Selection of Treatment Modalities or Observation of Arteriovenous Malformations

This article provides management guidelines for arteriovenous malformations (AVMs). Management options include observation, surgical excision, endovascular embolization, and radiosurgery. Each of these can be used individually or combined for multimodal therapy based on the characteristics of the lesion. The article stratifies each lesion based on the AVM and patient characteristics to either observation or a single or multimodal treatment arm. The treatment of an AVM must be carefully weighed in each patient because of the risk of neurologic injury in functional areas of the brain and weighed against the natural history of hemorrhage.

Cerebral arteriovenous malformations (AVMs) have variable modes of presentation. The hemorrhagic presentation of these lesions, as identified in the prospective trials of the New York Islands AVM Hemorrhage Study and Northern Manhattan Stroke Study groups, is 0.42 per 100,000 and 0.55 per 100,000, respectively. Evidence based on observation of patients with known lesions shows that the natural history of the disease is unchanged without intervention, even in symptomatic patients. Ondra and colleagues showed in a study of 160 patients followed over 24 years without intervention that the combined major morbidity and mortality was 2.7% per year. This per year risk percentage remained unchanged throughout the course of observation. In light of this, intervention risk should be lower than the natural history of the disease. This article elucidates ideal treatment modalities based on patient-specific factors.

Factors influencing modality of treatment

Once the decision is made to treat an AVM, the type of treatment is based on lesion-specific factors (size, location, and angiographic anatomy) and patient-specific factors. The first patient-specific factor to consider is age. A younger patient, who is more likely to benefit from long-term cure and symptom relief, is often at lower treatment-related risks compared with older patients. The overall health and neurologic state of the patient must also be considered preoperatively, because those with significant medical comorbidities may have reduced longevity, which is taken into account when considering the natural history of these lesions. Lawton and colleagues discuss multiple factors influencing outcome in the setting of hemorrhage, with a positive correlation with younger age, AVMs in noneloquent territory, and Spetzler-Martin Grades 3 or less. AVMs that display diffuse hemispheric or bilateral cerebral involvement may not be amenable to any treatment modality, and thus goals of care may be more palliative or toward symptom relief ( Fig. 1 ).

Seizures associated with AVMs can occur as the presenting symptom or de novo posttreatment. Patients who have had prior hemorrhage, particularly in the temporal lobe, are more prone to seizures. Piepgras and colleagues report that in patients with preoperative seizures, 83% were seizure-free on follow-up, and the remaining had interval improvement in regards to occurrence. In patients presenting without seizures, 6% developed de novo seizures postoperatively. These findings show significant improvement compared with previous studies, likely relating to improvements in surgical techniques. In patients who undergo surgical resection, it is advocated to resect the surrounding gliotic and hemosiderin-stained tissue if possible, which can improve seizure control in patients with this as a presenting symptom. Cao and colleagues used intraopertive electrocorticography to delineate additional surrounding epileptic discharges for bipolar electrocoagulation with significant improvement in seizure control noted on follow-up.

The location of a lesion, in particular with reference to eloquence of cortex, also influences treatment strategies. Historically, lesions in the sensory-motor strip have been less amenable to surgical intervention because of a presumed increase in morbidity. Kato and colleagues demonstrated improvement in 15 of 17 patients treated with surgery. Motor-evoked potentials were used intraoperatively to demonstrate a functional shift in the cortex away from the AVM nidus, thus allowing safe resection. Additionally, in multiple other studies patients with occipital lobe vascular malformations treated with surgical resection or stereotactic radiosurgery have been shown to have seizure improvement or resolution posttreatment with minimal risk to worsening visual function. Based on this, all modalities of therapy may currently be considered for lesions in eloquent cortex.

Giant AVMs, which compromise less than 10% of all malformations, present a unique situation in that these lesions are more prone to ischemia secondary to steal phenomenon. This is thought to result from the loss of the autoregulatory compensation caused by the prolonged dilated state from persistent high flow through these lesions. These lesions often have extensive collateral recruitment from multiple vascular beds, which can be demonstrated with angiography ( Fig. 2 ). Treatment goals are based on the location, extent of hemispheric involvement, and overall condition of the patient. Staged endovascular treatments may be used at times, but the goal of treatment needs to be carefully planned. Volume reduction may be possible with subsequent radiosurgery to the remaining nidus, yet conclusive proof of this therapy is not available. At times, it is necessary to target an intranidal aneurysm with endovascular therapy if the lesion is the obvious source of hemorrhage. Angiographic cure may not be possible in such cases where diffuse involvement is noted, and in such cases, no therapy is recommended. In a review of 53 patients with either Spetzler-Martin Grade IV to V AVMs, Chang and colleagues reported good outcomes in patients with symptomatic giant AVMs (hemorrhage [n = 20]; seizure [n = 18]; progressive neurologic decline [n = 7]; and headache [n = 8]) with multimodal treatment strategies ( Table 1 ). Yet even in this report, there is a bias toward patients with lesions favorable for treatment.

