Risk for Intracranial Hemorrhage

It is important to inform patients with AVMs of what their future is likely to be if untreated. Such information is complicated by the difficulty associated with understanding a disease that is diagnosed annually in just 0.9 to 1.5 per 100,000 population,^4,5 the complexity of identifying important angioarchitectural features such as aneurysms,^6,7 and the biases inherent in database research.⁸ Despite these difficulties, some information is available to help patients make an informed opinion on management. The literature would suggest that on diagnosis, the risk for future hemorrhage would be on the order of 4% per year with an annual mortality from hemorrhage of 1% (selected series results are reported in Table 385-1).^9–16 Based on patients prospectively observed until hemorrhage, further stratification into higher risk for hemorrhage (recent hemorrhage, presence of arterial or intranidal aneurysms, exclusive deep venous drainage, deep location, and increasing age) and lower risk for hemorrhage (asymptomatic lesions in the absence of a hemorrhagic history) is possible.^{10–14,16,17} Stratification of risk beyond this point may be inappropriate because of the small number of patients in any series, the nature of referral bias, the noted complexity in assigning angioarchitectural features, and the absence of trial results. It has been argued that we do not have enough information to predict the natural history of AVMs with accuracy and that randomized control trials would be necessary to provide an informed opinion.¹⁸ However, the obstacles and ethics of mounting such a trial, which would require sufficient length of follow-up to predict a course, often in excess of 20 years, would be extremely challenging for this rare and complex disease for which treatment is possible and outcomes with nonintervention can be catastrophic. Furthermore, there is sufficient concordance among studies (in different centers and time periods) to have some confidence that for the purpose of informed consent, a reasonable opinion of the natural history can be made.

TABLE 385-1 Reports Relevant to the Natural History of Hemorrhage from Arteriovenous Malformations

What assumptions can be made regarding the information that we have regarding the natural history? Although AVMs are congenital lesions, their risk for rupture is probably not constant. This is reflected in prospective studies that have indentified increased risk for hemorrhage after recent hemorrhage,¹⁷ with associated aneurysms (usually acquired),^14,17,19 and with increasing age.^11,12,20 Moreover, retrospective data suggest that acquired venous outflow stenosis (a condition that is more likely to produce a change in pressure within the AVM nidus where such venous outflow is limited, e.g., exclusive deep venous drainage) is also associated with increased risk.^21–23 A unifying theme for these risk-altering characteristics is the wear-and-tear changes occurring over time with the obligate high shear stress associated with AVMs. Therefore, predicting the risk for hemorrhage should be considered both from the short-term point of view (based on the history and angioarchitectural features present at diagnosis) and from the long-term point of view (where wear-and-tear changes have yet to develop but may do so in the future). Having both short-term and long-term conservative management outcomes in management templates is important for AVMs because the expected hemorrhage rate during the latency period (i.e., short term) after focused irradiation will need to be considered when determining the risks associated with focused irradiation. A long-term perspective is also of importance given that the majority of AVMs are diagnosed in patients with an expected survival of more than 2 decades in the absence of an AVM. Although a hemorrhagic manifestation is known to increase the risk for rupture in the short term, this may not be of importance when considering the risks over decades. In the only long-term study, 40% of the AVM population suffered at least one new hemorrhage (mean follow-up of 23.7 years).⁹ In this Finnish study, in contrast to the results from studies with a mean follow-up of less than 10 years, an initial hemorrhagic manifestation was not seen to be an important predictor for future hemorrhage. This underscores the potential for wear-and-tear changes to be acquired over time such that a low-risk AVM can be changed into one with greater risk for rupture.

