AVMs are vascular abnormalities consisting of fistulous connections of arteries and veins without normal intervening capillary beds. Typically, they are triangular with the base toward the meninges and the apex toward the ventricular system.³ There is an abrupt transition between the arteries that contains variable amounts of smooth muscle and elastic laminae and the dilated veins. The veins have a thickened wall that appears arterialized because of the proliferation of fibroblasts. Within the arteries and veins, evidence of previous thrombosis and recanalization may be evident. Residua of previous hemorrhages, such as dystrophic calcification and blood breakdown products, may surround the AVM along with histologic evidence of hemosiderin-laden macrophages.³ Marked surrounding gliosis may be present as a result of hypoperfusion related to the high-flow, low-resistance AVM shunt “stealing” blood away from surrounding tissue. The overlying leptomeninges are often thickened and sometimes stained with hemosiderin.^3,⁴

AVMs appear as serpiginous isointense or slightly hyperintense vessels that strongly enhance on computed tomography (CT) after the administration of contrast material.⁵^–⁷ Calcification is identified in 25% to 30% of cases.^7,⁸ In addition to identifying the AVM, CT is useful for demonstrating acute hemorrhage and showing a mass effect or displacement of normal anatomic structures. Magnetic resonance imaging (MRI), however, is superior in sensitivity and specificity. It is useful to determine the size, location, and evidence of previous symptomatic or subclinical hemorrhage, as well as secondary changes such as mass effect, edema, and ischemic changes in adjacent brain tissue.⁹ The typical AVM appears as a tightly packed “honeycomb” of flow voids on T1- and T2-weighted images as a result of high–flow velocity signal loss. Increased signal may be seen in thrombosed or low-flow vessels.⁷ Phase-contrast magnetic resonance angiography can be useful in the depiction of flow, although complete definition of complex lesions and their internal angioarchitecuture requires a cerebral angiogram.^10,¹¹

On cerebral angiography, parenchymal AVMs appear as tightly packed masses of enlarged feeding arteries and dilated tortuous veins with little or no intervening parenchyma within the nidus. There is little or no mass effect unless an associated hemorrhage or venous varix is present. Arteriovenous shunting with abnormal early filling of veins that drain the lesion is characteristic of AVMs.⁷ Angiography allows the identification of feeding and draining vessels, as well as associated vascular abnormalities such as aneurysms. Superselective angiography is preferred to delineate the internal angioarchitecture in detail.¹² Although angiography is highly sensitive, it may be negative after acute hemorrhage¹³^–¹⁶ or after spontaneous AVM thrombosis.¹⁷

AVMs are thought to be primarily congenital in origin because of the lack of development of intervening capillary beds.¹⁸^–²⁰ Although this may be part of the story, other influences may predispose to the development of AVMs. De novo formation may occur rarely,²¹ and recurrent AVMs have been documented after negative angiography, particularly in young adults.²² In addition, AVMs may be associated with other vascular malformations.²³ These examples raise questions about whether abnormalities in venous outflow,²⁴^–²⁷ angiogenic humoral factors, and hormonal influences play a role in pathogenesis.²²

AVMs may also have a genetic basis in some populations. In addition to the association with hereditary hemorrhagic telangiectasia (Osler-Weber-Rendu disease), a few rare pedigrees of familial AVM have been described both in the Asian population ²⁸ and in non-Asian populations.²⁹ Some have shown an autosomal dominant pattern of inheritance, although this is variable and no gene has yet been identified. Low statistical power in the available genome-wide association studies may be contributing to the negative analyses.³⁰ No clinically distinguishing features have been found between familial and congenital AVMs, although the familial type may occur at an earlier age.^28,^29,³¹^–³⁴ In families with brain AVMs in successive generations, the age at diagnosis in later generations may be younger than the age of earlier generations.²⁹

Epidemiology

The exact incidence of AVMs in unknown.³⁵ Large autopsy series estimate the frequency of AVM detection to be 1.4% to 4.3%.³⁶ In one population-based study in Olmsted County, Minnesota, the sex- and age-adjusted incidence rate was 1.11 per 100,000 persons.³⁷ In the New York Islands Study, the detection rate of AVMs was 1.34 per 100,000 person-years, and for first-ever AVM-related hemorrhage, it was 0.51 per 100,000 person-years.^38,³⁹ In the Scottish Intracranial Vascular Malformation Study,⁴⁰ the detection rate for an AVM was 0.56 per 100,000 adults per year.^41,⁴² The Australian population-based vascular malformation study revealed an AVM detection rate of 0.89 per 100,000 person-years.⁴³ AVMs are the most frequently detected symptomatic vascular formation⁴⁴ and account for 2% of all strokes^45,⁴⁶ and 38% of all intracerebral hemorrhages in patients between 15 and 45 years of age.⁴⁷ AVMs are a seventh as common as aneurysms, the prevalence of which has been estimated to be 0.2% to 0.8% in the general population.^2,^37,^48,⁴⁹ The prevalence may be slightly higher in Asian populations.⁵⁰

Patients are typically seen initially between 20 and 40 years of age.⁵¹^–⁶⁰ Older patients may be more likely to have hemorrhage and focal neurological deficits and less likely to have seizures.⁶¹ Most studies report an equal^46,⁵⁸ or slight male preponderance in patients with AVMs.⁵¹^–⁵⁴ ^,62 ^,63

The majority of AVMs are located supratentorially.^46,^54,^57,^64,⁶⁵ Less common sites include the cerebellum and brainstem and within the ventricles. In the posterior fossa, the cerebellum is the most common site.⁶⁶ Extracranially, AVMs may be present in the spinal cord, as well as in various organs and soft tissues.

Although AVMs are typically solitary, multiple AVMs have rarely been described.⁶⁷^–⁶⁹ The Cooperative Study of Intracranial Aneurysms and Subarachnoid Hemorrhage reported a less than 1% incidence of multiple AVMs.⁴⁶ Willinsky and colleagues, however, reported a higher incidence of 9% in 203 consecutive patients.⁶⁸ Occasionally, embolization or definitive treatment of a larger dominant AVM will unmask others as a result of changes in hemodynamics.⁶⁸

Multiple AVMs may be present without apparent cause or in association with hereditary hemorrhagic telangiectasia (Osler-Weber-Rendu disease),^46,^67,^68,^70,⁷¹ Wyburn-Mason syndrome,⁷² or soft tissue vascular malformations.⁷³ Cases of AVM in the brain, as well as in the spinal cord, have also been reported.^58,⁷⁴

AVMs may be associated with other vascular malformations.²³ Mixed vascular malformations are those that contain angiographic or imaging characteristics of more than one type of vascular malformation. Of 280 consecutive patients with vascular malformations, 14 had mixed lesions.²³ Although rare, their presence may generate hypotheses regarding a common pathogenesis or evolution of causation among different types of lesions. AVMs have been reported in association with CMs,²³ venous angiomas,^23,^27,⁴⁶ and aneurysms.^55,⁷⁵^–⁸⁵ Nussbaum and coauthors reported the case of a patient with an AVM that drained into a developmental venous anomaly.²⁷ After regression of the initial AVM, new malformations that drained into the same venous anomaly subsequently appeared, thus suggesting the possible importance of venous outflow in the pathogenesis of AVMs. The association with arterial aneurysms is discussed later.