Table 1

37-month follow-up in 53 patients treated with Spetzler-Martin Grade IV or V AVMs

Neurologic Status	Excellent	Good	Poor	Deceased
Pretreatment	31 (58%)	17 (32%)	5 (9%)	—
Posttreatment	27 (51%)	15 (28%)	3 (6%)	8 (15%)

The association of aneurysms in conjunction with cerebral AVMs is an issue in about 20% to 25% of patients with AVMs. These may be incidental aneurysms in a separate vascular bed from the AVM. Aneurysms occurring on expected locations, such as the circle of Willis, that ultimately give supply to the AVM are likely to be flow-related aneurysms and are thought to arise secondary to a hyperdynamic circulatory state induced by the AVM. Yasargil and coworkers comment on the association, including aneurysms less than 3 mm, as being approximately 10.8%, although no direct relationship between either size or degree of flow through the AVM as determined angiographically was associated with the presence of aneurysms. Cortical AVMs, in particular frontal and occipital, seem to have higher preponderance of associated aneurysms ( Table 2 ), and deeper locations, such as basal ganglia and brainstem lesions, were less likely to have flow-related aneurysms ( Fig. 3 ).

Table 2

Site of AVM and frequency of associated aneurysm

Site of AVM	Total No.	No. Patients with Aneurysms	% with Aneurysm
Frontal	48	12	25
Parietal	49	2	4.1
Temporal	53	6	11.3
Insular	23	5	21.7
Occipital	30	7	23.3
Cerebellar	58	6	10.3
Hippocampal	17	2	11.8

Data from Yasargil M. Microneurosurgery – AVM of the brain, clinical considerations, general and special operative techniques, surgical results, nonoperative cases, cavernous and venous angiomas. In: Microneurosurgery, IIIB. New York: Theime; 1988. p. 137.

In one study, the risk of hemorrhage of an AVM associated with an aneurysm was shown to be 7% at 5 years versus a baseline risk of AVM hemorrhage without aneurysm reported at 2% to 4% per year. This remains controversial, because separate studies by Miesal and Thompson and colleagues have not shown any increase in correlation. Although intranidal aneurysms have been associated with increased hemorrhage from an AVM, this has not been shown conclusively. Intranidal aneurysms adjacent to a ventricle may be at increased risks of recurrent hemorrhage ( Table 3 ). Batjer and colleagues also noted a higher likelihood of rupture associated with pedicle or feeding artery aneurysms. In the setting of hemorrhage from a cerebral AVM associated with a nidal, pedicle branch, or flow-related aneurysm, Yasargil and coworkers comment on multiple studies that the AVM is more likely the source of hemorrhage, except in the posterior fossa. Treatment should be focused on the aneurysm if it is the cause of hemorrhage. In the setting of a nonhemorrhagic presentation, aneurysms that are distal or flow-related may regress with AVM treatment of the aneurysm and those arising from the circle of Willis are unlikely to regress with time ( Fig. 4 ).

Table 3

AVM-associated aneurysm classification

Aneurysm Classification	Location
I	Anatomically unrelated
II	Proximal portion of major feeding vessel
III	Pedicle branch of major feeding vessel
IV	Intranidal