The short-term indicators of increased risk are unified by their relationship to hemodynamic “wear and tear” and fall into three categories: features demonstrating that the vasculature has had a history of being breached (history of hemorrhage), features suggesting that degeneration is occurring (presence of aneurysms, increasing length of time, increasing age),^{11,12,20,24,25} and those indicating increased vulnerability to breach because of limitations in potential alternative outflow in the event of acquired flow-related venous occlusion (exclusive deep venous drainage).²⁵ Although patients with associated aneurysms and AVMs would expect to have a higher risk for hemorrhage, the magnitude of the increase in risk is significantly greater than the summation of the risks.^10,14,25 The cause of the increased risk for hemorrhage from such an association is likely to be explained by vascular “wear and tear,” with the aneurysm being a marker for this progression in pathology.¹⁴

A relevant potential criticism of management decisions based on the natural history derived from clinical series is the potential for a significant number of AVMs to remain undiscovered throughout life. This would overestimate the true risk. However, the context in which we make management decisions is generally the same context in which the natural history data are derived. This allows assumptions derived from the known population with AVMs to be reasonably applied to all people with AVMs. Furthermore, some evidence exists that the pool of undiscovered AVMs remains small.¹⁵

In an effort to consolidate evidence presented in this section for use in constructing a simple paradigm of management, the following assumptions have been made (while recognizing the wide 95% confidence interval [CI] for most of the raw data) (Table 385-2):

TABLE 385-2 Chance of Remaining Hemorrhage Free from the Time of Diagnosis

* Presuming a 2% risk per year for the first 5 years and increasing to 3% per year after 5 years

† Presuming a 7% risk per year for the first 2 years and declining to 3% per year after 5 years)

Consequences of Intracranial Hemorrhage

Prediction of the likelihood of death and disability as a result of an intracranial hemorrhage from an AVM is the most important factor influencing the opinion of best management. Although it is often useful to look at morbidity and mortality as collective complications, patients and families frequently want these adverse outcomes detailed separately. The emphasis on avoidance of death versus avoidance of death and disability will vary among patients. For those who have already experienced morbidity because of repeated hemorrhages (which continue to compound the existing deficits), avoidance of death may be more important than the possibility of morbidity from surgery.

Death as a Result of Hemorrhage from Arteriovenous Malformation

In the only long-term study, death from hemorrhage occurred in 23% of subjects over a 23.7-year period.⁹ The annual mortality from hemorrhage in this study was 1%, and 25% of hemorrhages from AVM resulted in death. On a hemorrhagic incidence basis, mortality of nearly 30% with each hemorrhage has been reported.^13,24,25 In a population study, mortality secondary to hemorrhage from an AVM was 18% with a 95% CI of 3.8% to 43.3%.²⁶ Although recurrent hemorrhage has been reported to have higher mortality than initial hemorrhage,^24,27,28 this observation may reflect referral bias and sample size given that a patient sustaining a recurrent hemorrhage is more likely to be under observation with a correct diagnosis of AVM at the time of hemorrhage; in contrast, the source of a devastating intracranial hemorrhage may not be investigated at the initial evaluation of such a patient. The report of Fults and Kelly of 13.6% mortality for a first hemorrhage and 25% for a third or subsequent hemorrhage²⁴ is a risk very close to that of the 29% reported by Brown and colleagues after a first hemorrhage from an AVM.¹³ Although evidence exists that recurrent hemorrhage may stereotypically mimic the first hemorrhage (and thus survival from hemorrhage would predict survival from further hemorrhage),²⁹ this may be applicable only in the short term.^29,30 An estimated 20% to 25% mortality from any hemorrhage secondary to an AVM is not unreasonable for the purposes of constructing a useful predictive model of outcomes after AVM.

Morbidity as a Result of Hemorrhage from Arteriovenous Malformations

Permanent new neurological deficits arising as a consequence of AVM hemorrhage are more difficult to estimate than the categorical outcome of death. Estimates for permanent adverse outcomes vary.²⁹ However, there is some concordance among several series, with an approximately 50% risk for death and disability for each intraparenchymal hemorrhage from an AVM.^{13,24,25,30,31} The distribution of death and morbidity differs among series. There is a tendency for earlier series to report a greater proportion of cases resulting in death rather than morbidity. However, the combination of an approximately 40% to 50% risk for morbidity and mortality is consistent over time.