Clinical Findings

Asymptomatic

In one large autopsy series, only 12% of patients harboring an AVM had symptoms related to it.³⁶ Although the exact number of asymptomatic patients is unknown, clinical studies report that 2% to 4% of AVMs detected are incidental findings.^56,⁸⁶ In a population-based study of patients with intracranial vascular malformations, 40% were asymptomatic.³⁷ Because cross-sectional imaging is being performed more frequently, an increasing number of patients are initially being seen either in an asymptomatic state or with symptoms other than hemorrhage.⁸⁷

Hemorrhage

AVMs may be manifested as hemorrhage. In two population-based studies, 38%³⁸ and 65%³⁷ of patients with AVMs had hemorrhage, with a peak incidence in the fifth decade. In the cooperative study mentioned earlier,⁴⁶ 72% of patients with hemorrhage were initially seen before the age of 40. Before CT scanning, most studies reported a high incidence of subarachnoid hemorrhage.^46,^58,⁶² Since the advent of CT, distinction of hemorrhage type has been made clearer. Intraparenchymal hemorrhage is the most common, followed by intraventricular hemorrhage and subarachnoid hemorrhage.^5,^88,⁸⁹ Of 50 consecutive cerebral angiograms in patients with AVM and hemorrhage, 60% had intracerebral hemorrhage, 26% had intracerebral hemorrhage with intraventricular extension, 8% had intraventricular hemorrhage, 4% had subarachnoid hemorrhage, and 2% had evidence of subdural hemorrhage.⁸⁹ Subarachnoid hemorrhage is more common when an AVM is located cortically and can rarely be associated with vasospasm, depending on the location and thickness of the blood.^58,^90,⁹¹ Of patients with primary subarachnoid hemorrhage, 0.6% were due to AVMs.⁷⁵

Risk for Hemorrhage

The risk of bleeding from an AVM has been calculated in several studies. Retrospective risk is calculated by dividing the number of person-years’ exposure from birth to hemorrhage or intervention by the number of patients with hemorrhage. Retrospective risk is based on the assumption that the lesion is congenital and has a constant risk over time, which may not be accurate. Prospective risks are more useful; however, surgical selection and referral bias are common. In addition, follow-up is often inconsistent. For instance, some studies do not require follow-up imaging to confirm a bleeding episode. Furthermore, the length of follow-up is highly variable.

The risk for hemorrhage in patients with AVMs has been estimated to be 2% to 4% per year.^51,^54,^55,^63,^65,^92,⁹³ In a study of 168 patients with no previous history of bleeding and a follow-up greater than 8.2 years, 18% experienced a symptomatic hemorrhage during the follow-up period, for a crude risk of 2.2% per year.⁶⁵ Using life-table analysis, the annual risk was 1.3% at 1 year, 1.7% at 5 years, 1.5% at 10 years, and 2.2% at 15 years. Graf and associates also evaluated the prospective rate of hemorrhage.⁵⁵ The risk for a first hemorrhage in patients with seizures was 2% at 1 year, 13% at 5 years, and 30% at 10 years, for an overall rate of 2% to 3% per year. Similarly, other studies have also reported prospective hemorrhage risks of 1% to 3% per year.^53,^54,⁹⁴

Ondra and coworkers studied 160 patients with AVMs, which represented 90% of such lesions in the country of Finland.⁶³ Patients were initially evaluated because of hemorrhage, seizures, or other symptoms and were monitored over a 24-year period. This study found 147 new hemorrhagic events in 64 patients during the follow-up period, for an overall hemorrhage risk of 4% per year that was constant over time. The mean interval between diagnosis and hemorrhage was 7.7 years. This study started during the pre-CT era and included multiple hemorrhages in some patients. It is important to note that if one considers the risk for only the first hemorrhage, the hemorrhagic risk is 1.7% per year, similar to other studies. Similarly, Graf and Pollock found the initial risk for hemorrhage to be approximately 2% per year. Pollock and colleagues studied predominantly small AVMs (<3 cm).⁹² Of 315 patients, 196 were initially seen with hemorrhage, for a retrospective risk of 1.89% per year. A crude overall rate of 2.4% per year was determined when multiple hemorrhages were included.

A prospectively collected database consisting of 678 consecutive brain AVMs was monitored over 1932 person-years. The overall hemorrhage rate was 4.61% per year, but this included the 258 patients who initially had hemorrhage. Among these 258 patients, the annual risk for hemorrhage was 7.48% per year, as opposed to 4.16% for those initially evaluated because of seizures (n = 260). In terms of angiographic features, the annual risk for hemorrhage in those with deep venous drainage was 5.42% per year (n = 365), for those with an associated brain aneurysm it was 6.93% per year (n = 122), and for those without a brain aneurysm the risk was 3.99% per year (n = 556). In univariate analysis, a hemorrhagic manifestation (hazard ratio [HR], 2.21; P < .001), associated aneurysm (HR, 1.83; P = .01), and deep venous drainage (HR, 1.81; P = .01) were associated with an increased risk for hemorrhage. In multivariate analysis, a hemorrhagic manifestation (HR, 2.15; P < .01) was a significant predictor of future hemorrhage. There was a nonstatistically significant trend toward an increased risk for hemorrhage in patients with AVMs associated with the presence of an associated aneurysm (P = .07) and deep venous drainage (P = .07).⁹³

In a study of 305 consecutive patients observed prospectively, hemorrhage occurred in 26 of those initially seen with hemorrhage during 380 person-years, for a risk of 6.84% per year, and in 16 patients without initial hemorrhage during 512 person-years, for a risk of 3.12% per year.⁹⁵ In those with hemorrhage initially, the rate was higher in the first year (15.4%), 5.3% in the next 4 years, and 1.7% in the next 5 years. Patients initially evaluated because of headache had a 6.5% per year risk for hemorrhage versus an annual risk of 6.44% in asymptomatic patients, 2.2% in those with seizures, and 1.7% in those with neurological deficits. There was a higher risk for rebleeding in children (HR, 2.69), females (HR, 2.93), and those with deep-seated AVMs (HR, 3.07).

