Systematic Review of Methods for Assessing Leptomeningeal Collateral Flow

BACKGROUND AND PURPOSE: The importance of LMF in the outcome after acute ischemic stroke is increasingly recognized, but imaging presents a wide range of options for identification of collaterals and there is no single system for grading collateral flow. The aim of this study was to systematically review the literature on the available methods for measuring LMF adequacy. MATERIALS AND METHODS: We performed a systematic review of Ovid, MEDLINE, and Embase databases for studies in which flow in the leptomeningeal collateral vessels was evaluated. Imaging technique, grading scale, and reliability assessment for collateral flow measurement were recorded. RESULTS: We found 81 publications describing 63 methods for grading collateral flow on the basis of conventional angiography (n = 41), CT (n = 7), MR imaging (n = 9), and transcranial Doppler (n = 6). Inter- and/or intraobserver agreement was assessed in only 8 publications. CONCLUSIONS: There is inconsistency in how LMF is graded, with a variety of grading scales and imaging modalities being used. Consistency in evaluating collateral flow at baseline is required for the impact of collateral flow to be fully appreciated.

L eptomeningeal collaterals are anastomotic vessels providing alternative routes for blood flow in stroke. 1 In chronic hypoperfusion due to severe carotid stenosis or occlusion, flow via leptomeningeal vessels can maintain cerebral blood flow when primary collateral flow (via the arterial segments of the circle of Willis) is insufficient. 2,3 Better LMF is associated with less infarct growth and better outcome following acute stroke, 4,5 while poor collateralization is associated with hemorrhage after IA thrombolysis. 6 Numerous studies, using several imaging modalities and grading methods, suggest that leptomeningeal collaterals confer a benefit in stroke. Because the influential role of collaterals has been repeatedly reported, we conducted a systematic literature review to investigate the available LMF assessment methods.

Materials and Methods
MEDLINE and Embase were searched from inception to week 32,2009, by using the Ovid on-line portal for LMF assessments. The search strategy is shown in Appendix 1. The search was supplemented by review of journal electronic tables of contents and by searching the bibliographies of relevant articles; when full text was unavailable, authors were contacted. Studies that graded LMF, published in English and performed on humans, were considered, with assessments on patients with Moyamoya disease excluded. The target population included patients with acute stroke (Ͻ24 hours from onset) or patients with known cerebrovascular disease who had collateral flow assessed at later time points. Studies that graded collateral flow and provided a description of the assessment method were included. Terms such as "cortical/pial anastomoses" were judged as being synonymous with leptomeningeal collaterals and were assessed according to the same criteria. Positron-emission tomography and single-photon emission CT examinations, which indirectly evaluated collaterals, were excluded. Included publication dates ranged from January 1965 to October 2010.

