Arteriovenous malformations of the brain (AVMs) are a major cause of stroke in young, healthy individuals and present multiple diagnostic and therapeutic challenges, particularly in the acute setting. Although the flow hemodynamics, biology, epidemiology, and natural history of AVMs have been extensively studied, little data have been published on AVM surgery in the acute setting, and acute surgery has been claimed to possibly increase the risk of persistent neurological deficits. Although it is usually preferable to defer AVM surgery for a few weeks or months, acute surgical (open and endovascular) management is essential in specific clinical and radiological settings.
Arteriovenous malformations of the brain (AVMs) are a major cause of stroke in young healthy individuals and present multiple diagnostic and therapeutic challenges, particularly in the acute setting. They present as complex tangles of blood vessels with arterial blood flowing directly into the veins without interposed capillary networks and can be associated with other vascular lesions, such as aneurysms or arteriovenous fistulae, thus, adding to the complexity of patient management. At least 50% of patients with AVM present with hemorrhage with the resulting significant morbidity and mortality. Although the flow hemodynamics, biology, epidemiology, and natural history of these lesions have been extensively studied, little data have been published on AVM surgery in the acute setting, and acute surgery has been claimed to possibly increase the risk of persistent neurological deficits. Although it is usually preferable to defer AVM surgery for a few weeks or even months to allow brain swelling to decrease and gliosis to better delineate the lesion to be excised, acute surgical (open and endovascular) management is, nevertheless, essential in specific clinical and radiological settings. Large life-threatening hematomas, hydrocephalus, and the presence of an accessible and clear rupture site are all potential indications for acute procedural intervention. For a life-threatening hematoma related to AVM hemorrhage, it is generally advised to remove sufficient hematoma to achieve a slack brain without resecting the AVM. Given the differing acute natural history risk related to rehemorrhage, distinguishing between AVM hemorrhage and hemorrhage from an associated pedicle or circle of Willis aneurysm is paramount. Clearly identifying an interventionally accessible intranidal aneurysm may also be an indication for early embolization. Although experience with early intervention for ruptured AVMs is increasingly discussed in the literature, no guidelines have yet been published, and available data are mostly derived from anecdotal case series and limited to level V evidence.
Epidemiology of AVM Hemorrhage
AVMs are the most common cause of nontraumatic cerebral hemorrhage in patients aged younger than 45 years. Patients with AVM not only tend to be younger than the rest of the population suffering from cerebral vascular disorders but also seem to have fewer or no medical comorbidities. Hemorrhage in this population can lead to catastrophic outcomes and claim a heavy functional toll on previously healthy young individuals. Thorough knowledge of the natural history, predictors, and presentation of AVM hemorrhage is essential for prudent clinical decision making.
Initial hemorrhagic presentation
Approximately 50% of AVMs present with hemorrhage, with reported rates ranging between 32% and 82%, depending on the series. In stroke registries, around 1% of all strokes are attributed to AVM hemorrhage. Intraparenchymal hemorrhage is most commonly observed, but intraventricular and subarachnoid hemorrhage can also occur. Annual rupture rates among individuals presenting with symptoms other than hemorrhage range between 2% and 4%, with an estimated lifetime risk of hemorrhage of 17% to 90%. It should be noted that most available natural history data are derived from anecdotal case series (level of evidence V). Initial results from The New York Islands AVM study, a prospective population-based database, have been released, but these data are notable for a lack of long-term follow-up. The examination of 284 prospective AVM cases recruited in this database estimated the incidence of first AVM hemorrhage to be approximately 0.51 per 100,000 person-years (95% CI, 0.41 to 0.61) and the prevalence of hemorrhage among diagnosed cases to 0.68 per 100,000 person-years (95% CI, 0.57–0.79). Kondziolka and colleagues suggested a simple model based on life expectancy and the multiplicative law of probability to predict the lifetime risk of hemorrhage in a given individual assuming that, for that individual, the risk of hemorrhage is constant over time. According to their calculations, the lifetime risk of hemorrhage can be estimated using the following formula:
This formula would assume, for example, that for a patient with a life expectancy of 60 years and a yearly risk of hemorrhage of 4%, the lifetime risk of AVM bleed would be the following:
A simpler model assuming a constant 3% yearly risk of hemorrhage can also be used and still maintains a similar sensitivity:
For example, a 15-year-old boy with an AVM would have a lifetime risk of hemorrhage of
Although most studies report clinically relevant AVM ruptures, subclinical episodes of hemorrhage seem to occur more frequently than previously thought. Magnetic resonance imaging (MRI) frequently shows signs of hemosiderin deposits on T1- and T2-weighted sequences, a finding suggesting prior episodes of microbleeding.