In 622 consecutive patients with brain AVMs in the prospective Columbia AVM database, patients were monitored for a mean of 829 days. Intracranial hemorrhage was the initial finding in 282 (45%) patients. During follow-up, hemorrhage occurred in 39 (6%) patients, for an overall risk of 2.8% per year. Independent predictors of hemorrhage during follow-up in multivariate analysis included initial hemorrhagic AVM manifestation (HR, 5.38; 95% confidence interval [CI], 2.64 to 10.96), older age (HR, 1.05; 95% CI, 1.03 to 1.08), deep brain location (HR, 3.25; 95% CI, 1.30 to 8.16), and exclusive deep venous drainage (HR, 3.25; 95% CI, 1.01 to 5.67). Annual hemorrhage rates were 0.9% in patients without hemorrhage initially and in those without deep brain location or exclusive deep venous drainage and 34.4% in those with each of these factors.⁹⁶

Using the multiplicative law of probability, the lifetime risk for AVM rupture can be assessed by using the formula [1 − (risk for no hemorrhage)ⁿ], where n is the number of expected years of life remaining.⁹⁷ This formula assumes that the lesion is congenital and that the risk is constant over time. Assuming a 3% per year risk for hemorrhage, the formula becomes (1 − 0.97). The n value can be obtained from life tables. For example, a 35-year-old patient with 43 years of life remaining yields (1 − 0.97)⁴³ = 73%. An estimation that correlates well with this formula is simply 105 − age. In addition to the assumptions just mentioned, this formula also assumes that the risk is similar among different patients and that AVM morphology and angioarchitecture do not influence the rate of rupture.

Risk Factors Predicting Hemorrhage

Many studies have attempted to elucidate clinical and angiographic predictors of hemorrhagic AVM to further delineate which patients may be at higher risk. Clinical risk factors for AVM diagnosed after the occurrence of hemorrhage found by some include age,^46,⁶¹ sex,⁹⁴ pregnancy,⁹⁸ and hypertension.⁹⁹ Hemorrhage in infants and children is distinctly uncommon.⁴⁶ Before the age of 2 years, patients with AVMs are typically initially evaluated for congestive heart failure, hydrocephalus, or seizures.⁴⁶ Some data suggest that older patients are more likely to have hemorrhage or a focal neurological deficit and less likely to have seizures initially. Older patients were more likely to have concurrent arterial aneurysms, smaller AVMs, and AVMs in an infratentorial location.⁶¹ Regarding gender issues, in a multivariate analysis Mast and coauthors reported that male sex is a risk factor for a hemorrhagic manifestation of AVM.⁹⁴ Conversely, another study showed a slight preponderance of females initially seen with hemorrhage (62 males, 72 females),⁵⁵ but this was not statistically significant. Another study suggested that female sex may be a risk factor for hemorrhage in those with previous hemorrhage.⁹⁵ Pregnancy may increase the likelihood of rupture, but this is controversial and discussed in subsequent sections. One study found a history of hypertension to be independently associated with a hemorrhagic manifestation of AVM.⁹⁹ The association with recurrent hemorrhages, however, was unclear. Furthermore, no other studies have confirmed this observation. Race/ethnicity may also be of importance in predicting the risk for hemorrhage. In a prospective and retrospective study, Hispanic race/ethnicity (HR, 1.9; P = .02) was an independent risk factor for intracranial hemorrhage. After adjustment for selected angiographic factors, the HR was 3.1. There were trends observed for blacks (HR, 2.1; P = .09) and for Asians (HR, 2.4; P = .11).^100,¹⁰¹

Angiographic risk factors for AVMs manifested as hemorrhage have also been studied but are subject to certain biases. Angiography performed after acute hemorrhage must be interpreted carefully because hemorrhage may change the size and morphologic features of AVMs.¹⁵ Furthermore, patients undergoing angiography after hemorrhage had survived to that point. That is, patients with early death were excluded from the populations studied after hemorrhage. Early mortality may also be excluded in some studies because the patients represent a selected subgroup of patients referred to a center with angiographic capabilities.

One large study of 168 patients with AVMs detected before rupture evaluated angiographic characteristics predictive of future rupture. Despite meticulous analysis of AVM size, site, and flow characteristics, no angioarchitectural features were found to be predictive of rupture in this multivariate analysis.⁶⁵

Several studies have indicated that location influences the risk for hemorrhage. The presence of an AVM in a deep location,^59,^77,^95,^96,¹⁰² such as the basal ganglia,^12,^56,¹⁰³ posterior fossa,^56,¹⁰⁴ intraventricular and periventricular areas,^105,¹⁰⁶ and arterial border zones,¹⁰⁷ may predispose to a hemorrhagic manifestation. Some studies report an increased risk for hemorrhage with cerebellar lesions.^84,^105,^106,^108,¹⁰⁹ The higher risk in one study was attributed to the relatively high incidence of associated aneurysms.¹⁰⁸ In contrast to these studies, others found location to be inconsequential in predicting hemorrhagic risk.^60,^63,⁶⁵ Some hypothesize that deeply located AVMs may simply not cause symptoms other than hemorrhage until they reach a size where cortical irritation is possible.

Size has been controversial. Several large studies, including one of unruptured AVMs at the start of follow-up,⁶⁵ found no difference in risk for hemorrhage based on the size of the AVM.^12,^53,^60,^63,^65,^102,¹⁰⁵ Others have found that small AVMs (<3 cm) pose a higher risk for a hemorrhagic manifestation.^{55–58,62,64,99,110–112} One study found that 90% of patients with small AVMs initially came to medical attention because of hemorrhage.⁵⁶ Hemodynamic assessment of small AVMs revealed distinct differences in flow patterns and pressure. Spetzler and associates found higher intra-arterial pressure in smaller AVMs, thus suggesting a potential role in hemorrhage.¹¹⁰ In addition, the transnidal pressure gradient is higher in smaller AVMs.¹¹³ Others question whether the AVM has just not achieved a size at which it can cause other symptoms, thereby leading to a higher proportion of patients initially being seen with hemorrhage. In contrast to these studies, however, retrospective¹¹⁴ and prospective data¹¹⁵ suggest that larger AVMs have an increased risk for hemorrhage. Furthermore, a study of predominantly small AVMs found the hemorrhage risk rate to be similar to that in other studies, which suggests that there may not be a difference in terms of size.⁹²

Feeding arteries and venous drainage have also been assessed. Norris and colleagues evaluated a number of angiographic features in 31 patients, including size and several arterial and venous parameters.¹¹⁶ The studies were performed several months after hemorrhage to minimize changes caused by acute hemorrhage. The only difference in those with hemorrhage was slower arterial filling with contrast material, thus suggesting high feeding arterial pressure. Mean feeding arterial pressure was confirmed to be an important factor in the pathophysiology of AVM hemorrhage by the Columbia AVM study group.⁵⁶ This factor was independent of size and location; however, mean arterial pressure was not assessed in patients with small AVMs. Patients may also be predisposed to a hemorrhagic manifestation depending on which artery is feeding the nidus. Various arteries have been implicated, including perforating arteries^12,^56,⁷⁷ and the vertebrobasilar trunk.¹²

Many AVMs that rupture do so from the venous drainage system. Recent studies have focused on the venous drainage pattern of AVMs and its importance in the pathophysiology of hemorrhage. Contributing features include deep venous drainage, often with accompanying stenosis and occlusion, the number of draining veins, and turbulent venous flow, perhaps leading to enhanced platelet aggregation and thrombosis.⁵⁶