Nonhemorrhagic Decline in Health from Arteriovenous Malformations

Annually, 1.5% of patients with an AVM of the brain will undergo functional decline.¹³ The mechanism of the decline is usually either seizures (new or progression) or progressive neurological deficits caused by regional arterial hypotension or venous hypertension, or both.³¹ For such progression to take place it is likely that the AVMs would be large, supratentorial, and cortically based.

Grading System

Grading systems attempting to estimate the risk associated with surgery have been used for more than 20 years.^{26,33–35,41,48–50} For grading systems to be of use, however, they must be applicable in the context in which the patient is to be managed. Although the Spetzler-Martin grading system can be criticized for lack of weighting of variables and lack of independence of variables,³⁵ the need for further subclassification,⁴¹ intraobserver and interobserver error,^6,7,35,40 and biasing from patient exclusion because of perceived risks,^38,47 it is simple (grade I for a <3-cm nidus, grade II for 3 to 6 cm, grade III for >6 cm, with a point added for deep venous drainage and a point added if located in or adjacent to one of the following: sensorimotor cortex, language cortex, visual cortex, thalamus, hypothalamus, internal capsule, brainstem, cerebellar peduncles, or deep cerebellar nuclei), popular, reasonably robust, and a reliable discriminator of relative risk for patients undergoing surgery.⁴⁹ Even though the relative risk between Spetzler-Martin grades is appropriately discriminatory, the absolute risk cannot be used without taking into consideration patients excluded from surgery because of perceived operative difficulty. With an increasing proportion of patients excluded with an increasing grade of AVM, estimation of risk becomes increasingly misleading. As the grade increases and the nonsurgical proportion increases (Fig. 385-2) (e.g., only 5% of patients with Spetzler-Martin grades IV and V undergo surgery at Barrow Neurological Institute), the absolute risk related to surgery from published series probably increasingly understates the true risk.⁴⁶ This cannot be used as an argument against surgery because ancillary selection criteria can produce an informed decision and result in the reported permanent new neurological deficit rates in the literature. It is important to establish what selection criteria are being used before patients are selected for surgery. Biasing is likely to be specific to the institutional management norms, and this needs to be considered when informing patients what the risks associated with surgery for a specific grade may be. Biasing is evident in my experience (Table 385-4; also see Fig. 385-2). It is reasonable to conclude that for my series, risks associated with surgery for AVMs of grade I and II and for AVM grades greater than II in noneloquent brain regions are representative of all AVMs of their specific grade and type. However, the 21% (95% CI, 16% to 28%) of patients experiencing new permanent neurological deficits (modified Rankin score >1) as a result of surgery may seriously underestimate the risks for all eloquently located AVMs of Spetzler-Martin grades greater than II because there is a 14% incidence of surgery not being performed for risk-related reasons.

TABLE 385-4 Results of Treatment by Surgery (Including Preoperative Embolization As Indicated)

In conclusion, with regard to grading systems it is reasonable and appropriate to use Spetzler-Martin grading as the basis for stratifying and communicating surgical risks. It is also important to examine the Spetzler-Martin grades in the context in which the patient is to be managed and to incorporate additional variables demonstrated to have an impact on patient management (such as a diffuse nidus and lenticulostriate arterial supply)^35,38,40 as second-tier criteria for decision making.

Efficacy of Surgery

The primary aim with most surgery is prevention of hemorrhage. However, other goals of surgery need to be considered for some patients. Management of acute hemorrhage in the presence of an AVM, management of seizure disorders (inadequately controlled by medications), and management of neurological deficits caused by the combination of focal arterial hypotension and venous hypertension (often referred to as “steal”) may each be considerations for surgery in individual patients. Because of the small number of such cases, the evidence base for recommendations is limited for these secondary recommendations.

