Natural History and Management Options of Unruptured Brain Arteriovenous Malformation

2 Natural History and Management Options of Unruptured Brain Arteriovenous Malformation

Michael Kerin Morgan

Abstract

The annual risk of intracranial hemorrhage (ICH) from an unruptured brain arteriovenous malformation (bAVM) is likely to be 1.8% with a 20-year risk of 29%. There is a strong case for venous outflow stenosis (VOS) associated proximal intracranial aneurysms (APIA), and increasing age to be included as risk factors for bAVM ICH. The risk associated with ICH during pregnancy is inconsistent and confusing, although it is reasonable to state that pregnancy provides no protection against ICH. The consequence of ICH in association with bAVM very much depends upon the context. For patients who cannot be treated and therefore may sustain multiple ICH, the risks of cumulative morbidity and mortality is 70% (95% confidence interval [95% CI]: 63–76) including a 42% (95% CI: 36–49) risk of death. For an initially conservatively treated patient with an unruptured bAVM who will be treated if an ICH occurs (i.e., they will have a single ICH), there is a 42% (95% CI: 33–51) risk of morbidity and mortality, including a 9% (95% CI: 6–15) risk of death. The grading systems that should be utilized to assist decision-making in the management of bAVM include the Spetzler–Martin Grade (SMG) and its simplification into the three-tier Spetzler–Ponce Classification class (SPC classes A, B, or C), the Lawton–Young Supplementary (LYS) score, and the Virginia Radiosurgery AVM Scale (VRAS). Familiarity with cohort data allows guidance for the best recommendation of treatment for the individual patient. Unfavorable outcomes (unFOs; persistent bAVM, permanent complication of treatment or posttreatment hemorrhage) following surgery are expected to occur in 1 to 11% of cases for SPC class A (complication increasing with increasing size within class) and 10 to 38% for SPC class B (complication increasing with increasing size within class). Unruptured SPC class C bAVMs normally have such a high complication rate that they do not warrant consideration for surgery. Following radiosurgery, unFO is predicted to occur in 19, 25, 34, 52, and 59% for VRAS scores of 0, 1, 2, 3, and 4, respectively. It is important to remember that the proportion of permanent neurological deficit contributing to the unFO is greater for surgery than that for radiosurgery and the proportion of persisting bAVM contributing to the unFO is greater for radiosurgery than that for surgery. Embolization may be appropriate in some centers that have results comparable to those achieved by other treatments. However, as yet, embolization has not achieved the same level of evidence warranting its general support as an intent-to-cure treatment option for unruptured bAVM.

Keywords: brain arteriovenous malformation natural history risk of hemorrhage laminar wall shear stress surgery radiosurgery embolization

2.1 Introduction

A brain arteriovenous malformation (bAVM) may cause no problems (an unknown proportion of the population with bAVM) or cause intracranial hemorrhage (ICH; presentation in ~50% of cases), seizures (presentation in ~25% of cases), headaches (presentation in ~16% of cases), neurological deficits not associated with ICH (presentation in ~8% of cases), or a combination of these events.1^, ² Therefore, defining the natural history to be studied can be very complex given the variety of event types and variations in the risk factors for differing events. As an example, posterior fossa bAVMs are very unlikely to present with seizures, neurological deficits without ICH are most likely to occur in large bAVMs near critical brain, and headaches are not likely to be associated with small bAVMs unless associated with venous obstruction. Furthermore, consequences of these events may be highly variable and may produce cumulative problems.

There is no study that could claim to be a study of the true natural history, that is, a study that captures a sufficiently large number of representative cases from the overall population at the inception of the disease and follows these cases without intervention until death. This is not to say that a good estimate of the risk of ICH cannot be made. A number of studies do shed some light on the future risk of ICH. However, the limitations of these studies need to be understood so that, as far as possible, the inevitable biases that distort the results may be taken into consideration and discrepancies between studies explained.

Prevention of ICH is the most important management goal in bAVMs and is the reason for preemptive treatment. Although the various other ways that bAVM can cause symptoms are of importance, for the most part, these on their own do not warrant treatment. Therefore, knowing the risk of ICH is important for bAVM management. The purpose of this chapter is to provide reasonable estimates of the future risk of ICH in unruptured bAVMs and to discuss how this is best prevented.