Although AVMs are classically reported to carry a lower risk of hyperacute rebleeding compared with intracranial aneurysms, the risk of recurrent hemorrhage seems to increase temporarily after the initial episode and then subsequently renormalize. In a retrospective series of 191 patients harboring AVMs whereby 102 patients presented with hemorrhage and were followed for up to 37 years, Graf and colleagues found that the risk of rehemorrhage increased to 6% during the first year and then renormalized to 2% per year for up to 20 years. Forster and colleagues, in their retrospective series of 150 patients with AVMs, identified 106 patients who presented with acute rupture and followed them for an average of 15 years. After a single episode of bleeding, the risk of rehemorrhage increased to 25% in the 4 years following the episode. That same risk quadrupled after an additional hemorrhage to reach 25% per year. In their retrospective analysis of 43 patients with AVM who survived their first hemorrhage, Fults and Kelly found that 67.4% suffered from an additional bleed. The risk of rehemorrhage was evaluated at 17.9% during the first year following the hemorrhage and declined to 3% per year after 5 years and 2% per year after 10 years. Mast and colleagues provided even higher risks of hemorrhage in their prospective cohort. They followed a series of 281 consecutive patients with AVM whereby 142 presented with hemorrhage for a mean duration of 8.5 months (0.1–96.4). The average annual risk of hemorrhage was determined to be 17.8%, with a 32.9% risk of rerupture during the first year that declined to 11.3% in subsequent years. In comparison, patients who did not present with acute hemorrhage had a 0% risk of bleeding 1 year after presentation and a 2.9% risk for subsequent years. It should be noted, though, that only 20 untreated patients were still being followed 1 year after their hemorrhage. This finding should be contrasted with Ondra and colleagues’ retrospectively analyzed series of 114 untreated patients with AVM that were followed over a duration of approximately 23.7 years. The rate of major rebleed was estimated at 4% per year, which surprisingly showed little difference compared with the 2% to 3% per year classically reported rate.
Risk factors for AVM hemorrhage
Evidence describing predictors of AVM hemorrhage is largely derived from anecdotal case series or nonrandomized cohort studies using historical controls (levels of evidence V and IV). Most studies only retrospectively examine features related to hemorrhage and many convey conflicting results. Radiological parameters predicting hemorrhage can be categorized into (1) morphological, (2) arterial, and (3) venous factors.
AVM nidal size or volume has been inconsistently reported as a risk factor for hemorrhage. Although smaller AVMs may tend to have higher pressure in their feeding arteries and, thus, a higher propensity to bleed, it has been suggested that although large and giant AVMs most frequently present with steal or seizures because of their size, smaller AVMs were more likely to be detected in the context of hemorrhage, with size actually being a confounding factor. A prospective analysis of 73 consecutive patients with Spetzler-Martin grade (SMG) IV and V AVMs estimated a 1.5% annual bleeding risk, a number lower than those classically reported for grades I through III. On the other hand, a prospective study of 390 patients by Stefani and colleagues found that large AVMs tended to bleed more frequently than smaller lesions (odds ratio: 2.5, 1.41–4.35, P <.0001). Similarly, Jayaraman and colleagues suggested, after the retrospective examination of the medical records of 61 patients with SMG IV and V lesions, that the annual hemorrhage risk for these lesions could be as high as 10.4%, a number substantially higher than that reported for all other AVMs. Diffuse AVM morphology has also been suggested as a possible factor for rupture. Intraventricular, periventricular, basal ganglia, and deep brain AVM locations have also been suggested as predisposing factors for hemorrhage.
Second, on the arterial side, increased feeding artery pressure, the presence of intranidal aneurysms, and a perforator origin for arterial supply have all been postulated as potential risks factors for AVM rupture.
Finally, venous predictors of hemorrhage include the presence of deep venous drainage, a single draining vein, draining vein stenosis, or other causes of impaired venous drainage. A deep venous component has also been linked to subsequent episodes of hemorrhage.
