Relationship of Cerebral Arteriovenous Malformation Hemodynamics to Clinical Presentation, Angioarchitectural Features, and Hemorrhage

Abstract

Cerebral arteriovenous malformations (AVMs) have about a 2% to 4% annual risk of intracranial hemorrhage and are the most common cause of intracranial hemorrhage in younger people, who are subsequently at risk for long-term morbidity and mortality.1-3 Thus, understanding hemorrhage risk is crucial for determining appropriate treatment. Despite the hemodynamic pathophysiology underlying AVMs, current AVM characterization and hemorrhage risk assessment are based largely on anatomical features derived from digital subtraction angiography rather than on flow parameters.4,5 Specifically, the presence of intranidal or feeder artery aneurysms, which are thought to be angioarchitectural surrogates of high AVM inflow, can be predictive of hemorrhage.1 Additionally, venous outflow obstruction manifested anatomically by venous stenosis, deep venous drainage, deep AVM location, and a single draining vein has been associated with increased hemorrhage risk.1 However, the relationship between AVM flow and AVM clinical and angioarchitectural features has not been clearly examined. In this study, we measured flow volume rate in AVM feeders using quantitative magnetic resonance angiography (QMRA), and we analyzed the impact of AVM clinical and anatomical characteristics on flow. METHODS Patient Selection After institutional review board approval was granted, clinical data for all patients with a supratentorial cerebral AVM who underwent QMRA at our institution between 2007 and 2014 were collected and reviewed (n = 80). Patients were categorized on the basis of AVM clinical presentation into 1 of 4 groups: hemorrhage, seizures, incidental, or numbness/weakness. Patients presenting with aneurysmal subarachnoid hemorrhage were included in the incidental group. Patients with > 1 AVM, patients receiving prior AVM treatment, and those with AVMs with feeders from the external carotid artery were excluded. A total of 64 patients were included in the study cohort. Blood Flow Measurements All patients in this study underwent quantitative flow measurements of the extracranial and intracranial arteries and veins with QMRA before any treatment. Zhao et al6 have previously described this technique of blood flow quantification by QMRA. Specifically, all subjects underwent phase-contrast QMRA performed on a 1.5- or 3.0-T magnetic resonance system (Sigma VHi, GE Healthcare, Milwaukee, Wisconsin) using a 4-channel neurovascular coil. The volume flow rate measurements were acquired with the Noninvasive Optimal Vessel Analysis (NOVA) software (VasSol, Inc, River Forest, Illinois). A 3-dimensional (3D) MRA time of flight of the head was obtained with the following parameters: repetition time/echo time, 23/3.3 milliseconds; flip angle, 20°; field of view, 220 mm; section thickness, 1 mm; and matrix, 512 × 256. MRA time-of-flight images were received by the NOVA software on a separate workstation to reconstruct a 3D surface rendering of the vasculature for determining the perpendicular scan plane to vessels of interest. Volume flow measurements based on these positions were performed (repetition time, 10-15 milliseconds; echo time, 4-7 milliseconds; flip angle, 15°; number of excitations, 4; slice thickness, 3 mm for intracranial arteries and 5 mm for neck arteries; field of view, 140 mm for intracranial arteries and 180 mm for neck arteries; and matrix, 256 × 192 for intracranial arteries and 256 × 128 for neck arteries). Velocity encoding was automatically adjusted by the NOVA software. All QMRA flow measurements were performed with an oblique 2-dimensional fast phase-contrast sequence with retrospective gating. Volumetric flow rate (mL/min) in each artery was processed on the NOVA workstation after phase-contrast images had been acquired. This QMRA technique has been validated with in vitro and in vivo models and has demonstrated utility in the hemodynamic evaluation of cerebrovascular pathologies and interventions, including extracranial carotid artery stenosis, intracranial angioplasty/stenting, carotid endarterectomy, and extracranial-intracranial bypass.7-11 Total AVM blood flow was derived from the aggregate flow within the primary arterial feeders relative to flow in their contralateral counterparts, according to the following equation: Alternatively, flow was measured from single draining veins when possible. AVM flow index was calculated as follows: Statistical Analysis Analyses involving patient demographics and angioarchitectural features were conducted with the χ2 test. Univariate analysis to assess the relationship between AVM hemodynamic characteristics and clinical presentation or angioarchitecture was performed with analysis of variance. All analyses were performed with SPSS version 23 (IBM, Inc, Armonk, New York). RESULTS Patient Characteristics The study cohort consisted of 64 patients with a mean age of 39 years, and 28 patients (44%) were female. Fifty-three percent of patients were white, 14% were black, 25% were Hispanic, and 8% were Asian. Twenty-five percent of patients presented with hemorrhage, 39% with seizures, 35% incidentally, and 1% with numbness/weakness. Twenty-seven percent of AVMs were Spetzler-Martin grade 1, 37% were grade 2, 20% were grade 3, 11% were grade 4, and 5% were grade 5. The mean volume of these AVMs was 5.4 mL (range, 0.2-54.0 mL; median, 4.8 mL). Patient Demographics vs Clinical Presentation Although clinical presentation (hemorrhage, seizures, incidental, numbness/weakness) did not vary significantly with age, hemorrhagic presentation was significantly associated with male sex (P = .001). An analysis of race vs clinical presentation revealed that hemorrhage, seizures, and incidental finding were all significantly related to white race (P = .04, P = .006, and P = .005, respectively). Arterial Angioarchitectural Features vs Clinical Presentation The number of feeder arteries, presence of intranidal aneurysms, presence of arterial ectasia, and arterial steal were not associated with clinical presentation (Table 1). However, hemorrhage and seizures were significantly more likely to occur in the absence of a feeder aneurysm (P = .02 and P = .02, respectively). Additionally, the absence of an intranidal fistula was significantly associated with hemorrhage and incidental presentations (P = .03 and P = .01, respectively).TABLE 1: Angioarchitectural Features vs Clinical PresentationVenous Angioarchitectural Features vs Clinical Presentation Although clinical presentation did not vary significantly with the mean number of draining veins or exclusive deep venous drainage, hemorrhage and seizures were significantly more likely to occur in the absence of superficial drainage (P < .001 and P < .04, respectively; Table 1). Hemorrhagic presentation was also significantly associated with the absence of dual superficial and deep drainage (P = .007). Absence of a single draining vein (P = .002) and absence of venous stenosis (P = .01) were more likely to result in incidental finding of an AVM (P = .002). Absence of venous ectasia and a venous varix were significantly associated with hemorrhage and incidental presentations (P = .007, P = .007, P = .04, and P <.001, respectively). Hemodynamic Characteristics vs Clinical Presentation Mean AVM flow was significantly lower among patients who presented with hemorrhage (P = .005) but significantly higher in patients with seizures (P < .001; Table 2). Flow index, however, was comparable between the 4 clinical presentation groups. Venous index was significantly higher in patients with seizures (P = .001).TABLE 2: Hemodynamic Characteristics vs Clinical PresentationHemodynamic Characteristics vs Angioarchitectural Features When angioarchitectural features were evaluated in relationship to AVM flow, presence of an intranidal fistula (P = .05), venous ectasia (P = .006), and venous varix (P = .03) was associated with significantly higher flows, whereas mean flows were not significantly different in the presence or absence of venous stenosis, intranidal aneurysms, arterial ectasia, and arterial steal (Table 3). A similar analysis conducted with the AVM flow index showed that none of the aforementioned angioarchitectural features exhibited a significantly higher flow index (Table 4).TABLE 3: Arteriovenous Malformation Flow vs Angioarchitectural FeaturesaTABLE 4: Arteriovenous Malformation Flow Index vs Angioarchitectural FeaturesaDISCUSSION The role that hemodynamics plays in the pathophysiology of cerebral AVM hemorrhage remains unclear but is of significant interest because hemorrhage risk assessment can guide the management of AVMs. In 1992, Spetzler et al12 demonstrated an inverse relationship between AVM size and feeding artery pressure measured intraoperatively, thereby providing a hemodynamic cause for the observation that small AVMs have a higher hemorrhage risk than large AVMs. This work affirmed the findings of several preceding clinical studies that correlated AVM anatomical features of high AVM inflow and venous outflow obstruction with hemorrhagic presentation.1,13-15 In other words, increased inflow and resistance to flow within an AVM have been hypothesized to contribute to AVM rupture. Previous studies that applied transcranial Doppler ultrasound to measure flow velocities within cerebral AVM feeder arteries, however, generally found no association between flow velocities and hemorrhagic presentation.16-18 More specifically, in a large cohort of 449 patients, Kader et al16 demonstrated insignificantly lower mean flow velocities in the hemorrhage group. They also measured feeding artery and draining vein pressures and concluded that higher feeding artery pressure, but not draining vein pressure, is associated with AVM rupture. Similarly, Mast et al17 and Diehl et al18 determined the lack of predictive value of transcranial Doppler ultrasound findings for clinical sequelae of AVMs, including hemorrhage and steal phenomenon. More recent studies have investigated cerebral AVM hemodynamics using mean transit time acquired by digital subtraction angiography, 4-dimensional flow magnetic resonance imaging, and time-resolved spin-labeled MRA.5,19-21 For instance, Todaka et al19 identified a significant association between hemorrhagic presentation and the presence of a single draining vein. These hemorrhagic cases with a single vein had significantly higher mean transit times within the feeder artery but no difference within the draining vein. Another study reported significantly lower venous-to-arterial time-to-peak ratios, which correspond to higher flow velocities, among high-rupture-risk (exclusive deep venous drainage or deep location) and ruptured AVMs.5 On the other hand, the study by Illies et al20 using 3D time-resolved MRA in a cohort of 72 patients found no significant alteration in transit times within AVMs that had deep location, deep drainage, venous stenosis, or a single draining vein. Ansari et al21 also found no association between AVM hemodynamics and Spetzler-Martin grade. Thus, no consensus has emerged from the studies outlined above on AVM hemodynamics and rupture risk. Here, we present the first study to evaluate AVM flow measured with QMRA relative to AVM hemorrhage, angioarchitectural features, and venous drainage patterns. We found that the number of feeder arteries and