2.2 Selected Papers on the Natural History of Unruptured bAVM

●Gross BA, Du R. Natural history of cerebral arteriovenous malformations: a meta-analysis. J Neurosurg 2013;118(2):437–443

●Kim H, Al-Shahi Salman R, McCulloch CE, Stapf C, Young WL; MARS Coinvestigators. Untreated brain arteriovenous malformation: patient-level meta-analysis of hemorrhage predictors. Neurology 2014;83(7):590–597

●Hernesniemi JA, Dashti R, Juvela S, Väärt K, Niemelä M, Laakso A. Natural history of brain arteriovenous malformations: a long-term follow-up study of risk of hemorrhage in 238 patients. Neurosurgery 2008;63(5):823–829, discussion 829–831

●Al-Shahi Salman R, White PM, Counsell CE, et al; Scottish Audit of Intracranial Vascular Malformations Collaborators. Outcome after conservative management or intervention for unruptured brain arteriovenous malformations. JAMA 2014;311(16):1661–1669

2.2.1 Comparing Future Risk of ICH for Unruptured bAVM

From two recent pooled data analyses, the annual risk of ICH following diagnosis of an unruptured bAVM is reported to be 1.3% (95% CI: 1.0–1.7) for the “Multicenter AVM Research Study” (MARS) and 2.2% (95% CI: 0.9–5.5) for the Gross and Du study.2^, ³ The Gross and Du study was a meta-analysis of a number of cohorts. The MARS was formulated from a combination of two large cohort studies (UCSF AVM database and Columbia AVM database) and two population-based studies (“Scottish Intracranial Vascular Malformation Study” [SIVMS] and the Kaiser Permanente of Northern California) with analysis of individual patient data.³ Utilizing the decay equation (e^{–1.annual risk as}^{proportion.years} ),⁴ the projected cumulative incidence of a patient experiencing a first intraventricular hemorrhage (IVH) for the Gross and Du study was 11% at 5 years, 20% at 10 years, and 35% at 10 years. For the Mars, these results were 6% at 5 years, 12% at 10 years, and 23% at 10 years (Fig. 2.1). The methodological differences of these two pooled data studies do not entirely explain the different outcomes. Because the risk of ICH may not be constant with respect to time, these projected data from annualized risks need to be tested against large series with long-term follow-up. Results of short-term follow-up for unruptured bAVMs suggest a higher rate of ICH in the first 12 months compared with longer periods (Fig. 2.1).¹^, ⁴ Therefore, the use of cohorts with long-term results is more appropriate to test the projected results from which to compare the projected annualized long-term risks from MARS and Gross and Du.²^, ³ When confining our search to cohort studies with a minimum of 90 cases and at least 10% of the initial cohort remaining assessable at 10 years of follow-up (this provides reliable data for a minimum of 10 years), four studies are of interest (Fig. 2.1).⁵^, ⁶^, ⁷^, ⁸ These cohort studies provide moderation to the risks calculated from the projected annualized ICH rates derived from the two pooled data studies of MARS and that of Gross and Du. The longer-term studies lay between the projected annualized rates of MARS and Gross and Du.⁵^, ⁶^, ⁷^, ⁸ Therefore, it is reasonable to assume (apart from the very early period after diagnosis when some studies report a higher rate of ICH) that fitting the curve of a projected constant ICH rate in the decay equation over 10 to 20 years is a good fit in long-term studies and that MARS underestimates the risk of ICH and the Gross and Du meta-analysis overestimates the risk.²^, ³ Underestimating the risk of ICH may be explained partly by event identification (a lower capture rate for patients experiencing ICH as compared with those returning for routine follow-up) and censoring patients to treatment with a higher risk of future ICH.⁴ Overestimating the risk may be partly explained by event identification (a higher capture rate for patients experiencing ICH as compared with those returning for routine follow-up) and studies weighted for short-term follow-up (see Fig. 2.1). The study by Hernesniemi et al,⁷ from a time when most unruptured bAVMs were untreated, in which there was a 9.3-year median follow-up and only 1% of cases were lost to follow-up, is worthy of special attention.⁷ This cohort found an almost exact intermediate result between the two pooled data studies with cumulative incidence of a patient experiencing a first IVH of 8, 16, and 29% at 5, 10, and 20 years from diagnosis, respectively (Fig. 2.1).⁷ Applying the decay function to Hernesniemi et al’s results approximates an annualized ICH rate for patients with unruptured bAVM to be 1.8%.⁴ This would seem to be a reasonable intermediate rate between the annualized rates of 1.3% from MARS and 2.2% from the Gross and Du meta-analysis.²^, ³ Therefore, we believe a reasonable estimation of annual risk of ICH for unruptured bAVM is 1.8% and the 20-year risk is 29%.