Suggested clinical predictors of AVM hemorrhage include patient presentation with seizures ; male sex; increasing age, especially the fourth, fifth, and sixth decades ; and, most importantly, a history of prior hemorrhage. Increasing age has also been linked to the risk of rehemorrhage.
Although individual risk factors for AVM hemorrhage have been suggested, it seems that organizing findings into a personalized risk profile may provide an optimized model for predicting the risk of rupture and rerupture. Olivecrona and colleagues used the history of prior hemorrhage, angiographic findings of a single draining vein, and the presence of diffuse AVM morphology to define a high-risk patient population. They found that these patients had an 8.9% yearly risk of rupture compared with 1% in triple-negative controls. Similarly, in a prospectively collected cohort of 622 consecutive patients with AVM, Stapf and colleagues identified initial hemorrhagic AVM presentation, deep brain location, and exclusive deep venous drainage as individual risk factors for subsequent AVM bleed, with respective hazard ratios of 5.38, 3.25, and 3.25. When these 3 factors were found in combination, the annual risk of hemorrhage increased dramatically to 34.4% compared with 0.9% in individuals who had none of these 3 factors.
Natural History of AVM Hemorrhage
Although considerable information related to cerebral AVM pathophysiology and management is available, surprisingly scarce publications address the issue of morbidity and mortality from repeat bleeds. Most available studies have produced level V data and provide assumptions from the analysis of small anecdotal case series or retrospective chart reviews of patients with AVM.
Mortality from a first hemorrhage is thought to range from 10% to 30%, although lower rates have been reported. Overall morbidity is generally estimated at around 50%, with a long-term permanent disability rate of 10% to 20%. Fults and Kelly, in a retrospective analysis of 131 patients with AVM, discovered a tendency of increasing mortality with subsequent episodes of hemorrhage from 13.6% to 20.7% to 25.0%, after the first, second, and third episodes respectively, although statistical significance was not reached. Svien and McRae also found that the mortality rate from AVM rupture doubled from 3% to 6% after subsequent hemorrhages. This trend was not found in the analysis by Hartmann and colleagues of 115 prospectively enrolled patients with AVM who presented with hemorrhage as part of the Columbia-Presbyterian Medical Center AVM Study Project. In that study, 84% of the patients did not suffer from any neurological deficit or were independent in their daily activities (modified Rankin score of 1) after a single episode of hemorrhage, with only 16% severely disabled (modified Rankin score ≥4). Of the 27 patients who had recurrent hemorrhages, 74% were neurologically intact or independent in their daily activities, with only 4% severely disabled. These findings seem to go against the hypothesis of a cumulative impact of recurrent hemorrhage on prognosis. Interestingly, no fatalities were recorded after either episode. Although this data may seem reassuring, it does go against observations from many other studies and anecdotal observations. The lower morbidity of AVM rupture compared with intracranial hemorrhage attributed to other causes, such as hypertensive or aneurysmal bleeds, may be attributed to several factors. Although small AVMs with high intranidal pressure may lead to extensive bleedings, larger lesions with low-resistance arteriovenous shunts and low feeding artery pressure would tend to cause less damage. AVMs also rarely tend to cause vasospasm after they rupture. Rupture of deep draining veins often results in purely intraventricular hemorrhage. Additionally, bleeding limited to the nidus tends to spare normal brain parenchyma and minimally disrupt functional neuronal tissue. On the other hand, parenchymal AVM hemorrhages have been shown to have the highest rate of associated focal deficits (51.9%), followed by exclusively subarachnoid (41.2%) or purely intraventricular (27.8%) locations. Finally, patients with AVM tend to be younger, a fact that may contribute positively to recovery.
Medical Management of Acute AVM Hemorrhage
Recommendations for secondary intracranial hemorrhage management have been detailed in the American Heart Association–American Stroke Association guidelines. Although a thorough discussion of these recommendations is beyond the scope of this article, the authors briefly present available evidence.