Impaired venous drainage may lead to a higher risk for hemorrhage as a result of increased pressure transmitted through the shunt. This was evaluated mathematically ¹¹⁷ and clinically suggested by several studies.^58,^77,^118,¹¹⁹ Vinuela and colleagues found that 21 of 41 patients with intracranial hemorrhage secondary to deep AVMs had vessel wall irregularity, stenosis, or occlusion in the deep venous system.¹¹⁹

Deep venous drainage has frequently been shown to increase the risk for a hemorrhagic AVM manifestation.^12,^56,^77,^94,^96,^99,^102,^105,¹¹⁸ Because many lesions with deep venous drainage are noncortical and unlikely to cause seizures, some believe that hemorrhage is the only initial manifestation possible. The Columbia AVM study group examined a large number of physiologic indices in 449 patients to determine the relationship of AVM hemorrhage and venous drainage, as well as other parameters.¹²⁰ Multivariate analysis revealed that size and deep venous drainage were independent risk factors for bleeding, thus arguing against beliefs that deep venous drainage increases risk because of deep location or size. In fact, even large cortical AVMs with deep venous drainage were more likely to hemorrhage. Furthermore, equal draining pressure was found in both the deep and superficial AVMs that bled. From this study, four groups of patients emerged on the model’s prediction of the probability of intracerebral hemorrhage in terms of size and venous drainage: (1) small AVM size and the presence of deep venous drainage only had a probability of 96%, (2) medium or large AVMs and deep venous drainage only had a probability of 80%, (3) small AVM size and superficial venous drainage had a probability of 69%, and (4) medium or large AVMs with superficial venous drainage had a probability of 29%. In a later study of 622 patients in the Columbia AVM database, independent predictors of a hemorrhagic manifestation included AVM size, deep brain location, border zone location, exclusive deep venous drainage, and the presence of associated arterial aneurysms.⁹⁶ As noted earlier, in a prospective analysis of risk for hemorrhage, there was a trend toward deep venous drainage being an independent risk factor in several studies.^93,⁹⁶

A single draining vein was predictive of hemorrhagic risk in some studies ^102,^109,¹¹⁸ but was not confirmed by others.^12,^65,⁹² This may reflect the fact that smaller AVMs are more likely to have single draining veins.⁵⁶ The presence of, but not the size of, fragile venous aneurysms was significantly associated with risk for hemorrhage in two studies,^77,¹²¹ but this finding, too, is controversial.^12,^65,¹¹⁸

Although used primarily as a surgical risk scale, the Spetzler-Martin scale has also been used in considering the risk for hemorrhage.¹²² In patients with grade IV or V AVMs, the annual pretreatment hemorrhage rate was 10.4% per year, 13.9% in those with hemorrhage and 7.3% in those without hemorrhage at initial evaluation. After treatment, the overall risk was 6.1% per year for all patients, 5.6% for those initially seen with hemorrhage and 6.4% for those without hemorrhage, thus suggesting that the risk for hemorrhage after treatment was either less than or equal to the pretreatment rate.¹²³

Seizures

Approximately 15% to 35% of patients with AVMs will have a seizure as the first symptom.^46,^48,^49,^54,^55,^57,^65,⁹³ Seizures may be the result of a mass effect with cortical irritation or flow characteristics leading to steal, ischemia, and neuronal damage or be due to associated hemorrhage and gliosis.^49,¹²⁴ In a series by Morello and Borghi, 35% of patients were initially evaluated because of seizures.⁵⁷ Of these, 57% had seizures alone and 43% had seizures associated with hemorrhage. The majority of these patients had less than six seizures per year. The seizures are most commonly focal (simple or partial complex) but may also be generalized.

Some risk factors have been delineated for predicting the occurrence of seizures in patients with AVMs. Ninety percent of the seizures were supratentorial. Superficial, large (>6 cm) AVMs ^57,⁸⁶ and those in a frontal or temporal location^46,⁵⁷ were more likely to be manifested as seizures. Turjman and colleagues identified six angioarchitectural features predictive of the development of seizures in a multivariate analysis of 100 patients with AVMs.¹²⁴ The predictive features included cortical location, feeding by the middle cerebral artery, a cortical feeding artery, absence of aneurysms, presence of varices in the venous drainage, and the association of a varix in the absence of an intranidal aneurysm. Interestingly, size and high-flow shunting failed to reach statistical significance and were not predictive. The risk for future hemorrhage in those initially seen with seizures is probably less than in those with hemorrhage.⁹³

Headache

Headaches are a common complaint in patients with AVMs, even in the absence of hemorrhage. Approximately 15% of unruptured AVMs are initially manifested as headache.⁶⁵ The headache is typically located hemicranially (ipsilateral or contralateral to the lesion) or in the occipital region, and the quality is similar to migraine. Interestingly, the incidence of migraine headache in patients with AVMs does not exceed that of the general population¹²⁵ and therefore makes it difficult to know which headaches are related to the AVM unless treatment is performed. In fact, successful embolization of AVMs in patients with headache has been reported. The pathologic etiology of the headache is hypothesized to relate to long-standing meningeal artery involvement and recruitment of blood supply by the AVM. Occipital AVM location may be a risk factor for headache.⁴⁹

Neurological Deficit

Less than 10% of patients will initially have transient, permanent, or progressive focal neurological deficits not ascribed to hemorrhage or seizure.^51,^54,^65,¹²⁶ Progression of neurological dysfunction may be a result of the long-term effects of recurrent small hemorrhages, mass effect of the AVM, hydrocephalus, or ischemic complications and “steal.” Steal is the term used to describe blood flow away from a region of the brain toward the shunt of the AVM. This flow may cause hypoperfusion, ischemia, and symptoms in the region from where the blood was “stolen.” Consequently, steal may lead to focal or more global neurological deficits. Learning disorders have been documented in 66% of adults with AVMs, thus suggesting that functional brain deficits may be present before other clinical signs.^49,¹²⁷

Risk factors for progressive neurological deficits include size ¹¹⁰ and shunt characteristics.¹²⁸ Large AVMs are more likely to result in neurological symptoms attributed to steal. Spetzler and coworkers hypothesized that the low feeding artery pressure associated with large AVMs provides low perfusion pressure to the surrounding cortex, thereby producing a relative ischemia.¹¹⁰ Patients with progressive deficits are also more likely to have extremely fast shunts with higher flow volumes noted on transcranial Doppler ultrasound.¹²⁸ In contrast to these studies, Mast and colleagues believe that steal is a rare phenomenon.¹²⁶ They found no relationship to size or flow velocities in patients with focal deficits.¹²⁶ Furthermore, although positron emission tomography has shown reduced cerebral blood flow around AVMs, oxygen extraction fractions remain normal, which suggests that surrounding parenchyma compensates for the reduced flow.¹²⁹