Fig. 2.1 The percentage free of first intracranial hemorrhage (ICH) with respect to time following diagnosis of an unruptured brain arteriovenous malformation (ubAVM). This graph represents data provided by series with greater than 90 cases followed with a reliable number of patients at risk (10% of the initial patient cohort): orange (Brown et al⁶)—168 cases; black (Hernesniemi et al⁷)—99 cases; green (Crawford et al⁵)—217 cases; blue (Al-Shahi Salman et al⁸)—101 cases; red thin (Mohr et al⁴⁰)—125 cases; yellow (Stapf et al¹⁹)—340 cases; brown (Halim et al⁹)—423 cases; purple (da Costa et al¹⁵)—420 cases. Studies with short-term data (Mohr et al,⁴⁰ Stapf et al,¹⁹ Halim et al,⁹ and da Costa et al¹⁵) have also been included to demonstrate their inclusion would suggest a higher risk of first hemorrhage as compared with longer-term studies. The meta-analyses (blue—Gross and Du²; red—Kim et al³), represented by the broken dashed and dotted lines, are extrapolated from the annualized 2.2%² and 1.3%³ calculated by the decay equation (e ^{–1.annual risk as}^{proportion.years} ).⁴

2.2.2 Factors that Impact on the Risk of First ICH

There are a number of factors that have been proposed to increase the risk of subsequent ICH in unruptured bAVMs. These include venous outflow stenosis (VOS) an associated proximal intracranial aneurysm (APIA), older age, and pregnancy (see below).

●VOS: VOS may be associated with venous ectasia and may be more common with bAVMs associated with deep venous drainage (DVD).2^, ³ Although acute venous thrombosis may be a cause for VOS, the most common cause is an acquired intimal hyperplasia.¹⁰^, ¹¹^, ¹² There is considerable evidence implicating VOS as a risk factor of ICH as well as a reasonable underlying physiological basis for the association (reduced laminar wall shear stress [LWSS]). However, there are no prospective data following untreated patients with VOS compared with those without VOS. Such studies may not be feasible given the widespread acceptance of this condition as a risk factor suggesting the likelihood of a selection bias (due to the preference for intervening when VOS is identified) and the small proportion of cases with VOS in a rare disease making statistical significance difficult to achieve. A consequence of VOS is its effect on the upstream bAVM vasculature. The impact is twofold. There will develop an increased pressure arising once the progression of stenosis is significant and there is a reduction in LWSS with the consequent vascular degeneration. Both of these physiological perturbations, either on their own or in combination, may be a reason for ultimate vascular integrity failure and ICH.

