Arteriovenous malformations (AVMs) occur within the central nervous system as congenital anomalies that allow blood to be shunted directly from arteries to veins without an interposed capillary network. Afferent and efferent vessels are dilated, and they lead to and from a tangle of malformed channels (the nidus) containing arterial blood. Because these abnormal vessels receive blood at arterial pressure, bleeding can result. AVMs can also grow over time through a combination of vessel enlargement and recruitment of new vessels, perhaps involving mechanisms of angiogenesis. Congestion of draining veins with arterial blood flow can also lead to neurologic symptoms by causing venous hypertension and reduced perfusion in adjacent brain areas. Excision of the malformation will restore normal perfusion to the uninvolved brain.
AVMs can be small, with only a single feeding artery, or they can encompass several lobes of the brain and have arterial feeders from multiple sources. Cerebral AVMs are usually conical, with the apex near the ventricles and a broad base at the cortical surface; large cortical AVMs commonly present with seizures caused by irritation of adjacent cortical tissue. Small AVMs are more likely to present with hemorrhage, causing headache, neurologic deficits, coma, or even death. AVMs can be associated with flow-related aneurysms on feeding arteries or can have aneurysms on vessels within the nidus of the AVM itself. If aneurysms are present, they are usually the source of hemorrhage. AVMs can occur in the cerebellum or brain stem but are much less common. Rarely, AVMs occur within the spinal cord, or they may involve only the dura (intracranial or intraspinal).
Patients with AVMs usually develop symptoms before age 40 (but usually not during childhood). The most common presentation is hemorrhage (50–65%), followed by seizures (15–35%). Mortality and morbidity with each hemorrhage are 10–30% and 15–50%, respectively. For the first year after a hemorrhage, the risk of another hemorrhage is 6%, but for all subsequent years the annual risk drops to 2%, which is the risk of hemorrhage in an unruptured AVM that presents with other signs.
Symptoms and Signs of Arteriovenous malformations
Hemorrhage may cause symptoms ranging from headaches to focal neurologic deficits to coma, or the defect can be clinically silent. The site of the hemorrhage will determine the neurologic deficit but the size of a hemorrhage is often irrelevant, as small hematomas in eloquent areas cause more impairment than large ones in silent areas of the brain. Hemorrhage into the ventricles can cause hydrocephalus that must be managed with ventricular drainage. Hemorrhage into the subarachnoid space can cause headache and nuchal rigidity, much like subarachnoid hemorrhage from an aneurysmal hemorrhage. However, the subsequent risk of vasospasm is usually much less after AVM-related subarachnoid hemorrhage than after aneurysmal subarachnoid hemorrhage. AVMs leak less blood that is distributed over the convexities, while aneurysms leak more blood into the basilar cisterns, adjacent to the origins of the major brain arteries as they emerge from the skull base.
Seizures can be focal or general and may cause temporary postictal neurologic deficits. The clinical importance of AVM presentation with seizures is that it may allow for treatment of an AVM prior to symptomatic hemorrhage. There are insufficient data in the literature to predict either the seizure frequency or the prognosis of seizure control based on the size or location of an AVM, though frontal or temporal lobe AVMs are more commonly associated with seizures.
Headaches are common in patients with AVMs. It is difficult to show a precise relationship between headache and AVMs, and the literature is inconclusive on the subject of which (if any) headache syndromes reliably predict an underlying AVM.
Focal neurologic deficits can occur in the absence of hemorrhage or seizure. These symptoms were once thought to be due to “steal” of arterial blood from adjacent areas of functional brain by the high-flow arteriovenous shunt. However, recent studies using measurements of intra-arterial and intravenous pressure and blood flow in AVMs and adjacent cortical territories in patients suggest that focal neurologic symptoms may be produced by venous congestion leading to venous hypertension and ischemia in adjacent cortical territories.
It is worth noting one uncommon but well-known complication of large AVMs in infants: arteriovenous shunting leading to left ventricular dysfunction and high-output cardiac failure. Frequently, these children will have an audible cranial bruit and can present with noncommunicating hydrocephalus due to obstruction of the cerebral aqueduct by enlarged basal veins surrounding the midbrain.
If a patient presents with sudden-onset headache or neurologic deficits, a CT scan is usually all that is required for diagnosis of an intracranial hemorrhage (see below). If this is not diagnostic, lumbar puncture may be required to establish the presence of subarachnoid blood that would prompt rigorous search for a cause.