presence of intranidal aneurysms, arterial ectasia, and arterial steal were not associated with clinical presentation. Absence of a single draining vein (P = .002) and absence of venous stenosis (P = .01) were more likely to result in incidental finding of an AVM. Flow index did not vary with presentation, but venous index was higher with seizures. Presence of an intranidal fistula, venous ectasia, and venous varix was associated with significantly higher flows, whereas flows were not significantly different in the presence or absence of venous stenosis, intranidal aneurysms, arterial ectasia, and arterial steal. Overall, our findings concur with the multiple other studies based on transcranial Doppler ultrasound and magnetic resonance techniques that did not reveal a clear correspondence between AVM hemodynamics and rupture or venous outflow characteristics. Indeed, our results suggest that cerebral AVM rupture is unrelated to the presence of increased AVM flow. However, we also demonstrated that the absence of a single draining vein and absence of venous stenosis are more likely to result in asymptomatic presentation, supporting the relative importance of AVM outflow resistance over total inflow in AVM hemorrhage risk. These results differ from those of Todaka et al19 and Raoult et al5 that demonstrated essentially higher flow velocities within feeders of AVMs that had ruptured or had a single draining vein, exclusive deep venous drainage, or deep location, which implicates fast nidal inflow in AVM rupture. This discrepancy may be explained by differences in hemodynamic assessment techniques, namely digital subtraction angiography and 4-dimensional flow magnetic resonance imaging vs QMRA. Additionally, those 2 studies had smaller sample sizes (n = 30 and n = 16, respectively), and Raoult et al combined multiple hemorrhagic risk factors into high- and low-risk groups in their analysis. Although QMRA measurements may have lower accuracy in the presence of arterial stenosis and turbulent flow, there were no such cases in the present study; in fact, QMRA measurements have a higher accuracy at higher flow rates, as is the case in AVMs.7 QMRA also has important advantages over phase-contrast techniques alone by providing 3D vessel localization capability and therefore more exact flow measurements. Our QMRA technique has been validated with in vitro and in vivo models and has demonstrated utility in the hemodynamic evaluation of cerebrovascular pathologies and interventions, including extracranial carotid artery stenosis, intracranial angioplasty/stenting, carotid endarterectomy, and extracranial-intracranial bypass.7-11 Nonetheless, the present study and many other publications cited earlier have shown no significant association between increased flow within AVMs and AVM rupture, suggesting a few possibilities. First, the failure to identify a relationship between flow and AVM angioarchitectural characteristics thought to be related to increased hemorrhage risk may potentially be a consequence of inadequate sample size; additional studies with larger sample sizes may be better powered to uncover this association. In fact, despite the sizeable cohort in our study, only 2 AVMs had intranidal aneurysms, 4 had exclusive deep venous drainage, and 5 had deep location. Interestingly, our group recently showed that posterior fossa AVMs, as opposed to supratentorial AVMs, are more prone to developing feeder vessel aneurysms and that such aneurysms are more likely to be the source of posterior fossa hemorrhage.22 We also formerly published that wall shear stress is significantly higher among cerebral AVM feeders harboring aneurysms.23 Another possibility is that flow measurements in the proximal feeder arteries may not reflect the flow that would be attained from measuring distal feeder arteries supplying only the AVM. The proximal feeder vessels measured in this study (A2, M1, P2), for example, supply some blood flow to adjacent brain tissue, not solely the AVM. Indeed, our group also previously demonstrated a progressive decrease in AVM flow after multiple preoperative embolization sessions and diminished wall shear stress after embolization and surgery, subsequently highlighting the important interplay between changes in the nidus itself and AVM hemodynamics after embolization and surgical resection.24,25 Finally, AVM rupture may be related to other hemodynamic factors such as pressure, increased flow velocity, or venous outflow obstruction, as alluded to above. Although elevated feeder artery pressures have been linked to AVM hemorrhage,12,16 feeder artery pressures have not been extensively assessed relative to venous anatomy and drainage patterns. Further inquiry is warranted because hemorrhagic risk assessment is critical in the evaluation and management of cerebral AVMs. CONCLUSION Cerebral AVM rupture appears unrelated to increased AVM flow. However, the absence of a single draining vein and absence of venous stenosis are more likely to result in asymptomatic presentation, supporting the relative importance of AVM outflow resistance over total inflow in AVM hemorrhage risk. Disclosures Dr Alaraj received a research grant from the National Institutes of Health and is a consultant for Cordis Codman. Dr Aletich received a research grant from Micrus and is a consultant for Cordis-Codman. Dr Amin-Hanjani received a research grant from the National Institutes of Health and research support (no direct funds) from GE Healthcare and VasSol Inc. Dr Charbel is a stockholder for VasSol Inc and a consultant for Transonic. The authors have no personal, financial, or institutional interest in any of the drugs, materials, or devices described in this article.