Outcome

Morbidity and Mortality

The mortality associated with the initial symptomatic hemorrhage is 4% to 29%.^37,^48,^55,^60,^63,^65,^93,¹³⁰^–¹³² More recent data, however, suggest that mortality and morbidity may be somewhat lower than previously reported.^96,^133,¹³⁴ In a study published in 1988, the 30-day mortality with the initial hemorrhage was 29%, and long-term disability was noted in 23% of survivors.⁶⁵ In a recent series of 678 patients, 89 hemorrhages occurred during follow-up; 5 patients (6%) died of the hemorrhage and 35% had significant functional impairment.⁹³ There were no specific demographic or angioarchitectural predictors of a poor outcome. In a study of 622 patients monitored prospectively, 39 hemorrhages occurred and caused just one death and a median Rankin score of 2 in survivors.⁹⁶ In a separate report of the morbidity and mortality in 241 patients initially seen with hemorrhage in the Columbia AVM database, the median Rankin scale score was 2 and the neurological deficit was mild. There was no increase in morbidity if hemorrhage recurred during follow-up.¹³⁵ Parenchymatous AVM hemorrhage was associated with higher stroke morbidity (odds ratio [OR], 2.9) than nonparenchymatous hemorrhage was but had a better outcome than did non–AVM-related parenchymatous hemorrhage (P < .0001). A population-based study from Australia noted that the initial hemorrhage carried a 4.6% case fatality rate.

Many other studies have reported complete recovery or just mild disability in more than 50% of patients after an initial hemorrhage.^131,^55,^130,¹³⁶ Outcome may be dependent on the location and type of hemorrhage, but not on the size of the AVM.^54,⁵⁵ Patients were more likely to have neurological deficits if the hemorrhage was parenchymal rather than subarachnoid or intraventricular.^55,^131,¹³⁶ In addition, hemorrhage in the posterior fossa was associated with a higher mortality rate. Mortality was reported to occur in 66.7% of patients with posterior fossa hemorrhage in one study.⁵⁴

In a long-term follow-up study from the Australian population-based study, patients were monitored for a mean of 10.1 years. Assessment of 209 survivors with the Oxford neurological disability scale showed that 74.1% were grade 0 to 2, 17.2% were grade 3, and 9.5% were grade 4 or 5.⁴³

Recurrent Hemorrhage

Recurrent hemorrhage is noted in 23% to 44% of patients, and the risk may be higher in the first year after the initial hemorrhage.^54,^55,⁹³^–⁹⁵ ^,137 Because the average time from initial evaluation to hemorrhage is 7 to 12 years,^55,^63,⁶⁵ few studies have assessed the risk for recurrent hemorrhage. Graf and colleagues observed 134 patients with ruptured AVMs for a median time of 2 years.⁵⁵ Twenty-four percent of the patients had recurrent hemorrhage during the follow-up period, for an annual rebleeding risk of 6% at 1 year and 2% thereafter. Although this study found size to be predictive of initial hemorrhage, size was not a predictor of subsequent hemorrhage.

Of 315 patients in one study, 196 were seen because of an initial hemorrhage.⁹² Recurrent hemorrhage was found in 44%, for a recurrence risk of 7.45% per year. Four AVM groups were constructed to predict risk for hemorrhage on the basis of three significant variables in the multivariate analysis. The low-risk group had no history of bleeding, more than one draining vein, and a compact nidus. The intermediate- to low-risk group had no history of bleeding and one draining vein or a diffuse nidus (or both). The intermediate- to high-risk group had a history of bleeding, more than one draining vein, and a compact nidus. The high-risk AVM group had a history of bleeding and one draining vein or a diffuse nidus (or both). Annual rates of hemorrhage were 1.31% in the low-risk group, 2.4% in both intermediate groups, and 8.99% in the high-risk group.

Other studies have confirmed that previous hemorrhage is a risk factor for subsequent hemorrhage.* The Columbia AVM study group found an 18% per year risk for hemorrhage in patients with previous hemorrhage versus 2% per year for all others.¹²⁰ In a prospectively collected database from Toronto, the annual risk for hemorrhage was 4.61% for the entire cohort and 7.48% for those with hemorrhage.⁹³ Hemorrhage was an independent risk factor for future hemorrhage in a multivariate analysis (HR, 2.15). In a study of 305 patients, the annual risk for hemorrhage was 6.84% per year in those with hemorrhage as opposed to 3.12% per year in those without hemorrhage.⁹⁵

Mortality did not necessarily increase with subsequent hemorrhagic episodes.⁵⁴ The mortality associated with recurrent hemorrhage has been estimated to be 12% to 15%.^46,^51,^60,¹³⁰ In a population-based study with small numbers, 4 of 17 patients had recurrent hemorrhage, with mortality reaching 50%.³⁷

Seizures

Information on risk for the de novo development of seizures over time and the outcome of patients with epilepsy treated conservatively is scant. Crawford and coworkers monitored 245 patients with symptoms other than seizures for a median period of 7 years.⁸⁶ Ninety-six patients were treated surgically, and these patients had a higher risk for the development of epilepsy. The 20-year risk was 57% in the surgical group and 19% (<1% per year) in the conservatively treated group. In three fourths of the patients in whom seizures developed, they did so within 2 years of treatment. The surgically treated group was more likely to experience seizures if they were younger at diagnosis and if the AVM was in the frontal or parietal lobes. Of the conservatively treated group, those with hemorrhage at a younger age and those with temporally located AVMs were most likely to have seizures. The size of the AVM did not play a significant role. In none of the patients who had nonhemorrhagic focal neurological deficits or were initially asymptomatic did seizures develop. The reason that the surgically treated patients have a higher risk for epilepsy is unclear. It may be a result of selection bias inasmuch as most patients who were selected for surgery had hemorrhage on initial evaluation and the lesions were located superficially.

Piepgras and colleagues studied 280 patients over a period of 7.5 years.¹¹² Fifty-six percent of the patients were initially seen with hemorrhage, 25% had seizures, and 19% had other symptoms. Sixty-nine patients had both seizures and hemorrhage. Patients were treated surgically mainly to reduce the risk for bleeding, not specifically because of seizures. Of the surviving 136 patients who had no history of preoperative seizures, 94% were seizure free and new seizures developed in just 6% after surgery. Of patients with preoperative seizures, 83% were seizure free after surgery (50% were taking antiepileptic medications) and 17% had intermittent seizures, although the majority were improved. The difference between this and older studies probably relates to improved diagnostic and surgical techniques.

In conservatively treated patients, the majority are well controlled with antiepileptic medications.^46,¹³⁹ Murphy found that only 16% of conservatively treated patients were incapacitated by their seizures.¹³⁹ Similarly, Perret and Nishioka found that just 4 of 39 patients treated with medications alone were unable to work.⁴⁶

Radiographic Features/Lesion Behavior

AVMs may increase in size, remain stable, decrease in size, or completely regress or thrombose over time.^16,^58,^60,^111,¹⁴⁰^–¹⁴⁵ Minikawa and associates found that the patients with enlarging AVMs had their first angiogram at a young age (0 to 11 years).¹⁴¹ Subsequently, three of the four patients in that study rebled. The AVMs that regressed were relatively small and fed by few feeding arteries.