All patients should be admitted in an intensive care unit where they can be monitored for erratic variations in intracerebral and systemic pressures and where ventilator support is available (class I, level of evidence B). Antiepileptic treatment should be administered promptly at the first sign of clinical seizures (class I, level of evidence B) and may be continued for a brief period after the onset of intracerebral hemorrhage because it may reduce the risk of subsequent epileptic episodes, especially in cases of lobar hemorrhage (class IIb, level of evidence C). Fever and hyperthermia of any cause should be appropriately treated, and antipyretics should be administered to prevent additional brain injury because elevated cerebral temperature has been shown to increase ischemic brain damage (class I, level of evidence C). Persisting early hyperglycemia is associated with poor outcome and should be managed with insulin administration if levels go more than 185 mg/dL (class IIa, level of evidence C). Incomplete evidence is available on blood pressure management in the setting of intracranial hemorrhage, but intravenous drug administration to reduce extremely high variations and maintain cerebral perfusion pressure more than 60 to 80 mm Hg is recommended (class IIB, level of evidence C). Medical treatment of elevated intracranial pressure should start with simple measures, such as elevating the patient’s head and appropriate analgesia, but may include sedation, osmotic diuretics, hyperventilation, and cerebrospinal fluid drainage (class IIa, level of evidence B). Early mobilization of clinically stable patients is recommended to prevent deep vein thrombosis and pulmonary embolism (class I, level of evidence C), and hemiparetic/hemiplegic individuals should have intermittent pneumatic compression of their lower limbs (class I, level of evidence B). Subcutaneous heparin should only be administered after cessation of bleeding has been documented (class IIB, level of evidence B).
Acute Surgical Management After AVM Rupture
Early versus delayed surgical removal
Although most intracranial aneurysms are operated on acutely because of the high risk and morbid consequences of early rerupture, surgeons have been classically reluctant to operate acutely on patients with AVM presenting with hemorrhage unless the hemorrhage is aneurismal in origin or the hematoma is life threatening. Reports in the literature suggest that acute surgical removal of ruptured AVMs may be hazardous and lead to persistent and possibly preventable neurological deficits. AVMs are reported to have a lower risk and morbidity from acute rerupture compared with intracranial aneurysms, a fact that lead to recommendations to defer surgery weeks or even months after the initial bleed to allow brain swelling to subside and patients to recover. Ruptured AVMs often result in intraparenchymal clots and brain edema that is most severe during the first days after the hemorrhage, a fact that may require retraction and manipulation of noncompliant brain tissue to improve visualization. Compression of the AVM by the hematoma may also partially mask parts of the lesion that may not be evident using digital subtraction angiography because contrast may not completely fill the nidus causing incongruence between the radiological image and operative surgical anatomy. Waiting 3 to 6 weeks for the hematoma to resolve or liquefy may improve visualization of the AVM, with the resulting hematoma cavity usually creating a well-defined dissection plane between the lesion and normal brain parenchyma. An exception to this principle may be extremely superficial and small AVMs whereby the anatomy necessary for lesion access and removal is straightforward.
Early surgery for intraparenchymal brain hemorrhage can quickly reduce mass effect and potentially spare healthy neuronal tissue from a prolonged exposure to toxic blood degradation products. Kuhmonen and colleagues reported a large series of 45 patients with AVM presenting acutely with hemorrhage who were treated with surgery within 4 days. Although two-thirds of the patients were admitted with a Hunt-Hess score of 4 to 5, 55% had good functional outcome 2 to 3 months after surgery. It was the investigator’s conclusion that aggressive and early surgical management of the lesions with evacuation of the hematoma can lead to favorable outcomes and accelerated rehabilitation compared with the natural history of the disease or delayed surgical management. One should consider, though, that 60% of the AVMs that were treated were classified as SMG I and II. Pavesis and colleagues retrospectively reviewed 27 SMG I and II AVMs that presented acutely with hemorrhage and were surgically treated within 6 days and came to the same conclusions: early surgery with hematoma evacuation provided complete lesion excision, immediate decompression, bleeding control, shorter hospital stay, and favorable functional outcome. Acute surgery for lesions with SMG greater than or equal to III may require further study before any conclusions are reached because the risk of direct surgery increases in that population and because careful planning of multimodal therapy is more likely needed.
All this being said, it should be noted that instances of ultra-early rebleeding have been reported. In cases whereby the origin of bleeding can be determined with certainty (eg, obvious intranidal or prenidal ruptured aneurysms), early embolization can provide security against rehemorrhage. Microsurgical or endovascular repair of a proximal saccular aneurysm should also be considered early when indicated.