Discussion
The quality of LMF is reported to be an independent predictor of outcome after acute ischemic stroke, after adjustment for other known prognostic factors such as age, clinical stroke severity, baseline imaging characteristics, occlusion site, treatment, and recanalization/reperfusion 4,5,43,63,65,69,72 and suggests that, as a minimum, there is a need to account for its influence on outcomes after stroke. Good collateral flow is assumed to be associated with favorable outcome as a consequence of maintaining the ischemic penumbra for longer until reperfusion occurs, though the effect of collaterals appears to be independent of conventional indices of penumbra such as arterial recanalization/reperfusion. 5,65 It is unclear whether the collateral grade represents an inherent characteristic of individual subjects or a potential therapeutic target. Collateral flow grades on CTA are reportedly better in patients who undergo imaging later after symptom onset, while better collateral flow grades on conventional angiography have been reported in patients treated with statins before stroke, 18 suggesting that collateral flow is dynamic and could potentially be modified.
The fact that LMF is not accounted for in occlusion classifications may be important in defining arterial occlusions at the entry to a clinical trial and adoption of scoring systems from coronary artery disease; notably, the TIMI 87 system or minor modifications of such systems (eg, thrombolysis in cerebral ischemia) ignore fundamental differences in the acquisition of images and the anatomy of the different vascular beds. For example, when applied to the cerebral circulation, a "good" TIMI score on CTA (eg, TIMI 2) could actually represent a complete arterial occlusion (no anterograde flow) with extensive retrograde flow via collaterals. A consistent method for assessment and grading is required to investigate collaterals in acute stroke. Our review revealed wide variation in the methods for grading LMF, few of which are supported even by measurement of observer agreement.
Conventional angiography, considered the criterion standard for assessing cerebrovascular anatomy, can reveal retrograde collateral perfusion in a dynamic fashion and has been used for LMF assessments in the largest number of patients. The most frequently used scale was proposed by the American Society of Interventional and Therapeutic Neuroradiology in an effort to homogenize grading with angiography 16 , but an assessment of interobserver agreement has not yet been reported. Good intraobserver agreement has been demonstrated with angiography when LMF was graded according to the anatomic extent of retrograde flow ( ϭ 0.81). 6,14 The Qureshi scale also demonstrates good interobserver agreement but does not focus on LMF independently. One collateral grading scale quantified collateral flow according to the time taken for contrast to travel from the ICA to the M2 segment of the MCA via collaterals but described flow through primary collaterals of the circle of Willis rather than through cortical anastomoses. 88 Although not truly grading LMF, it provides a quantitative time-based measurement that could potentially be used for collateral assessments. Because LMF is derived from neighboring arterial territories, its quality may only be fully evaluated when the contribution of all potential inflow sources is assessed. Descriptions of arterial injection sites are infrequently provided, and even when available, the contribution of the whole cerebral circulation is seldom evaluated. Conventional angiography is invasive and is usually performed when a patient is being considered for IA therapy which, in general, is reserved for those with contraindications to intravenous treatment (eg, presentation beyond 4.5 hours with favorable appearances on CT), meaning that angiographic LMF assessments are predominantly restricted to this group. Because multimodal CT and MR imaging are increasingly used in clinical practice and before entry to clinical trials, they offer a larger potential population in which LMF can be assessed noninvasively.
Although lacking dynamic information, CTA permits visualization of the extent of LMF. The independent predictive value of collaterals has been confirmed with different CTA methods, and interobserver agreement within different grading scales has been assessed. 64,69,71,72 Retrograde flow relative to a proximal arterial occlusion provides a measurement of LMF adequacy, but grading LMF this way for more distal occlusions may be more difficult. A collateral scoring system based on contrast enhancement in defined regions of interest provides a scale not dependant on a specific occlusion, which could potentially be applied in a larger patient population. 64 The addition of CT perfusion to CTA adds important dynamic information to confirm that collateral flow is truly retrograde and demonstrates excellent interobserver agreement. 5 New multidetector scanners that enable simultaneous acquisition of both CTA and CTP allow dynamic collateral flow assessment with CT. 89 LMF assessments with MR imaging use different imaging characteristics to infer the presence of collaterals. FLAIR vascular hyperintensities due to retrograde flow in leptomeningeal vessels have been associated with larger mismatch volumes and smaller subacute infarct volumes, while abnormal vessels on T2* imaging may be due to deoxygenated blood in collaterals and are associated with smaller infarct volumes. 80,81 ASL, by using different criteria, has also been used to grade collateral flow with MR imaging. 13,51,78 These and other LMF assessments with MR imaging have not been replicated nor has interobserver reliability been graded, and it remains to be seen if they represent robust means of assessing collateral flow.
Relative blood flow velocity and vessel pulsatility have been used as surrogate markers for leptomeningeal collateral flow by using TCD in a small number of studies, but the criteria for defining LMF varied among publications ( Table 5). The lack of an agreed definition for LMF on TCD, absence of direct  collateral visualization, and difficulty in finding acoustic windows are limitations of TCD, though these are offset by the lack of radiation and contrast requirements. 90 Flow diversion on TCD, defined as increased flow velocity in ipsilateral ACA/  PCA, did correlate with angiographic collateral grade when methods were compared, suggesting a possible role for TCD to measure LMF. 35 When collaterals were measured by using DSA, CTA, and MR imaging, CTA compared favorably, but the methods used for grading LMF on CTA were not clearly stated, so this finding must be interpreted with caution. 36

Conclusions
The presence of flow in leptomeningeal collaterals is linked with positive outcomes after stroke, but there is little consistency in the methods used to grade the efficacy of collateral flow. Although the importance of leptomeningeal collaterals is consistently reported, the inconsistency in imaging methods and grading currently limits the emphasis that can be placed on collaterals. For targeting collateral vessels in stroke therapeutic strategies, consistency in examining their extent at baseline is required to permit further expansion of this area. At present, conventional angiography remains the method that can best measure collateral extent and number, but CT-based techniques, which have demonstrated good interobserver reliability and correlation with clinical outcome, may provide an accessible and reliable assessment method for grading collateral flow in a larger patient population, particularly with the development of dynamic CTA combined with perfusion imaging. MR imaging and TCD have been used less frequently than angiography or CT but can also provide noninvasive measurements of LMF.