Spontaneous regression can be acute or gradual and occurs in approximately 2% to 3% of patients with AVMs.¹⁷ Causes include low flow; occlusion of feeding vessels by atherosclerosis, emboli, or dissection; or regression in association with hemorrhage because of mass effect or vasospasm.^16,^140,^142,¹⁴⁶ Thrombosis is thought to be protective. Thrombosed AVMs rarely cause hemorrhage, although the risk for seizures remains.^48,¹⁴⁴ Caution is suggested in concluding that an AVM is thrombosed after hemorrhage with negative angiographic findings. Transient regression after acute hemorrhage may be secondary to vasospasm and not thrombosis.¹⁵ Others have reported recanalization after thrombosis.¹⁴⁷

Arteriovenous Malformations And Aneurysms

The association of aneurysms with AVMs has been established.^46,^55,^{75–83,85,93,148–152} Numerous studies have attempted to classify the types of aneurysms associated with AVMs, and in general, each classification system takes into account the distance and flow relationship to the AVM. Aneurysms may be flow related, intranidal, or unrelated.⁷⁷ Flow-related aneurysms refer to saccular aneurysms arising along the course of arteries that eventually supply the AVM. A proximal flow-related aneurysm is one located on the supraclinoid internal carotid artery, the circle of Willis, the middle cerebral artery up to the bifurcation, the anterior communicating artery up to the anterior cerebral artery, or the vertebrobasilar trunk. All flow-related aneuryms beyond these locations are distal flow-related aneurysms.⁷⁷ The distal type generally consists of those on the main feeding artery of the AVM, and they are also known as pedicle artery aneurysms.⁷⁶ Some further break the classification down to differentiate whether the feeding artery is deep or superficial.⁸² Intranidal aneurysms are aneurysms that lie within the AVM nidus. Unrelated or dysplastic aneurysms are those remote to the AVM.

The pathogenesis of aneurysms associated with AVMs is unknown. Three theories have evolved. The first proposes that the association of aneurysms and AVMs is coincidental without a causal relationship. This is probably not the case because the prevalence of aneurysms in patients with AVMs exceeds that of the general population.^83,⁸⁵ The second theory hypothesizes that both AVMs and aneurysms are congenital vascular abnormalities. The third and more favored theory is that aneurysms result from hemodynamic factors caused by the increased flow through the AVM.⁸¹ In favor of this hypothesis is the increased occurrence of aneurysms on the feeding arteries,⁸⁵ evidence of endothelial changes and internal elastic lamina rupture proximal to the AVM shunt, and the observation that distal flow-related aneurysms often regress with definitive treatment of the AVM. Brown and associates, however, found that even patients with low-shunt AVMs harbored pedicle aneurysms.⁸⁵ They further hypothesized that the pathogenesis of aneurysms may be related to the size of the AVM inasmuch as all aneurysms associated with small AVMs were found at atypical locations. Thus, the pathogenesis may relate to a combination of factors, such as an underlying vascular defect, hemodynamic stimuli, vasoactive substances, and locally generated growth factors.

Depending on the series, aneurysms are present in 2.7%⁷⁸ to 34%¹⁵² of patients with AVMs.^46,^55,^75,^76,^{78–85,93,150–152} The average figure of 8% to 10% exceeds the 2% prevalence of aneurysms in the general population.⁸³ The mean age at initial evaluation may be slightly older than that of the general AVM population. Miyasaka and coworkers found the mean age at diagnosis to be 41 years in those with aneurysms versus 31 in those without.⁷⁹ The male-to-female ratio is similar to that of the entire AVM population.⁸⁰ Males may be more likely to harbor flow-related and intranidal aneurysms, whereas females may be more likely to have associated dysplastic or remote aneurysms.⁸⁰ As with AVMs in isolation, AVMs with associated aneurysms are typically manifested as hemorrhage.^82,⁸⁴ Of 39 patients with 64 aneurysms, Cunha and coauthors reported intracerebral hemorrhage as the initial finding in 63%.⁸² Forty-six percent of these hemorrhages were secondary to the aneurysm, 33% were related to the AVM, and in 21%, the site of hemorrhage was unclear. Redekop and colleagues reported that 36% of a total of 632 patients with AVMs were initially seen with hemorrhage.⁷⁷ In those with intranidal aneurysms, hemorrhage occurred in 72%. Similar to the overall group, 40% of patients with flow-related aneurysms experienced hemorrhage. Of patients with an AVM and flow-related aneurysm, 12 of 29 (41.4%) bled from the aneurysm (5 distal, 7 proximal), 15 of 29 (51.7%) bled from the AVM, and 2 of 29 were undetermined.

Aneurysms may be multiple in 30% to 50% of patients with AVMs and aneurysms.^79,^82,^84,¹⁵¹ The majority of studies reveal no difference in the size of the AVM and the type, number, and size of the associated aneurysm.^77,^82,⁸⁵ Only one study found that larger AVMs were more likely to be associated with aneurysms.⁷⁹ The size of the associated aneurysm ranged from 3 mm to 2.5 cm and averaged approximately 7.2 to 8 mm.^77,⁸⁵ The frequency of aneurysm type varied according to the study, with the majority being flow related.^46,^75,^77,^79,^80,^82,¹⁵¹ The frequency of distal feeding artery aneurysms ranged from 37% to 69%.^46,^75,^77,^79,¹⁴⁹ ^,151 Remote aneurysms are less common and found in 1.6% to 43% of patients.^46,^77,⁷⁹ ^,80 ^,82 ^,151 Intranidal aneurysms represent approximately 20% of aneurysms.⁷⁷ This figure, however, is variable and depends on the type of study performed. Superselective angiography is more likely to detect intranidal aneurysms but was not uniformly used in these studies. Typical locations (at artery bifurcations) were found in association with high-shunt malformations.⁸⁵ Atypical locations were noted in association with high- or low-flow and high- or low-shunt AVMs.