Perioperative anesthetic considerations
Although AVM surgery is usually not emergent and allows for the optimization of patient parameters and careful operative planning, acute resection often requires strict blood pressure control, placing patients in a barbiturate-induced coma with vasopressor administration in cases of myocardial depression as well as invasive monitoring of systemic and intracranial pressures. Increases in intracranial pressure have been associated with higher morbidity and mortality in patients with intracranial hemorrhage. Close monitoring and control of intracranial pressure with hyperventilation, hyperosmotic treatment, and barbiturate administration is, therefore, essential to ensure good postoperative functional outcome. Intracranial pressure monitoring is also essential during a barbiturate coma because an increase in intracranial pressure may be the only sign of a postoperative hematoma and impending herniation. In their series of 10 patients who underwent emergency AVM surgery in the context of hemorrhage, Jafar and colleagues concluded that the 2 conditions necessary to ensure good surgical results were prompt decompression with hematoma evacuation and aggressive perioperative management of intracranial pressure.
Predictors of outcome after acute surgery
Although some risk factors for elective AVM surgery have been identified, such as increasing patient age, lesion size, and eloquent location, little data are available on predictors of good functional outcome in the acute surgical setting. In their 3-month follow-up of 49 patients with ruptured AVMs that were operated within 4 days of hemorrhage, Kuhmonen and colleagues identified the severity of the hemorrhage at presentation reflected by the Hunt-Hess score ( P = .001), as well as increasing patient age ( P = .006), as clear predictors of worse postoperative outcome. The concomitant presence of intraventricular hemorrhage also seemed to correlate with increasing morbidity and mortality ( P = .049).
Clinical Case Presentations
Clinical case I
A 50-year-old man presented to the authors’ emergency department complaining of headaches. Clinical examination was normal, but computed tomography revealed signs of subarachnoid hemorrhage ( Fig. 1 ). An angiogram showed a left inferior/posterior frontal lobe and left basal ganglia AVM, measuring 4.5 × 3.0-cm with superficial venous drainage ( Fig. 2 ). Five intracranial aneurysms were also detected, including an 8 × 5-mm flow-related aneurysm ( Fig. 3 A) that was discovered at the posterior communicating (PCOM)-P1 segment of the posterior cerebral artery junction and was determined to be the cause of the bleeding. The patient underwent immediate coiling of his PCOM aneurysm (see Fig. 3 B) and was discharged home in normal neurological condition. Four months later, the patient was readmitted and underwent surgical clipping of his 4 remaining aneurysms (a 2-mm aneurysm of the anterior communicating artery, a 3-mm aneurysm adjacent to an enlarged left choroidal artery infundibulum, a 4-mm wide-necked aneurysm on the M1 segment of the left middle cerebral artery, and a 3-mm left medial paraclinoid aneurysm). Subsequently, the patient elected to undergo staged embolization sessions followed by staged radiosurgery over 2 sessions. Thirty months later, he was neurologically intact, and his AVM was completely obliterated on digital subtraction angiography (see Fig. 3 C).
Clinical case II
A 30-year-old woman was brought unresponsive to our emergency department. She had complained of headaches before losing consciousness. Her computed tomography scan showed a 3-cm intraparenchymal cerebellar and vermian hemorrhage, with blood in the fourth ventricle ( Fig. 4 ). She was taken to the operating room for an immediate posterior fossa craniectomy and underwent partial hematoma evacuation. An external ventricular drain was placed initially and opened once the posterior fossa dura was exposed. An AVM was noted but not removed because of significant brain swelling. Postoperatively, a vermian AVM was determined to be the source of the bleeding ( Fig. 5 ). She was discharged home complaining only of mild gait incoordination and left-hand apraxia, with plans to be later readmitted for AVM resection. Three months later, the AVM was uneventfully resected microsurgically and she made a full recovery.
Clinical case III
A 51-year-old woman was transferred to the authors’ institution from a peripheral hospital unresponsive, with intracranial bleeding and acute hydrocephalus, which was treated with external ventricular drain placement. Her computed tomography scan revealed a large vermian and cerebellar hematoma with mass effect ( Fig. 6 ). A vertebral angiogram revealed a small (<1 cm) AVM in the midportion of the vermis with a single cortical draining vein ( Fig. 7 ). Her posterior fossa was decompressed with a craniectomy and partial hematoma evacuation ( Fig. 8 ). Two months later, the AVM was resected uneventfully and she made an excellent recovery.