●APIA: Aneurysms may be intranidal, APIAs, or coincidental (occurring in an artery not contributing to the bAVM). The association of aneurysms with an increased risk of ICH was first reported prior to 1991.¹³^, ¹⁴ The source of an ICH may be either the APIA or aneurysm. The predominant source of ICH when APIAs are present is the bAVM.¹³ More recent cohorts have observed this association inconsistently.¹⁵^, ¹⁶^, ¹⁷ An identified association reported in an earlier study of a cohort was not confirmed in later analysis of a larger cohort from the same database.¹⁸^, ¹⁹ This association was identified in the Gross and Du meta-analysis (hazard ratio: 1.8; 95% CI: 1.6–2.0) but was studied in only two of the nine series incorporated in the study. For one of the studies that identified the association of aneurysms with ICH, the mean follow-up was only 2.8 years.² In MARS, the association was significant for the first 12 months but was not significant beyond this period.³ Therefore, the association of an increased risk with APIA is complex and may be present for only a short period after diagnosis, disappearing beyond the initial 12 months.⁴ APIAs are acquired.²⁰ As with intracranial aneurysm formation in the absence of bAVM, high LWSS is likely to be responsible for aneurysm formation (although low LWSS may be associated with aneurysm rupture).²¹ LWSS may only be elevated at the time of bAVM formation and early growth and may be normal or low at other times. Therefore, there may be only a limited period of time for the development of APIA after which the conditions favor vascular stability. Following treatment by radiosurgery, prior to bAVM occlusion, there is an increased risk of bAVM rupture in the presence of APIAs.²² This may be considered support for the proposition that APIA increases the risk of bAVM ICH because the period between treatment and obliteration is considered to follow the natural history of untreated bAVMs. However, the alternate explanation is that during progressive thrombosis of the bAVM as a consequence of treatment, LWSS is reduced, leading to degeneration and disruption of the bAVM. It may be argued that those bAVMs with an APIA are, overall, more likely to have had periods of high LWSS in their past as compared with bAVMs without APIA. If so, they have the vascular angioarchitecture that may lead to a greater reduction in LWSS and greater increase in pressure arising as the consequence of the progressive obliteration by radiosurgery. The reduction of LWSS may increase the inflammation of the vascular wall and the increase in pressure may challenge the integrity of the vascular wall. Therefore, the data from postradiosurgery ICH cannot be legitimately extrapolated to an untreated bAVM with APIA.
In summary, with regard to APIAs, it is reasonable to suggest that there may be an association with the risk of bAVM ICH. However, this may be transient. Furthermore, APIA may be a surrogate marker for age, and it may prove to be that increasing age is the factor associated with increasing the risk of bAVM ICH rather than the APIA.

●Older age: Age was found to be a factor associated with an increased risk of bAVM ICH in the combined cohort and population series of MARS.³ However, this was not found in the Gross and Du meta-analysis. This suggests that this association, if true, is likely to be weak.² The importance of bias is significant as younger patients are more likely to be offered treatment because of their lower chance of complications of treatment and the longer time they have to live with a threat of ICH. Therefore, there is an age-related selection bias. In order to reduce this bias, evidence from a time when conservative management was more prevalent is important. Two such studies supported the effect of increasing age being associated with an increasing risk of bAVM ICH,⁵^, ¹³ but a third failed to demonstrate this association.⁷ Overall, it would seem reasonable to accept there is an association between increasing age and bAVM ICH, but suggest that such additional risk, in comparison with that of presentation with ICH, is small. Furthermore, increasing age has been reported to be associated with both the development of VOS and the diagnosis of APIA.¹¹^, ¹² Therefore, there may well be a complex interaction of these factors. This is perhaps best illustrated by contrasting a study from Toronto that found APIA, but not age, was associated with increased risk of bAVM ICH,¹⁵ in comparison with a study from Columbia that found the converse result.¹⁸^, ¹⁹ Furthermore, the Columbia database reported an association with APIA and presentation with bAVM ICH,²⁴ but subsequently failed to confirm this association.¹⁹

●Pregnancy: Few areas related to bAVM have resulted in such confusing results and consequentially inconsistent management recommendations (including pregnancy termination, avoidance of pregnancy, and no change in management).²⁵^, ²⁶^, ²⁷^, ²⁸^, ²⁹ Studying the risk of pregnancy in patients with bAVM presents many challenges. This is highlighted by the variation in annualized risk of first ICH during pregnancy for cohorts ranging from 3.0% (95% CI: 1.7–5.2) to 30% (95% CI: 18–49).²⁷^, ²⁸ Even in a study in which the same methodology was employed for two different cohorts, the results vary widely with one cohort having 8.6% (95% CI: 1.8–25) annualized risk of first ICH during pregnancy and the second 30% (95% CI: 18–49).²⁷ Furthermore, no reasonable explanation for why a normal pregnancy would increase the risk for hemorrhage has been provided. The argument that a hormonal or increased blood volume has an influence on the predisposition to ICH from a bAVM is still speculative. The issue of the risk of bAVM ICH during pregnancy remains to be determined. However, it is reasonable to state that pregnancy does not offer any reduction in risk of bAVM ICH.