Several studies reported an increased risk for hemorrhage when an AVM is associated with an aneurysm.^77,^85,^93,^148,^151,¹⁵² Brown and collaborators studied 16 patients with 26 aneurysms associated with unruptured AVMs.⁸⁵ The risk for hemorrhage in a patient with a coexisting aneurysm was 7% per year at 5 years. In those without a coexisting aneurysm, the rate was 3% per year at 1 year and then declined to 1.7% per year at 5 years. In a multivariate analysis of 678 people with AVMs, the risk for hemorrhage in those with an aneurysm was 6.93% per year versus 3.99% per year in those without an aneurysm. In a multivariate analysis, there was a trend toward significance of the presence of an aneurysm as a predictor of hemorrhage (HR, 1.59; P = .07).⁹³ The association of hemorrhage at initial evaluation and coexisting aneurysms has not been consistently noted.¹⁵²

Risk factors for hemorrhage in patients harboring both lesions have been evaluated. Neither the size of the AVM nor the size of the aneurysm was important, although the number of patients was small. Several studies have reported that patients with associated intranidal aneurysms are at increased risk for an initial manifestation of hemorrhage and subsequent rebleeding.^77,^148,¹⁵¹ All the patients in a study of Marks and colleagues with intranidal aneurysms were initially seen because of hemorrhage, as opposed to 58% of those without such aneurysms.¹⁴⁸ As mentioned previously, 72% of the patients with intranidal aneurysms in the study by Redekop and associates had hemorrhage initially.⁷⁷ Thirteen of the patients with intranidal aneurysms in this study could not be treated surgically. Fourteen subsequently hemorrhaged over a period of 143 patient-years, for an annual rate of hemorrhage of 9.8%. This was compared with a risk of 5.3% per year for flow-related aneurysms that could not be treated by surgery. They concluded that patients with intranidal aneurysms had a high risk for rupture and a hemorrhagic manifestation, as well as a high risk for rebleeding. In addition, for flow-related aneurysms, patients had an equal chance of bleeding from the aneurysm as from the AVM. In the Columbia AVM databank, 117 of 463 (25%) prospectively identified patients had a concurrent aneurysm, including 54 with aneurysms on the feeding arteries, 21 with aneurysms in an intranidal location, 18 with aneurysms unrelated to the AVM, and 24 with more than one aneurysm type.¹⁵¹ In multivariate analysis, the presence of a feeding artery aneurysm was an independent predictor of a hemorrhagic manifestation (OR, 2.11), but such was not the case with intranidal or unrelated aneurysms. Pedicle artery aneurysms were considered a risk factor for hemorrhage in other small studies, but no multivariate modeling was available.^76,⁸⁴ Perata and associates demonstrated the significance of pedicle artery aneurysms in 4 patients.⁷⁶ They hypothesized that short perforators (thalamoperforate or lenticulostriate) are exposed to higher pressure and flow rates and are more likely to undergo aneurysm formation and rupture during the development of an aneurysm. Batjer and coworkers confirmed the danger of pedicle or feeding artery aneurysms. Of 9 patients with intracranial hemorrhage, 7 bled from the aneurysm.⁸⁴ All 7 aneurysms were atypical aneurysms located on the major feeding vessel.

Aneurysms may increase in size, remain the same, or regress over time. With definitive AVM treatment, distal flow-related aneurysms are the most likely to regress.⁷⁷ Proximal aneurysms on the circle of Willis or remote aneurysms are unlikely to change.^81,⁸³ Furthermore, in patients with untreated AVMs, new aneurysms may arise over time.

Arteriovenous Malformations in Pregnancy

Cerebral hemorrhage during pregnancy is the third leading cause of maternal death from nonobstetric causes. Maternal mortality as a result of intracranial hemorrhage has been reported to be as high as 40% to 50%.¹⁵³ Intracerebral hemorrhage is most commonly caused by eclampsia, but it may also be due to AVM or aneurysm rupture, venous thrombosis, hypertension, or choriocarcinoma.¹⁵⁴ Although referral sources may bias the frequency, estimates of cerebral hemorrhage secondary to AVM or aneurysm rupture during pregnancy range from 0.01% to 0.05%.¹⁵⁵

Controversy exists whether pregnancy increases the risk for bleeding from an AVM.^53,^98,¹⁵³^–¹⁵⁸ Robinson and coauthors reported a 10% risk for subarachnoid hemorrhage from AVMs in nonpregnant females of childbearing age and an 87% risk in association with pregnancy.⁹⁸ Hemorrhage “associated with pregnancy” was defined as pregnancy within 2 years after diagnosis, hemorrhage during pregnancy or delivery, or hemorrhage less than 2 years after pregnancy. Although Robinson and colleagues noted that there was an 87% risk for hemorrhage during pregnancy, the data actually indicate that among AVMs that are diagnosed during pregnancy, 87% will hemorrhage. With conservative treatment, the fetal death rate was 26% and 29% of mothers died or were disabled.

In contrast to these findings, Horton and coworkers studied 451 women (540 pregnancies) with AVMs, 17 of whom had pregnancies complicated by hemorrhage.¹⁵⁶ There was no difference in the hemorrhage rate between the months of pregnancy and all other months of the women’s lives. However, the pregnancy term used was somewhat arbitrary, and a definition of pregnancy that was slightly shorter would have led to a “statistically significant” hemorrhage difference. Patients with no history of hemorrhage had a 3.5% risk of bleeding during pregnancy, which is similar to the risk of bleeding during the nongravid months in the same patients with unruptured AVMs. The risk increased to 5.8% in patients with a history of hemorrhage. Of 17 pregnancies complicated by hemorrhage, 4 resulted in maternal disability and 3 in fetal deaths. This study selected patients chosen for proton beam therapy, thereby potentially eliminating early deaths, severely disabled women, or vascular malformations that would not have been proton beam candidates.

Several studies evaluated risk factors for hemorrhage from AVMs during pregnancy. When compared with patients with hemorrhage from aneurysms, patients with AVMs were more likely to be younger,^153,^154,¹⁵⁹ to have borne fewer children,¹⁵⁹ and to initially have been seen commonly between 15 and 20 weeks.¹⁵⁹ Dias and Sekhar reviewed 154 cases of hemorrhage during pregnancy as a result of ruptured AVMs or aneurysms.¹⁵³ They found that 77% of the hemorrhages were due to aneurysm rupture and 23% to AVM rupture. Patients with AVM rupture were more likely to be younger than those with aneurysms, but in this study no difference in parity or gestational age was found. It was also noted in this¹⁵⁹ and other studies that there is an increased frequency of hemorrhage with advancing gestational age, possibly caused by a combination of hemodynamic, coagulation, and hormonal factors.