Prevention of Hemorrhage

When management is planned as a single surgery (with or without preoperative embolization), the early postoperative angiogram usually confirms ablation of the AVM. An adult with an AVM that has been ablated, as evidenced by angiography, is assumed to be cured and not at risk for future hemorrhage. However, the planned surgery does not always result in early AVM ablation on postoperative angiography. Furthermore, ablation on early postoperative angiography does not always result in cure. Data on early postoperative angiography are limited. I have experienced a 4% rate (95% CI, 2.6% to 6.0%; 21 of 526 surviving surgical patients) of unplanned residual AVM after surgery. In our series, all Spetzler-Martin grades were represented and there was no correlation with any particular clinical manifestation or angioarchitectural feature. Management of this residual is normally a simple matter to deal with during the early postoperative period. However, that further unintended intervention may be required for AVM ablation in 4% of patients needs to be factored into consideration for the patient to make an informed decision.

Delayed detection of a previously obliterated AVM also needs to be considered in management decisions. Errors can be made in interpretation of the postoperative angiogram. In addition, it is not always certain that the absence of early venous drainage on angiography confirms cure of the disease. It is possible, for example, that vasospasm, temporary occlusion of venous outflow (with unstable thrombus that may recanalize at some interval after angiography), or unobliterated angiomatous (or perinidal) feeding vessels may be complicit in the development of a future AVM nidus.^51,52 Delayed angiography has identified new, residual, or recurrent AVMs in 3.5% of children whose AVMs were previously confirmed to be angiographically obliterated after surgery.⁵³ This incidence of new, residual, or recurrent AVMs was not part of routine screening in the series of Kader and colleagues,⁵³ and thus their results probably represent a minimal incidence. The occurrence of new, residual, or recurrent AVMs in adults cannot be excluded. Such findings have also been detected, after previously confirmed to be angiographically obliterated, after focused irradiation.^54,55 Despite the suggestion that risk for late AVM recurrence is confined to children,⁵³ this has not been my experience. In 101 patients with radiologic follow-up of more than 2 years after confirmed ablation, 5 were confirmed to have a new AVM (1 of whom was seen because of a late new hemorrhage) after an early postoperative angiogram confirmed “cure.” Only 1 patient was younger than 20 years at the initial surgery. This may over-represent the true risk for late new, residual, or recurrent AVMs because the discovery was made after a hemorrhagic event in 1 of these 5 patients, with just 4 new malformations being detected in 100 patients subjected to routine screening (i.e., 4% with a 95% CI of 1.6% to 9.8%). The optimal timing of delayed angiography has not yet been determined and the rate of late residual, recurrent, or new AVMs still has to be established, but in the absence of these details, for the purpose of decision making it is best to consider the initial surgery as being curative in 90% of patients.

Surgery and Epilepsy

New Neurological Deficits

The development of new neurological deficits related to surgery is a key factor in determining management. New neurological deficits account for nearly 80% of the complications of surgery and are present immediately on awakening from surgery in more than 80% of the patients in whom deficits will develop.⁶⁰ Serious infections and complications of venous thrombosis should not be discounted as challenges to surgical management even though the focus is on neurological outcomes. However, the efficacy of management and the occurrence of new permanent neurological deficits are the most important determinants in decision making for AVM management. Examples of series reporting an incidence of new permanent neurological deficits are provided in Table 385-3.^{35,41,42–45}

These series demonstrate the relationship between Spetzler-Martin grade and outcomes, and the results may be generalizable to all AVMs of a specific Spetzler-Martin grade if the number of patients excluded for reasons of surgical difficulty is low. In my retrospectively analyzed, prospectively collected database (which included the reason for management other than surgery), in patients with Spetzler-Martin grades I and II who underwent surgery, permanent new neurological deficits occurred in less than 2.5%, whereas in patients with Spetzler-Martin grades III and IV in noneloquent brain regions who underwent surgery, new permanent neurological deficits developed in 17% (see Table 385-4). In both these groups the results are generalizable, with very few patients being refused surgery because of perceived operative difficulty. However, the risk for new permanent neurological deficits in patients with AVMs of Spetzler-Martin grade III or greater that are located in eloquent brain regions does not reliably reflect the risk for all such AVMs (because 14% of all such AVMs were excluded as a result of perceived operative difficulty). These AVMs are, in general, likely to have a significantly greater risk associated with surgery than the actual demonstrated risk for the group (21%).

Balance between Cure and Risks Associated with Treatment by Focused Irradiation

Radiotherapy has been used for the management of AVMs for nearly as long as surgery has.³ However, the radical change with the advent of focused irradiation has resulted in it being recommended as the preferred treatment option for many AVMs. As with any of the treatment strategies, the local context has a significant impact on outcomes. This well-tolerated procedure needs to be considered in terms of the risks related to treatment, the risk associated with the unobliterated AVM nidus (between initiation of treatment and cure), and the efficacy of treatment. These considerations are interactive and relate to the dose of treatment and the size and location of the nidus. The risk for permanent deficits increases when the brain receives doses escalating above 10 Gy in a single fraction. To be considered at the same time is that the chance of cure of the AVM nidus decreases when marginal doses to the nidus decrease below 25 Gy. The smaller the volume of the AVM nidus to be covered, the steeper the gradient in dose at the margins of treatment. Hence, smaller lesions can be treated with a higher marginal dose to achieve a high probability of cure with an acceptably low risk for brain damage. From more than 1000 Gamma Knife procedures performed at the Karolinska Institute,⁶¹ the probability of obliteration was found to be 35.69 × ln (marginal dose) − 39.66. This model predicts that for a marginal dose of 22 Gy there is a 71% probability of cure, which decreases to 55% with a marginal dose of 14 Gy. This equation is clearly accurate only over the range of marginal doses normally considered in treating AVMs. Linear accelerator (LINAC)-based focused irradiation can achieve similar results. However, the probability of cure is clouded by the difference between those who have been demonstrated to be cured and those who have not yet been demonstrated to be cured. That there is a gap between our knowledge of patients with demonstrated failure and those who have not yet been demonstrated to have been cured is understandable because of the referral pattern for most focused irradiation services and the latent period between treatment and cure. This is illustrated by Friedman and colleagues’ report on LINAC-focused irradiation for Spetzler-Martin grade I and II AVMs.⁶² At 3 years, 66% of patients were demonstrated to have been cured and 21% were known to have failed to be cured. This discrepancy is common and has led some to suggest an approximately 30% lower rate of cure for focused irradiation than what has been reported.⁶³ Thus, for an AVM nidus larger than 30 mm in diameter, the probability of obliteration at 2 years may be just 16% in highly regarded treatment institutions.

Innovative radiosurgical strategies (such as low-dose irradiation followed by repeated treatment) to increase the likelihood of obliteration continue to be developed in an attempt to deal with larger AVMs.^64,65 However, fractionated treatments have a low chance of success,⁶⁶ and any significant change in radiotherapy treatment protocols from single-fraction radiosurgery to multiple fractions will need to be validated with patient outcome data for these new protocols.

Another innovative combination is planned embolization to reduce the focused irradiation target. However, AVMs undergoing pretreatment embolization (in an attempt to reduce the focused irradiation volume) cannot be considered to be the equivalent of untreated AVMs of comparable volume. Postembolization AVMs are obliterated at 70% of the probability of obliteration of AVMs that have not undergone previous treatment when matched for AVM residual volume, marginal dose delivered, and location.^67,68

Of some concern is the report of delayed hemorrhage after documented cure with focused irradiation.^54,55 Delayed hemorrhage has also been reported after surgery on early postoperative angiography, but it is presumed that absence of the AVM on an angiogram performed several years after treatment would represent a cure with any form of therapy. The suggestion that there may be an ongoing small risk for hemorrhage after demonstrable occlusion assessed some years after treatment will require further study.

Radiosurgery for children can be performed without significant deviation from the results expected for adults.^69,70

Although there are limitations in the potential for cure because of the dose that can safely be administered to normal brain, there are also limitations resulting from the accuracy of target definition. Variability among planners in determining contouring may lead to underdosing.⁷¹ This underscores the importance for radiosurgery institutions to provide outcome data to ensure that performance is known and optimal.