Any young adult presenting with new-onset severe headache, neurologic deficits, coma, or seizures should first be studied by noncontrast CT scan, which will routinely demonstrate intracerebral or subarachnoid blood if a hemorrhage has occurred and, in the absence of bleeding, may suggest a mass lesion without contrast. CT scans also help to guide immediate management decisions such as the need for emergency hematoma evacuation or ventriculostomy for hydrocephalus and intraventricular hemorrhage. MR imaging with intravenous gadolinium contrast defines the relationships of the AVM to surrounding brain structures and can be critical in assessing the surgical risks and choosing the appropriate surgical approach.
Four-vessel cerebral arteriography is the standard test used to define the vascular anatomy of AVMs. This procedure identifies all feeding arteries, all compartments of the nidus, and all draining veins. In addition, angiography can identify associated aneurysms within the nidus or those on feeding arteries far removed from the nidus. Selective external carotid artery injections should be performed on the side of large lesions because many will have feeding artery contributions from the dura or from external-to-internal carotid anastomoses that can be embolized easily. If surgical resection is planned, angiography with endovascular embolization of large feeding arteries can significantly reduce arterial supply to an AVM prior to surgery and thereby make resection much less difficult.
It should be noted that any or all of the studies described above can fail to detect an AVM nidus immediately after acute hemorrhage because the intraparenchymal clot can compress the nidus and close it off, making it radiographically “silent.” Therefore, in any patient suspected of having an AVM-related hemorrhage but in whom MRI and angiography are not diagnostic, these studies should be repeated 1–3 months after the hemorrhage. By this time, the hematoma will have been absorbed and any underlying vascular pathology should be evident. On occasion, bleeding may be due to a cavernous malformation, another kind of vascular malformation whose characteristic appearance on MRI is described below.
Hemorrhage—intracerebral or subarachnoid—can be caused by hypertension, amyloid angiopathy, diabetes mellitus, hemorrhagic conversion of an ischemic stroke, but these diseases are unlikely in the age group usually affected by AVM-related hemorrhage. Therefore, in the absence of blood dyscrasia, spontaneous intracranial hemorrhage in a child is the hallmark of an AVM. Other diagnostic considerations include an intracranial aneurysm or a primary brain neoplasm or secondary metastatic lesion within the brain that has bled spontaneously. Each of these has a characteristic appearance on MRI or angiography and can be differentiated from an AVM. Cavernous malformations (“cryptic” vascular malformations) are the second most frequently seen cerebrovascular malformation but are nonetheless much less common than AVMs. Their low flow also makes them less likely to bleed (annual risk of 0.7%) and less likely to cause neurologic deficits when they do. These defects have a characteristic “mulberry” appearance on MRI and are usually surrounded by a ring of T2-weighted susceptibility that makes them easy to distinguish from AVMs. They typically present with seizures unless they are located in the brain stem, where even a small venous hemorrhage can cause a profound and sudden neurologic deterioration. These are treated with surgical removal.
A wide range of structural disorders can cause seizures although, as noted before, they usually give rise to a radiographic evaluation that can suggest or prove the presence of an AVM.
Treatment of Arteriovenous malformations
Medical treatment is only effective for relieving symptoms or for reducing posthemorrhagic sequelae. Anticonvulsants can control AVM-incited seizures, and blood pressure control is important after a hemorrhage, especially if the hemorrhage is due to an AVM-related aneurysm.
The ideal treatment of an AVM is excision and removal of any associated hematoma. However, treatment must be guided by two basic principles:
the AVM should be removed completely, because any residual nidus can still bleed;
neurologic function should be preserved.
Some malformations cannot be removed with reasonable risk, and some of these can be treated by endovascular embolization or focused irradiation (stereotactic radiosurgery; see below).
In an attempt to determine which patients should or should not be offered surgery, a number of grading systems have been developed. AVM grading systems are important for estimating the risk of operation compared with the risks associated with the history of the lesion, and for these reasons they should be simple but comprehensive enough to grade all AVMs, address all factors that influence treatment risk, and predict these treatment risks. Most grading systems are based on radiographic features of the lesion. The most widely accepted grading system is the Spetzler-Martin system, which attempts to predict the risk of surgical resection based on three variables: AVM size, pattern of venous drainage, and location in relation to eloquent brain. Size is scored by measuring the greatest diameter of the AVM on angiography, MRI, or CT and relates to the number of feeding arteries, the degree of flow. Venous drainage is scored by determining whether the AVM empties into the superficial cortical venous system or into the deep venous system (internal cerebral veins, basal veins of Rosenthal, or vein of Galen); this relates to the depth of the AVM. Eloquence is scored by the proximity of the AVM nidus to areas of brain that have easily identifiable neurologic function and, when injured, produce disabling neurologic deficits (motor and sensory cortex, deep gray matter, corticospinal tracts, or brain stem). A numerical AVM grade is then assigned by adding the scores for size, venous drainage, and eloquence. AVMs can then be stratified into five grades, with grade 1 lesions being small, superficial, and in noneloquent brain and grade 5 lesions being large and deep in eloquent brain.