2.2.3 The Expected Outcome from bAVM ICH

The outcomes of bAVM ICH are more complex to determine than other causes of spontaneous ICH. This is because there is an increased likelihood of multiple ICHs if the bAVM remains untreated with the risk of recurrent hemorrhage greater than that for the first ICH. Furthermore, the risks of complications from treatment can be significant and may be difficult to separate from those of the ICH itself. For all cases, the reported range of combined morbidity and mortality is 40 to 90% and the range for mortality alone is 0 to 60%.4 However, the specific situation is an important consideration. Patients presenting with an acute ICH have a 40% (95% CI: 32–48) risk of new permanent neurological deficit or death including a 15% (95% CI: 12–19) risk of death as a consequence of this single ICH.⁴ Untreated patients followed until ICH have a 42% (95% CI: 33–51) risk of morbidity and mortality including a 9% (95% CI: 6–15) risk of death after the next single ICH.⁴ Untreated patients followed long term who may be subject to multiple ICHs have a cumulative 70% (95% CI: 63–76) morbidity and mortality including a 42% (95% CI: 36–49) risk of death.⁴ Patients treated by radiosurgery followed until the next ICH or subsequent ICH have a 41% (95% CI: 33–50) risk of new permanent neurological deficit or death and a 27% (95% CI: 22–33) risk of death.⁴ These various situations need to be understood as the clinical pathway will vary between those patients who cannot reasonably be treated and those who will be treated when diagnosed. For example, it may be reasonable to assume that the risk of morbidity and mortality for a patient with an SPC class A unruptured bAVM would be about 40% if an ICH were to occur and that for a patient with an SPC class C case would be about 70% (because of multiple ICH or subsequent treatment following ICH). In order to simplify these data, an estimated risk of permanent neurological deficit or death from a first ICH of 40% is reasonable if the bAVM would be otherwise suitable to treat if it had presented with ICH rather than unruptured. This is because it is possible to change management pathways if ICH should occur.

2.2.4 Understanding the Cause for ICH Associated with bAVM

In order to better understand the natural history ICH associated with bAVM, the cause for vascular integrity failure needs to be understood. The associated risk factors of bAVM ICH discussed earlier do not directly explain why blood vessels break to cause ICH. VOS, APIA, and age are not independent of each other. Both the identification of VOS and occurrence of APIA are associated with increasing age. A unifying underlying physiological association of these factors is deviation of LWSS from normal. APIA is associated with high LWSS, VOS associated with a point of low LWSS, and age (if VOS progresses) may be associated with a reduction in LWSS. There is a reasonable basis to suspect that a reduction in LWSS explains the ICH. Although this is yet to be verified prospectively (in part because of the difficulties until recently in calculating LWSS), the hypothetical basis is compelling.

LWSS is tightly controlled within a normal range by the endothelial mechanoreceptor-vascular wall response and is likely to remain intact in patients with bAVM.³⁰ LWSS in the feeding vessels of bAVM may vary with the life cycle of a bAVM (Fig. 2.2). After the initiation of the fistulas, the increased velocity of the involved vasculature leads to increased LWSS resulting in vasodilation, angiogenesis, and arterial recruitment.³¹ These responses have the effect of returning the LWSS toward normal (although with persisting increased velocity of blood flow) within the dilated bAVM vasculature. The initial high LWSS may also explain the presence of APIA formation.²⁰ Studies of LWSS in the proximal arteries of bAVM confirm that values do not vary widely from normal.³²^, ³³ High or normal LWSS is associated with normal healthy endothelium.³⁴ As with arteriovenous shunts from any cause, points of low LWSS may be present to produce VOS.³⁵ Low LWSS is responsible for an inflammatory-degenerative process resulting in most of the histopathological degenerative process evident in bAVM and intimal hyperplasia in the VOS.³⁴ A vicious cycle may ensue to progress the degenerative process with ultimate rupture resulting from the combination of the vascular degeneration and the rise in pressure if the VOS is significant. This explanation is given in detail in Fig. 2.2.

