept of artery-to-artery embolism in the pathogenesis of stroke in CAD were the multiple acute DWI lesions in the majority of patients and the occurrence of pial artery or perforating artery territory stroke in 96% of patients.

If the simplification of the relationship between the mechanism and stroke pattern provides an easy approach to the presumed mechanisms of cerebral infarction in most patients, one should bear in mind that, even at the individual level, both mechanisms may be at play and may reinforce each other in causing the ischemic lesion.⁵⁹ As recently suggested, border-zone infarcts might result from mixed mechanisms by impaired clearance of emboli in hypoperfused regions.⁶⁰ Indeed, during the constitution of the mural hematoma, only the embolic mechanism may explain the occurrence of an infarction, whereas the hemodynamic mechanism may appear progressively with the narrowing of the lumen due to the growth of the mural hematoma. Once the lumen is severely stenosed, hemodynamic impairment may encourage the formation of secondary clots in border-zone regions, impede the clearance of distal clots, and increase the impact on cerebral tissues. The fact that, in the present study, border-zone infarction rarely occurred in isolation but mainly occurred in association with pial or perforating artery territory infarction supports the hypothesis of an impaired clearance of small emboli broken up to multiple small branches. Nevertheless, the small number of patients with border-zone infarcts prohibits any definitive statements on this issue.

Our study suffered from several limitations. In addition to selection bias of patients addressed to a referral center and the variability in the timing of imaging studies, we should note the imperfect sensitivity of the SVS on T2*,¹⁶ which we considered as an in vivo gold standard for direct visualization of the thrombus. In addition, an embolism may have lysed or migrated at the time of brain MRI. However, such misclassifications would have weakened the link between stroke and the embolic mechanism. The SVS has never been compared with DSA in patients with hyperacute stroke to determine its value in predicting the arterial occlusion site and extent. Consequently, one can argue that SVS may correspond not to an embolus but to a prolongation of the thrombus from the dissected cervical artery into the intracranial arteries. This third theoretical mechanism might have occurred in occlusive CAD. However, the autopsy of a patient⁴⁶ who died from a middle cerebral artery stroke did not provide evidence for this mechanism. The luminal thrombus extended from 4 cm distal to the right carotid artery to the supraclinoid portion of the carotid artery, but thrombotic material in branches of the middle cerebral artery was independent of the supraclinoid thrombus. In line with our results, the authors concluded that there was a secondary embolism from the primary cervical lesion.⁴⁶ A second limitation is the interindividual variability of the arterial territories.⁶¹ However, even supposing that some additional border-zone infarcts were missed on DWI, we did not use the stroke pattern alone to determine the stroke mechanism. Third, we did not systematically use DSA, the standard of reference for cervical artery stenosis measurement, given that DSA is rarely used in patients with CAD. We cannot exclude the possibility that, because of the imprecision of CE-MRA and Doppler ultrasound in measuring the degree of stenosis, group errors might have occurred. Such misclassifications would have particularly affected vertebral artery dissection, in which standard measurement criteria have not yet been established. Finally, the present article can only explore mechanisms of the primary clinical event and assume that, for each patient, subsequent events will follow the same mechanism. We should bear in mind that an exploration of the likely mechanism of the primary clinical event, no matter how reliable, may not apply to secondary clinical events.

In conclusion, our study suggests that thromboembolism, rather than hemodynamic infarction, is the most frequent cause of stroke in CAD. This implies that prevention of artery-to-artery embolism could play a crucial role in the management of these patients. Our findings may also help the design of a randomized controlled trial to obtain scientific evidence of the role of heparin, antiplatelet therapy, and stenting in CAD.