Prospective analysis of AVM grade and surgical results in 120 patients showed minimal risk associated with resection of AVMs in patients with grade 1 or grade 2 lesions, with no new neurologic deficits after surgery. We therefore recommend aggressive excision of these lesions. In contrast, surgical resection of grade 4 and grade 5 AVMs resulted in 15–20% incidence of new deficits. Patients with grade 4 and grade 5 AVMs are recommended for surgical treatment only when they suffer repeated hemorrhages or are experiencing a rapid and progressive neurologic decline. Grade 3 lesions remain ambiguous in that some studies suggest these lesions can be resected with low surgical morbidity while others report an unacceptable surgical risk. Patients whose neurologic condition is deteriorating are typically ready to accept surgical risks, whereas patients who are neurologically intact may prefer to postpone treatment until absolute indications are present and the surgical risks become more acceptable relative to the history of the AVM. Because the risk of hemorrhage accrues annually, young patients are exposed to substantial cumulative risk during their lifetimes and may be more amenable to a strategy of surgical excision.
Endovascular embolization as sole treatment of AVMs is unacceptable because complete occlusion rates are low and the long-term occlusion rates are unknown. Therefore, embolization should be used as part of an approach to either devascularize a large lesion before surgery or reduce its size before radiosurgery. In this context, embolization can significantly simplify surgical resection or can convert a large AVM into a much smaller.
Stereotactic radiosurgery is rapidly gaining attention as an alternative treatment for AVMs and has been shown to work well for small defects. However, radiosurgical obliteration is inversely related to AVM size, with larger lesions showing a lower rate of occlusion after delivery of radiation. Generally, radiosurgery takes up to 2 years to achieve full effect, and during that time studies suggest that the risk of hemorrhage from an irradiated AVM may actually increase compared with an untreated AVM. Therefore, we do not currently recommend stereotactic radiosurgery for small AVMs, as these lesions can be surgically resected with no morbidity and with immediate elimination of the risk of hemorrhage. Instead, stereotactic radiation can be used as part of a treatment plan for deep AVMs located in critical brain structures where there is no easy surgical access. Such an AVM can be reduced in size, either by endovascular embolization or partial surgical resection, and then obliterated with stereotactic radiation therapy applied to the much smaller residual nidus.
The complication rates associated with resection of AVMs of different grades are reviewed above. The most important postoperative complication relates to a phenomenon “normal perfusion pressure breakthrough.” Once an AVM has been resected, the shunting of blood through the nidus is gone, thus restoring normal perfusion and normal arterial pressure within vessels of adjacent brain. Because shunting through the AVM has altered the autoregulation of these vessels, these arteries are ill-suited for accommodation of the new influx of arterial blood and can hemorrhage. The actual risk of normal perfusion pressure hemorrhage is unknown, but because of the theoretical risk, blood pressure control is critical during the first 48 hours after surgery while arteries adjacent to the AVM nidus reestablish their normal autoregulation.
Postoperative hemorrhage should also sound the alarm for a possible residual AVM nidus. Nothing short of complete AVM resection will protect the patient from further hemorrhage, and all patients should therefore undergo postoperative angiography to verify complete obliteration of the AVM nidus. Intraoperative angiography should be considered in selected cases of either large or deep AVMs where it can be difficult to determine if resection is complete. In addition, any patient treated with stereotactic radiation should undergo follow-up angiography 3 years after treatment; if there is residual nidus, this may need to be treated with further stereotactic radiotherapy (staged radiosurgery) or with conventional microsurgical removal.
The annual risk of hemorrhage from an unruptured AVM is 2%. This risk doubles during the first year after a hemorrhage but then drops back to the prehemorrhage annual rate for all subsequent years. The morbidity associated with each hemorrhage is 15–50%, and the mortality is 10–30%. In contrast, the overall morbidity from AVM treatment is less than 10%, and the overall mortality is less than 5%, though this varies widely depending on the grade of AVM. The patient’s preoperative neurologic condition is also important. Patients who present with devastating hemorrhage may not deteriorate with treatment but will be left with disabling deficits that require prolonged rehabilitation. Age is also an important factor because young patients are much more resilient after treatment than elderly ones. However, with adequate time, full recovery is usually possible; studies demonstrated that most early neurologic deficits after surgery will resolve.