Fig. 2.2 The hypothetical progression over time of a brain arteriovenous malformation (bAVM) commencing with a trigger event and ending in a stable bAVM, intracranial hemorrhage (ICH), or bAVM occlusion. The progress from high laminar wall shear stress (LWSS) through to normal LWSS (after dilatation, recruitment, and angiogenesis) and a stable bAVM. With the development of venous intimal hyperplasia resulting in venous outflow obstruction that results in a reduction in LWSS and an increase in pressure upstream (within the abnormal vasculature). This will lead to endothelial oxidative stress and the changes that can lead to vascular wall degeneration challenged by the increase in pressure within these vessels. Measuring LWSS may prove to be an opportunity to more accurately predict the risk for an individual with a bAVM of ICH.

2.3 Selected Papers on the Treatment Options for Unruptured Brain AVM

●Spetzler RF, Ponce FA. A 3-tier classification of cerebral arteriovenous malformations. Clinical article. J Neurosurg 2011;114(3):842–849

●Patel NJ, Bervini D, Eftekhar B, et al. Results of surgery for low-grade brain arteriovenous malformation resection by early career neurosurgeons: an observational study. Neurosurgery 2019;84(3):655–661

2.4 Treatment Options for Unruptured Brain AVM

Cushing and Bailey described the first resection of an unruptured bAVM (a probable Spetzler–Martin Grade [SMG] III).36 The patient was treated by radiotherapy through a bone window created for surgery on a 5-cm dominant central sulcus convexity bAVM in 1924. At the time of surgery, Cushing judged that it was “inconceivable that [the bAVM] could be attacked without fatal hemorrhage” and had the patient treated with radiotherapy. Surgery on this patient was repeated by Cushing in 1927, following which the patient developed progressive dysphasia. At this second surgery, Cushing and Bailey³⁶ reported:

“The lesion proved to be of stony hardness. The tangle of pulsating vessels previously encountered was largely thrombosed and transformed into a multitude of small bloodless shreds… On finally tilting the growth out from its deep pocket, there passed …a large thick-walled, partly calcified vessel almost the size of a lead pencil…there was nothing to do but to ligate the vessel and to remove the lesion…. As a result of this operation the patient was rendered aphasic and hemiplegic.”

Cushing’s management decisions, surgical performance, and consequent outcome remain current. The decision to treat a 64-year-old patient with unruptured SMG III bAVM with combined therapy with an adverse outcome from each treatment would be controversial today. While treatment at this time may be considered controversial, it was based on a misunderstanding of the natural history of bAVM. Norlen asserted, in 1949, that “probably most, if not all, patients die of hemorrhage or are completely incapacitated.”37 Despite the sinister view of the natural history, the opinion of the role of surgery was equally dire. Walter Dandy’s review in 1928 of his clinical experience concluded that “the radical attempt at cure is attended by such supreme difficulties and is so exceedingly dangerous as to be contraindicated except in certain selected cases.”³⁸ Radiotherapy outcomes were also initially considered fruitless. Olivecrona and Riives concluded that there is no “proof that Roentgen treatment…in any way alters the spontaneous course (of bAVM ICH).”³⁹ What progress has been made in our understanding of various management pathways over the last nearly 100 years of attempted treatments?

The highest level of evidence upon which to base management recommendations is the randomized controlled trial (RCT). Such a trial for unruptured bAVM, the “A Randomized trial of Unruptured Brain Arteriovenous malformations” (ARUBA) found in favor of medical management in comparison to intervention with intent to cure.⁴⁰ However, lumping the three intent-to-cure methods together fails to acknowledge that there is in fact multiple treatment pathways rather than a dichotomy of pathways to consider for unruptured bAVM: conservative treatment, endovascular treatment, radiosurgery, surgery (with or without planned preoperative embolization), and a combination of the aforementioned treatments. Because of the likelihood of heterogeneity with regard to adverse outcomes among the treatment methods, it is reasonable to conclude that ARUBA has not resolved the question of management despite the expertise of the ARUBA researchers.⁴¹ The failure to identify the best management pathway is largely related to the small number of patients, which reflects the rarity of the disease (unruptured bAVMs are diagnosed in < 1 per 100,000 per year)⁴²^, ⁴³; the ethical dilemma in denying treatment to patients whose risk of treatment was perceived to be low; and randomizing patients with immediate high-stakes outcomes. By contrast, many pharmaceutical trials allow a crossover in management after the end of the trial period once the best course is known.⁴¹