Timing of treatment

The reported morbidity associated with the initial hemorrhage from an AVM is relatively low, as discussed by Hartmann and colleagues in a study of 115 patients with hemorrhage secondary to AVM rupture. Ninety-seven of these patients were found to be independent in their daily activities (Rankin Score 1), with 15 being Rankin Score 2 or 3, and only three patients being severely disabled. Data from the prospective Columbia AVM Databank showed the median Rankin Score to be 2 in a study of 241 AVM patients. Of the AVM hemorrhage subtypes, parenchymal hemorrhage was associated with a higher neurologic morbidity. The reported rerupture risk in the first year was reported at 6% to 18%, and returning to baseline risk after the first year. It was also noted that in 120 patients who presented with AVM hemorrhage and underwent surgery, there was a mean change in Modified Rankin Score (MRS) score of +0.89 after resection. This can be attributed to the hemorrhage masking the effect of the morbidity of a craniotomy. Hematoma absorption in patients operated weeks to months after rupture also generates cavities and encephalomalacia that improve surgical access to the nidus that might otherwise have necessitated corticectomy. Unless the AVM and hemorrhage have resulted in an expected major neurologic deficit in that territory, it is advisable to allow the patient time for neurologic recovery. In the urgent setting of a hematoma with resulting mass effect and a neurologic deficit, surgical intervention with evacuation of the hematoma is recommended. This should then be followed by appropriate diagnostic imaging and angiography to fully delineate and characterize the arterial supply and venous outflow in lieu of attempted resection in the same setting. Resection of the AVM at the time of hematoma evacuation could result in significant blood loss and a higher risk of ischemic complications secondary to resection of parenchymal branches along with the vasculature recruited by the AVM. In regards to the timing of radiosurgery after hematoma, Maruyama and colleagues noted in a study of patients status posthemorrhage with hematoma that waiting more than 6 months before treatment to allow for hematoma reabsorption put the patient at greater risk of rehemorrhage in the interval preceding treatment. However, the hematoma should be resolved to the point where it does not alter the accurate targeting for radiosurgery.

Timing of treatment

The reported morbidity associated with the initial hemorrhage from an AVM is relatively low, as discussed by Hartmann and colleagues in a study of 115 patients with hemorrhage secondary to AVM rupture. Ninety-seven of these patients were found to be independent in their daily activities (Rankin Score 1), with 15 being Rankin Score 2 or 3, and only three patients being severely disabled. Data from the prospective Columbia AVM Databank showed the median Rankin Score to be 2 in a study of 241 AVM patients. Of the AVM hemorrhage subtypes, parenchymal hemorrhage was associated with a higher neurologic morbidity. The reported rerupture risk in the first year was reported at 6% to 18%, and returning to baseline risk after the first year. It was also noted that in 120 patients who presented with AVM hemorrhage and underwent surgery, there was a mean change in Modified Rankin Score (MRS) score of +0.89 after resection. This can be attributed to the hemorrhage masking the effect of the morbidity of a craniotomy. Hematoma absorption in patients operated weeks to months after rupture also generates cavities and encephalomalacia that improve surgical access to the nidus that might otherwise have necessitated corticectomy. Unless the AVM and hemorrhage have resulted in an expected major neurologic deficit in that territory, it is advisable to allow the patient time for neurologic recovery. In the urgent setting of a hematoma with resulting mass effect and a neurologic deficit, surgical intervention with evacuation of the hematoma is recommended. This should then be followed by appropriate diagnostic imaging and angiography to fully delineate and characterize the arterial supply and venous outflow in lieu of attempted resection in the same setting. Resection of the AVM at the time of hematoma evacuation could result in significant blood loss and a higher risk of ischemic complications secondary to resection of parenchymal branches along with the vasculature recruited by the AVM. In regards to the timing of radiosurgery after hematoma, Maruyama and colleagues noted in a study of patients status posthemorrhage with hematoma that waiting more than 6 months before treatment to allow for hematoma reabsorption put the patient at greater risk of rehemorrhage in the interval preceding treatment. However, the hematoma should be resolved to the point where it does not alter the accurate targeting for radiosurgery.

Surgical intervention

The history of surgical resection of AVMs dates back to the 1920s, where outcomes were typically poor. The development of cerebral angiography by Moniz in 1927 provided an essential understanding of the vascular anatomy and hemodynamics of these lesions. The acceptance of en bloc resection of these lesions did not become part of standard surgical technique until 1957 after the publication of Olivecrona’s and Landenheim’s paper. Technologic advances, such as the introduction of microneurosurgery by Yasargil in 1969 and bipolar cauterization by Malis, have resulted in a significant reduction in morbidity and mortality compared to the initial documented operative resections.

The principal benefit of surgical resection is the immediate cure ( Fig. 5 ). Additionally, in a study of 54 patients with epilepsy and without a history of hemorrhage who underwent total surgical resection of the AVM, excellent (70%) and good (18.5%) results in terms of seizure control were noted. Treatment of AVMs even in eloquent areas, such as in the occipital lobe and dominant superior temporal gyrus, has had promising results when surgery has been used as the primary modality of therapy. This series had only 1 of 22 total patients with occipital AVMs who developed worsening of a homonymous hemianopsia. However, this and other reports are subject to selection bias.