Recurrent hemorrhage is not uncommon and increases the rate of mortality.¹⁵⁵ Few patients suffer initial AVM rupture during labor and delivery^53,¹⁵⁵; however, rebleeding may commonly occur during this time.¹⁵⁹ Subsequent pregnancies also carry an increased risk for rebleeding.^98,¹⁵⁴

Arteriovenous Malformations in Children

In large clinical series, children (<20 years of age) accounted for 15% to 33% of all patients initially seen with AVMs.^46,^55,¹³⁰ Symptomatic AVMs during childhood are rare, although the most common cause of intracerebral hemorrhage in children is AVM rupture.^34,¹⁶⁰ In the only population-based study, 3 of 20 patients with hemorrhage because of a ruptured vascular malformation were younger than 20 years.³⁷

Hemorrhage is by far the most common initial manifestation of the lesion (50% to 79%),¹⁶¹^–¹⁶⁶ followed by seizures (8% to 25%)^161,¹⁶³^–¹⁶⁵ and congestive heart failure (18%).^162,^167,¹⁶⁸ Children with initial hemorrhage may be at heightened risk for recurrent hemorrhage.⁹⁵ Symptoms of congestive heart failure from high left-to-right shunting through the AVM and hydrocephalus predominate in newborns.¹⁶⁹ Hemorrhage and seizures occur more frequently in children older than 2 years.^46,^164,¹⁶⁵ In addition to these symptoms, patients may rarely exhibit mental deterioration or progressive neurological symptoms secondary to steal phenomena.¹⁷⁰

D’Aliberti and colleagues compared the clinical and angiographic features of 19 children (mean age, 11) and 120 adults with AVMs.¹⁷¹ The main differences between the two groups were sex distribution and the size, depth, location, and complexity of the AVM. AVMs were more common in male pediatric patients. AVMs in pediatric patients tended to be smaller and located superficially. Even though 68% of pediatric patients with AVMs suffered hemorrhage, only 6% had deep venous drainage.

AVMs can grow in size, and this is usually most prominent in the pediatric age group. In addition, AVM recurrence after resection and negative postoperative angiographic findings has been reported.^161,^165,¹⁷² Hdlaky and associates hypothesized that persistent microshunts not evident on early postoperative angiograms may increase over time and predispose patients to recurrent hemorrhage.¹⁶¹

Morbidity and Mortality

Even with newer diagnostic techniques and therapeutic interventions, the mortality associated with AVM hemorrhage in children remains high. Clinical studies of pediatric patients with hemorrhage report a 6.5% to 35% mortality rate.^162,^164,^165,¹⁷³^–¹⁷⁵ Hemorrhage location and volume predict mortality.^165,¹⁷⁴ The mortality associated with posterior fossa hemorrhage was far greater than that with supratentorial lesions. In a study of cerebellar AVMs in children, 6 of 17 (35%) patients died.¹⁷⁴ In another series, mortality was 67% in 9 patients with brainstem AVMs.¹⁷² The higher mortality in pediatric series may relate to the distribution of posterior fossa AVMs in some studies. In addition, the cumulative morbidity and mortality caused by AVMs in children is high because of the prolonged risk period for potential hemorrhage. The risk for rebleeding has been reported to be between 22% and 29%,^165,¹⁷⁵ and it can increase mortality.¹⁷⁵

Despite high mortality rates, children are more likely than adults to improve after intracerebral hemorrhage caused by AVM rupture.¹⁷¹ D’Aliberti and colleagues found that 89% of children versus 79% of adults had good recovery based on the Glasgow outcome score. Similarly, another study found satisfactory outcomes in 81% of children with AVMs observed over a mean period of 8.5 years.¹⁶¹ In this study, 72% of the patients with hemiparesis recovered fully or were left with only mild residual effects. Neonates with congestive heart failure, however, have a very poor prognosis. In the series of seven neonates described by Melville and coworkers, six died in the first 17 days or at the time of intervention. Although some deaths were due to neurosurgical complications, the poor prognosis may also relate to residual shunting after treatment.¹⁶⁹

Cavernous Malformation

Synonyms: cavernous angioma, cavernous hemangioma, cryptic vascular formation,* occult vascular formation,* angiographically occult vascular malformation.*

Definition and Pathogenesis

CMs are well-circumscribed, multilobulated, angiographically occult vascular malformations. Pathologically, they are composed of sinusoidal vascular channels (caverns) lined by a single layer of endothelium. The caverns are separated by collagenous stroma devoid of elastin, smooth muscle, or other mature vascular wall elements. The lack of intervening brain parenchyma is a characteristic pathologic marker. Within the lesion, hyalinization, thrombosis with varying degrees of organization, calcification, cysts, and cholesterol crystals are common and lead CMs to be likened to mulberries. The surrounding parenchyma consistently exhibits evidence of previous microhemorrhage, hemosiderin discoloration, and hemosiderin-filled macrophages. A gliomatous reaction of surrounding parenchyma is characteristic and may form a capsule around the lesion.^3,^4,^17,¹⁷⁶^–¹⁸⁰ Electron microscopy reveals a defect in endothelial tight junctions, which has been hypothesized to be the reason that CMs tend to leak.¹⁸¹ There is no difference in pathology between sporadic and familial forms.¹⁸²

CT is 70% to 100% sensitive but less than 50% specific in detecting CMs.^178,¹⁸³^–¹⁸⁵ The typical appearance is a well-circumscribed nodular lesion with calcification. Hemorrhage and cystic components may also be present, and faint enhancement may be seen with the administration of contrast material. MRI is the most sensitive in detecting CMs,¹⁸⁶ specifically, gradient echo¹⁸⁷^–¹⁸⁹ and susceptibility-weighted imaging.¹⁹⁰ CMs appear as well-defined, lobulated lesions on MRI with a central core of mixed signal intensity surrounded by a rim of signal hypointensity.^178,¹⁹¹^–¹⁹⁷ The mixed signal reflects the lesion’s behavior. The decreased rim of T2 intensity is related to repeated subclinical intralesional and perilesional hemorrhage. Low T2 signal within the lesion reflects intralesional calcification. The hyperintensity reflects acute and subacute hemorrhage in different stages. Residua of previously expanded cisterns that have involuted with thrombus organization and resolution are reflected by the presence of cysts. Some studies have further characterized the MRI appearance of CMs for clinical study (Table 384-2).^187,¹⁸⁹

TABLE 384-2 Categories of Cavernous Malformation Based on Magnetic Resonance Imaging Characteristics

Although MRI is more specific than CT, the MRI appearance is not always pathognomonic. The differential diagnosis should also include thrombosed AVM, calcified tumor (e.g., oligodendroglioma), granuloma, or infectious and inflammatory nodules.^178,^184,^192,^194,^196,¹⁹⁸^–²⁰⁰ The absence of systemic neoplasm and surrounding edema and the presence of multiple lesions, calcification, and ossification offer clues distinguishing neoplasms from CMs. A more typical MRI appearance, family history, and multiple lesions may favor CM. Distinguishing CM from other cryptic vascular malformations on imaging studies and even by pathologic evaluation may be more difficult. The designation cryptic or occult vascular malformation refers to angiographically occult vascular malformations. They include CMs but may also include thrombosed AVMs, mixed lesions, or rarely, venous angiomas. The majority of cryptic malformations are CMs.^201,²⁰² In a study of 34 consecutive angiographically occult malformations undergoing surgical excision and pathologic examination, 21 were CMs, 3 were AVMs, 3 were VMs, 2 were capillary telangectases, and 5 were mixed lesions.¹⁷ One study attempted to find radiographic features that distinguish angiographically occult malformations. Vanefsky and associates found that cryptic CMs were more likely to have hemosiderin rings and absence of edema.²⁰³ Rigorous subdivision may not be necessary because angiographically occult vascular malformations of any type may not have a distinctive natural history.^17,^145,^198,²⁰⁴