In the absence of an RCT, what is the best evidence upon which management decisions can be made? Case-controlled series can be a good basis upon which to make decisions providing patients included in the analysis can be seen to be a reasonable representation of all similar cases both treated and untreated (i.e., there is an explanation of how selection bias may have impacted the results); can be reliably divided into risk categories that can be reliably reproduced by others (e.g., grading systems); have sufficient numbers that there is a great deal of confidence in the reported outcomes; and results can be reliably demonstrated to be reproduced by others. In order to compare different treatments from such case series, there must be an acceptable categorization of cases by grade of risk, as well as outcomes, that when applied across different treatments is reliable. Fortunately, such conditions can be met for some of the treatments.

The Spetzler–Martin grading system, developed for risk assessment of surgery, has gained the widest acceptance and has been validated independently by a number of authors for surgery, radiosurgery, and embolization.⁴⁴^, ⁴⁵^, ⁴⁶^, ⁴⁷^, ⁴⁸^, ⁴⁹^, ⁵⁰ SMG was established by allocating points for size (1 for maximum diameter < 3 cm, 2 for maximum diameter between 3 and 6 cm, and 3 for maximum diameter > 6 cm), the presence of DVD (adding 1 point if present), and location in eloquent brain (adding 1 point as defined earlier).⁴⁴ The SPC categories for bAVM are derived from combining SMG 1 and 2 into SPC class A, SMG 3 into SPC class B, and SMG 4 and 5 into SPC class C.⁴⁵ The inter- and intraobserver reliability of the SMG evaluation of the score is generally reasonable.⁴⁴^, ⁵¹^, ⁵² Unoperated cases have been incorporated into a sensitivity analysis to confirm that maximum diameter, eloquent location, DVD, and presentation in the absence of hemorrhage remain as risk factors, confirming that these factors are valid for use in risk assessment classification scales.¹^, ⁴⁶ The relative impact of these factors is high, with an odds ratio of 1.8 per cm (95% CI: 1.5–2.2) for maximum bAVM diameter as a continuous variable, 5.4 (95% CI: 2.5–11.5) for DVD, and 4.4 (95% CI: 2.1–9.2) for eloquent location.¹ Taking into consideration the size grouping in the SMG, the relative risk as depicted by the odds ratio of each of these variables is well represented by their weighting of the SMG points. Therefore, there should be a high degree of confidence as to their reliability and validity of SMG and SPC to predict risk of surgery.

The SMG is useful in a number of ways. These include audit, comparing cohorts, and informing recommendation of treatment. For the first two of these uses, it is common to report outcomes in a binary fashion (e.g., modified Rankin scale [mRS] > 1 or < 2) in order to allow for statistical analysis. However, with regard to its use in the recommendation of treatment, it is only a good first step of many. Adverse outcomes are diverse in both what is affected and how profound is the effect. For one patient, any permanent neurological impairment that may follow immediately from surgery would be considered unacceptable, but for another, avoidance of early death from subsequent bAVM ICH may be given greater priority.

Despite these considerations, the SMG is a useful first step in estimating the risk of surgical treatment and they can reliably predict the risk of new permanent neurological deficits.⁴⁴^, ⁴⁵ However, for the purpose of comparing treatments, new permanent neurological deficits are only one aspect of the sum of unFOs. Other considerations that need to be incorporated in the assessment of an unFO include failure to occlude along with delayed ICH following treatment. Furthermore, because of the latency period until occlusion, time needs to elapse for the results of radiosurgery to report unFOs. Large cohort series that have incorporated selection bias for surgery, and multicenter for radiosurgery, facilitate such comparison by having a common definition of unFO and have divided cases by SMG.⁴⁷^, ⁵³^, ⁵⁴

There are two assessment instruments specifically developed with the intent of predicting unFOs:

●The modified Pollock–Flickinger (mPF) score⁵⁴:

mPF score = 0.1 × Volume (mL) + 0.02 × age (years) + 0.5

(if located in basal ganglia, thalamus, or brainstem).

●The VRAS⁴⁷:
VRAS score = the sum of points awarded to bAVM volume (2–4 cm³ = 1 point; smaller = 0 points and larger =2 points) with additional 1 point awarded for each if there was a history of hemorrhage or the hemorrhage occurred in eloquent brain (as defined by the SMG defined below).

Both these scores correlate well with predicting unFOs after radiosurgery. For patients followed for more than 12 months (patients with a complication or hemorrhage occurring in the first year were also included) for a mean of 7 years, these are excellent methods of communicating expectations of radiosurgery.47 In addition, the SMG also correlated well with unFOs for radiosurgery.⁴⁷^, ⁴⁸^, ⁵⁵ Therefore, the SMG and the outcomes reported with respect to unFOs are a good basis upon which to commence with decision-making for the individual patient as it allows some comparison between the radiosurgery and surgery.

2.4.1 Embolization

Embolization continues to evolve rapidly. These changes can be explained by an increased experience with ethylene vinyl acetyl copolymer, innovation in catheter technology, and the introduction of the promising role of venous approaches.⁴⁹^, ⁵⁰^, ⁵⁶^, ⁵⁷ There is some support for selected cases to be treated by embolization with an intent to cure.⁴⁹^, ⁵⁰ However, because of the results from ARUBA, caution needs to be exercised with its widespread adoption outside those few highly specialized centers that are achieving outstanding results and are exploring the place of embolization as an intent-to-cure treatment.⁴⁰^, ⁵⁶ In the treatment arm of ARUBA, the breakdown of the adverse outcomes by treatment modality was not published. However, 62% of the intention-to-treat arm included cases for embolization (32% as the only treatment).⁴⁰ There was only a small number of surgical cases (5% surgery alone and 14% embolization followed by surgery) and a low likelihood that a significant number of complications would have occurred in those undergoing focused irradiation during a mean follow-up of 33 months. Therefore, it is reasonable to speculate that embolization was a major cause of adverse outcomes (occurring in 36 cases, i.e., 37% of all treated patients) in the treatment arm. A conclusion that may be drawn is that the role of embolization in unruptured bAVM is questionable if alternate management options are reasonable.⁵⁶

Despite the above note of caution, there are two case series of more than 100 treated patients that report (for those completing treatment) obliteration for SPC class A bAVM of 92 and 98% with a decline in neurological condition of 6 and 2.5%.⁴⁹^, ⁵⁰ However, not all cases in these series have completed treatment (reported as 19% for one of the series and an exclusion criterion from any form of analysis in the other). This allows for the possibility that the overall results may be worse than those reported for the completed treatment group.

Barring advances in technique, the role for embolization with intent to cure for SPC classes greater than A remains limited. The best results from a center that attempts maximum treatment with embolization reported an 11% new permanent neurological deficit rate with a minority of cases cured by embolization.⁴⁹ This is a deceptive figure given there will be a tradeoff between complications of embolization with effectiveness of treatment as well as the exclusion of cases not considered treatable by embolization.⁵⁰^, ⁵⁶

The role for embolization as a preoperative treatment is discussed in the section “Radiosurgery.”

2.4.2 Radiosurgery

The predictability of the physics and radiobiology of radiosurgery makes risk prediction for radiosurgery reliable.⁴⁷^, ⁵⁴ A consideration of radiosurgery needs to include the risks of radiation-related complications, time delay between treatment and cure (with a continuing risk of hemorrhage during this latency period), and the likelihood of cure. Because of the interplay between these variables (e.g., a higher marginal dose reduces the time to obliteration but may be at the expense of increasing radiation-induced complications), an overall assessment of unFO is a good method of looking at the overall expected outcome.⁵⁴ With the more consistent and reliable principles of the physics of radiosurgery than the much more operator-dependent approaches of surgery or embolization, it is not surprising that there has been only modest improvement in obliteration rate among experts and experienced radiosurgical centers with time.⁵⁵

The mPF score, at a mean follow-up of 70 months, predicts an unFO in approximately 10% for scores ≤ 1, 30% for scores higher than 1 and ≤ 1.5, 40% for scores higher than 1.5 and ≤ 2, and more than 50% for scores higher than 2.⁵⁴ In order to roughly calculate volume, the three orthogonal axes (that include the largest diameter) can be estimated by the following equation: