Treatment of Brain Arteriovenous Malformations: A Systematic Review and Meta-analysis

Context  Outcomes following treatment of brain arteriovenous malformations (AVMs) with microsurgery, embolization, stereotactic radiosurgery (SRS), or combinations vary greatly between studies.

Objectives  To assess rates of case fatality, long-term risk of hemorrhage, complications, and successful obliteration of brain AVMs after interventional treatment and to assess determinants of these outcomes.

Data Sources  We searched PubMed and EMBASE to March 1, 2011, and hand-searched 6 journals from January 2000 until March 2011.

Study Selection and Data Extraction  We identified studies fulfilling predefined inclusion criteria. We used Poisson regression analyses to explore associations of patient and study characteristics with case fatality, complications, long-term risk of hemorrhage, and successful brain AVM obliteration.

Data Synthesis  We identified 137 observational studies including 142 cohorts, totaling 13 698 patients and 46 314 patient-years of follow-up. Case fatality was 0.68 (95% CI, 0.61-0.76) per 100 person-years overall, 1.1 (95% CI, 0.87-1.3; n = 2549) after microsurgery, 0.50 (95% CI, 0.43-0.58; n = 9436) after SRS, and 0.96 (95% CI, 0.67-1.4; n = 1019) after embolization. Intracranial hemorrhage rates were 1.4 (95% CI, 1.3-1.5) per 100 person-years overall, 0.18 (95% CI, 0.10-0.30) after microsurgery, 1.7 (95% CI, 1.5-1.8) after SRS, and 1.7 (95% CI, 1.3-2.3) after embolization. More recent studies were associated with lower case-fatality rates (rate ratio [RR], 0.972; 95% CI, 0.955-0.989) but increased rates of hemorrhage (RR, 1.02; 95% CI, 1.00-1.03). Male sex (RR, 0.964; 95% CI, 0.945-0.984), small brain AVMs (RR, 0.988; 95% CI, 0.981-0.995), and those with strictly deep venous drainage (RR, 0.975; 95% CI, 0.960-0.990) were associated with lower case fatality. Lower hemorrhage rates were associated with male sex (RR, 0.976, 95% CI, 0.964-0.988), small brain AVMs (RR, 0.988, 95% CI, 0.980-0.996), and brain AVMs with deep venous drainage (0.982, 95% CI, 0.969-0.996). Complications leading to permanent neurological deficits or death occurred in a median 7.4% (range, 0%-40%) of patients after microsurgery, 5.1% (range, 0%-21%) after SRS, and 6.6% (range, 0%-28%) after embolization. Successful brain AVM obliteration was achieved in 96% (range, 0%-100%) of patients after microsurgery, 38% (range, 0%-75%) after SRS, and 13% (range, 0%-94%) after embolization.

Conclusions  Although case fatality after treatment has decreased over time, treatment of brain AVM remains associated with considerable risks and incomplete efficacy. Randomized controlled trials comparing different treatment modalities appear justified.

Brain arteriovenous malformations (AVMs) are abnormal connections between arteries and veins leading to arteriovenous shunting with an intervening network of vessels—the so-called nidus.¹ Brain AVM prevalence varies between 15 and 18 per 100 000 adults.² In approximately 0.05% of the population, they are incidental findings on brain magnetic resonance imaging screens.³ Their overall detection rate is 1 per 100 000 adults per year.⁴ Roughly half of patients with brain AVMs present with intracranial hemorrhage, resulting in a first-ever hemorrhage rate of 0.55 per 100 000 person-years.⁴ Little is known about the etiology of brain AVMs, but the etiology is likely to be multifactorial.⁵

A commonly used grading scale for brain AVMs is the Spetzler-Martin Grade (SMG) scale, which is a composite score of nidus size (<3 cm, 3-6 cm, >6 cm; 1-3 points), eloquence of adjacent brain (1 point if located in brainstem, thalamus, hypothalamus, cerebellar peduncles, or sensorimotor, language, or primary visual cortex), and presence of deep venous drainage (1 point if any or all drainage is through deep veins, such as internal cerebral veins, basal veins, or precentral cerebellar veins).⁶ The risk of subsequent hemorrhage is increased when the brain AVM presents with hemorrhage^7-9 or deep venous drainage,^7,8 when it is associated with aneurysms,⁸ or when it is in a deep location.^7,9 The annual hemorrhage risk may be as low as 0.9% in patients with unruptured, superficially located brain AVMs with superficial drainage and may be as high as 34% in patients with ruptured, deeply seated brain AVMs with deep venous drainage.⁷Quiz Ref IDBecause ruptured brain AVMs presumably have a higher hemorrhage risk (4.5%-34%) than previously unruptured ones (0.9%-8%),⁷ interventional treatment of ruptured brain AVMs is advisable,^10-12 despite the absence of evidence from randomized controlled trials (RCTs) that the benefits outweigh the risks.¹³

Quiz Ref IDThe goal of brain AVM treatment is typically the prevention of hemorrhage. However, seizure control or stabilization of progressive neurological deficits are occasionally treatment goals.¹⁰ Microsurgery (ie, craniotomy followed by resection) has been reported to have a low risk of complications for in SMG I and II brain AVMs (eg, small malformations in noneloquent areas) and result in immediate cure. However, microsurgery is invasive. Quiz Ref IDStereotactic radiosurgery (SRS), focused irradiation, can be effective for malformations that are smaller than 3.5 cm, but complete obliteration requires approximately 1 to 3 years after treatment and cure is not always obtained. Delayed complications such as hemorrhage in the latency period and radiation edema or necrosis can occur as late complications.

Embolization is used to obliterate small malformations or to make larger malformations amenable for (radio)surgery, or to eliminate a possible cause of hemorrhage (eg, associated aneurysms). For embolization, microcatheters are used to deliver embolic materials to feeding arteries or the nidus. Brain AVMs categorized as either SMG IV or V generally require multimodality treatment.^10,12 Variations in practice exist, due to referral practices, availability of technical equipment and expertise, personal preference, policies of health insurance systems, and absence of RCTs comparing treatment modalities.^13-15

In this systematic review and meta-analysis, we report rates of case fatality, long-term risk of hemorrhage, risk of complications, and successful obliteration after microsurgery, embolization, or SRS. Clinical characteristics associated with these outcomes are also reported for each treatment modality.

Dr van Beijnum searched PubMed and EMBASE up to March 1, 2011, using electronic search strategies for brain arteriovenous malformations (intracranial arteriovenous malformations /brain arteriovenous malformation) AND terms for treatment (surgery or radiosurgery or therapeutic embolization /artificial embolism or combined modality therapy). She also hand searched the following journals: Journal of Neurosurgery, Neurosurgery, Acta Neurochirurgica, American Journal of Neuroradiology, Stroke, and Neurology from January 2000 until March 2011. When multiple publications arose from the same cohort, we included the largest cohort.

Drs van Beijnum and Klijn, with 1 of 4 authors serving as a second reader, reviewed and extracted data on inclusion and exclusion criteria, study design, whether outcome assessment was performed independently of the treating team, whether standardized outcome scales were used, patient and brain AVM characteristics (Box), treatments, and outcomes using a structured data extraction form. Drs Van Beijunum and Klijn resolved discrepancies between reviewers in a consensus meeting.

Box Section Ref ID

The primary outcome was all-cause case fatality during treatment and follow-up. The secondary outcome was intracranial hemorrhage beyond 30 days of treatment (hemorrhage within 30 days was categorized as a complication of treatment). The tertiary outcomes were treatment complications and successful obliteration of the brain AVMs. Acute complications were defined as events that occurred within 30 days after treatment (eg, intracranial hemorrhage, ischemic stroke, intracranial infection, or hydrocephalus). Because the effect of SRS becomes apparent over time, we also recorded late complications occurring beyond 30 days for radiosurgical cohorts, including intracranial hemorrhage as a complication of the SRS, radiation edema, necrosis, or cyst formation. In addition, we assessed the proportion of patients with severe complications, defined as those leading to death or permanent neurological deficits. The proportion successfully obliterated was defined based on the presence of angiographically demonstrated obliteration and presented as a proportion of all treated patients.

In all analyses, we present results only for those characteristics that were reported in at least 5 cohorts.

We calculated case fatality and hemorrhage rates per 100 person-years for all cohorts and for each treatment modality separately with the number of outcomes and the number of person-years for each study as parameters in Poisson regression models. For assessment of the overall incidence rate, the intercept of the model was used. We assessed the relationship of summary measures of patient and brain AVM characteristics (ie, proportions of these characteristics) to each outcomes by calculating adjusted rate ratios (RRs) with corresponding 95% confidence intervals for all treatments together. We adjusted for the following prespecified factors: mid-year (the median of the range of years during which the study was conducted), patient age, the proportion of male patients, and the proportion of ruptured brain AVMs. Adjusted RRs are expressed per 1% increase in the proportion of patients with a study characteristic or per 1-year increase in age or midyear. If the confidence intervals did not include 1.0, we considered the result statistically significant.

Subsequently, we repeated our analyses for each treatment modality separately. For overall case fatality and hemorrhage rates, we performed prespecified sensitivity analyses in high-quality studies, which were defined as either a prospective study design or an independent outcome assessment. We also studied associations between study and participant characteristics and early case fatality. To assess heterogeneity, we calculated the quantity I²¹⁶ for case fatality and hemorrhage rate in all cohorts and in the 3 modalities separately.

We used Poisson regression analysis to investigate the associations of study characteristics on the occurrence of complications and on the proportion of patients in whom successful obliteration was achieved after each of the 3 treatment modalities.

We used SPSS Version 16.0 for all statistical analyses.

The literature search identified 137 articles reporting on 142 cohorts including 13 698 patients with a total of 46 314 patient-years of follow-up (eFigure 1, as are the electronic references that follow). Forty-one cohorts (29%) reported on microsurgery,^e1-40 14 (10%) on embolization,^e113-126 69 (48%) on SRS,^e41-105 7 (5%) on fractionated radiotherapy,^e106-112 and 11 (8%) described either multimodality treatment,^e132,e134-7 various treatments within 1 article,^{e127-8,e130-1,e133} or another treatment such as intraoperative embolization.^e129

Patient and brain AVM characteristics, treatments, and outcomes for all cohorts are listed separately in eTable 1 and summarized in Table 1 and Table 2. Of all study cohorts, none was an RCT, 23 (16%) were prospective cohorts, and 10 (7%) described independent assessment of outcome, either clinical (n = 6), radiological (n = 1), or both (n = 3). In all cohorts, I² values were 42% for case fatality and 54% for hemorrhage rate, for surgical cohorts, 0% and 0%, for radiosurgical cohorts, 30% and 66%, and for embolization cohorts 46% and 0%, respectively.

Case-fatality rates were 0.68 (95% CI, 0.61-0.76) per 100 person-years, 1.1 (95% CI, 0.87-1.3) after microsurgery, 0.50 (95% CI, 0.43-0.58) after SRS, and 0.96 (95% CI, 0.67-1.4) after embolization. Hemorrhage rates were 1.4 (95% CI, 1.3-1.5) per 100 person-years, 0.18 (95% CI, 0.10-0.30) after microsurgery, 1.7 (95% CI, 1.5-1.8) after SRS, and 1.7 (95% CI, 1.3-2.3) after embolization. Complications leading to permanent neurological deficits or death occurred in a median of 7.4% (range, 0%-40%) of patients after microsurgery, in 5.1% (range, 0%-21%) after SRS, and in 6.6% (range, 0%-18%) after embolization. Obliteration was achieved in 96% (range, 0%-100%) of patients after microsurgery, in 38% (range, 0%-75%) after SRS, and in 13% (range, 0%-94%) after embolization.

Adjusted RRs for the association of study characteristics on case fatality are summarized in Table 3 and hemorrhage rate in Table 4. More recent studies were associated with lower case fatality but higher rates of hemorrhage. Male sex and small brain AVMs or brain AVMs with strictly deep venous drainage were associated with both lower case fatality and lower hemorrhage rates. Analyses limited to high-quality studies showed essentially the same results, but in these analyses older age was also associated with a higher case fatality (eTable 2).

Older age, and a larger proportion of male patients, ruptured brain AVMs, and eloquent brain AVMs were associated with lower complication rates (Table3 and Table 4; eTable 3 and eTable 4). More recent cohorts and increasing proportions of deep and of SMG I through III brain AVMs were associated with higher complication rates (eTable 3).

Twenty-two of 69 cohorts (32%) reported on Gamma Knife SRS, and 36 (52%) on linear accelerator SRS. The remaining 11 cohorts (14%) reported on other radiosurgical modalities (n = 5), multiple radiosurgical modalities within 1 article (n = 5), or did not mention the modality (n = 1).

Increasing proportions of male patients, SMG I through III brain AVMs, small nidus size, and increasing margin dose were associated with lower case fatality and lower hemorrhage rate (Table 3 and Table 4). Preradiosurgical embolization was associated not only with an increased hemorrhage rate and higher risk of complications but also with a higher chance of obliteration (Table 4; eTable 3 and eTable 4).

The proportion of acute complications after SRS was low (Table 1). A higher risk of late complications was observed in earlier cohorts, in cohorts with older age and larger nidus size and in cohorts with higher proportions of ruptured brain AVMs, of high-grade brain AVMs, or of eloquent brain AVMs (eTable 3). Because of contradicting findings regarding the association of infratentorial location for complications, we repeated multivariable analysis of all late complications in 17 cohorts that also reported late severe complications (RR 1.00, 95% CI, 0.997-1.010).

The chance of brain AVM obliteration was higher in more recent cohorts and in cohorts with younger patients or with more unruptured brain AVMs (eTable 4).

Ten of 14 cohorts (71%) reported the number of embolization sessions (median 2, range 1-11). A larger proportion of SMG I through III brain AVMs was associated with lower case-fatality, and a larger proportion of brain AVMs with strictly deep venous drainage with lower hemorrhage rate (Table 3 and Table 4). For the analyses of complications, 4 additional articles could be included that separately reported complications after embolization as part of multimodality treatment.^{e37,e39,e91,e137} A more recent midyear, younger age, higher proportions of women, strictly deep venous drainage, smaller nidus size, and increasing number of embolization sessions were associated with a lower risk of complications (eTable 3). Higher obliteration rate was associated with more recent cohorts and with previously ruptured brain AVMs (eTable 4).

In this systematic review of brain AVM treatment, we found that male sex, small brain AVMs, and brain AVMs with strictly deep venous drainage were associated with both lower case-fatality and lower hemorrhage rates. Younger age and brain AVMs with SMG I through III were associated with lower case fatality, whereas lower proportions of eloquent brain AVMs and higher proportions of obliterated brain AVMs were associated with lower hemorrhage rates. More recent studies were associated with lower case fatality after brain AVM treatment but higher rates of hemorrhage.

Case fatality was low in patients treated with SRS. Hemorrhage rates during follow-up were high after SRS and embolization, and low after microsurgery. All 3 treatment modalities had considerable risks of severe complications. The proportion of patients in whom complete obliteration was obtained after treatment was high after microsurgery. Embolization alone achieved obliteration in a minority of patients. However, it is important to point out that characteristics of the brain AVM direct the choice for the type of treatment and none of the data reported herein are from randomized controlled trials.

Because all included studies were observational, without direct comparisons between treatment modalities, our results cannot be used to compare the risks and benefits of the treatments. Nevertheless, the quantitative information that we report is valuable to inform medical professionals and patients and to devise the design of future RCTs.

Quiz Ref IDThe decrease in case fatality associated with more recent procedures suggests that technical advances and increasing experience have improved outcome after treatment. The increased hemorrhage rate over time is probably caused by the increasing number of patients treated with SRS, in which obliteration rate is relatively low and hemorrhage rate is high in comparison with microsurgery. Overall complication rates after SRS and embolization decreased over time. The overall complication rate increased over time after microsurgery but not the rate of severe complications. Possible complications have been described more meticulously in more recent studies.^17-20

Older age was associated with higher case fatality but not with a higher risk of hemorrhage. The finding that older age was associated with both a lower complication rate after microsurgery and a higher obliteration rate after embolization might be explained by patient selection. Older age was also associated with a higher risk of late complications and a decreased chance of successful obliteration after SRS. This observation is in accordance with the radiosurgery-based brain AVM score,^21,22 which predicts occurrence of new neurological deficits and obliteration after SRS. Male sex was associated with lower case fatality and lower hemorrhage rates in the overall analyses and after SRS. Several studies have suggested that SRS may be less effective in women,^23-26 particularly in premenopausal women.²⁷

A higher proportion of ruptured brain AVMs was associated with a lower risk of surgical complications but was not associated with case fatality or hemorrhage rates. Adverse outcomes after microsurgery may have a more detrimental effect on patients with unruptured brain AVMs than on those who already experienced intracerebral hemorrhage.²⁸ Recently, an adapted scale was proposed including unruptured presentation as a predictor of increased disability.²⁹ The association between increasing proportion of ruptured brain AVMs and increasing risk of severe late complications after SRS and increasing risk of all complications after embolization contrasts with that of previous studies on predictors of outcome after SRS^11,26,30,31 or after embolization,^32,33 which found none of these associations. Our meta-analysis shows a reduced chance of obliteration after SRS when the proportion of ruptured brain AVMs increases. This also contrasts with previous studies.^24,26,34 The lower rate of obliteration in our meta-analysis may be explained by our strict definition of obliteration based on conventional angiography.

We found that increasing nidus size was associated with increasing case fatality and hemorrhage rates overall and after SRS. Previous studies also showed that smaller brain AVMs have a higher chance of obliteration after SRS along with a lower risk of recurrent hemorrhage.^24,35,36 However, our analyses regarding brain AVM size and obliteration yielded contradictory findings, possibly due to the small number of cohorts that reported size.

A higher proportion of low-grade brain AVMs was associated with decreased case fatality overall and after SRS and embolization but not after microsurgery. Also, low-grade brain AVMs were associated with a lower hemorrhage rate after SRS. In this meta-analysis, the surgical risk was not associated with the proportion of patients with low-grade brain AVMs in the included studies. This may also be explained by selection bias in the current report.

Strictly deep venous drainage has been reported as an independent predictor of brain AVM hemorrhage before treatment, during follow-up without invasive treatment,⁷ and with increased surgical risk.⁶ The finding in this meta-analysis that strictly deep venous drainage was associated with lower case-fatality and hemorrhage rates in all cohorts combined, and a lower hemorrhage rate and overall complication rate after SRS and embolization may be caused by selection of patients for certain treatments based on the venous draining pattern.

Embolization before SRS was associated with an increased hemorrhage rate and an increased risk of complications. This higher hemorrhage rate after SRS may in part be explained by the fact that previously embolized brain AVMs are generally larger and obliteration may be obtained later or not at all. Moreover, it is important to emphasize that the combined risks of multimodality therapy are only presented in the overall analyses (Table 3 and Table 4) and in eTable 1. We were not able to provide reliable estimates of the risk of multimodality treatment. Multimodality treatment may be the safest approach for some brain AVMs, but it may also result in accumulation of risks of the various treatments involved.

Preradiosurgical embolization was associated with a higher chance of complete obliteration. Preradiosurgical embolization may obscure brain AVM nidus delineation, both by the superimposition of embolic material and the presence of collateral feeding vessels.³⁷ Unfavorable effects of preradiosurgical embolization on obliterated proportions have been reported.^30,38,39 However, embolization is also performed prior to SRS to eradicate potential sources of hemorrhage, such as aneurysms or venous ectasia.¹⁰ Proximal flow-related aneurysms (ie, proximal on artery which supplies brain AVM) rarely regress with brain AVM treatment whereas distal flow-related aneurysms (ie, distal on feeding artery) have a high prospect of regressing with obliteration of the brain AVM nidus.⁴⁰

Quiz Ref IDAn important limitation of our meta-analysis is that the large majority of studies were retrospective and without independent outcome assessment. In addition, none of the studies compared treatments in a randomized design. Selection of patients for certain treatment options is likely to have been influenced by patient and brain AVM characteristics and, therefore, hampers comparison of treatments. In most analyses, heterogeneity was absent or at most moderate. The highest I² value was found in hemorrhage rate within radiosurgical cohorts. Additionally, important criticisms of meta-analysis also apply to ours: due to the combination of different kinds of studies the results are difficult to apply to the individual patient. In addition, we may have excluded studies that may be important but did not fulfill all inclusion criteria.⁴¹

The strength of our meta-analysis is that we studied associations of time trends and outcomes after brain AVM treatment. We also studied associations of important patient and brain AVM characteristics with outcomes in a large number of cohorts. In this review, microsurgery was associated with lower risk of (recurrent) hemorrhage. Case fatality after microsurgery may be higher than SRS due to selection bias because in patients who present with hemorrhage, surgery is more often performed in the acute phase,^42-45 whereas SRS is often only performed on patients who survive hemorrhage in a reasonable condition. Stereotactic radiosurgery was associated with a low case fatality at the expense of a low obliteration rate. The latter may in part be explained by our strict definition of obliteration and may reflect incomplete follow-up. In agreement with our findings, some authors have suggested that the 2-year obliteration rate after SRS is probably in the range of 40% rather than the commonly cited 80%.⁴⁶ Embolization alone achieves obliteration in a minority of patients and is mostly applied in conjunction with other modalities.

Despite the observed decrease in case fatality in more recent studies, treatment of brain AVMs remains associated with a substantial risk of death or disability, no matter which modality is chosen. Treatment of unruptured brain AVMs has been associated with poor outcome.^47,48 The ongoing ARUBA (A Randomized Multicenter Clinical Trial of Unruptured Brain AVMs) study compares the risks and benefits of conservative management and interventional treatment in patients with unruptured brain AVMs (http://www.arubastudy.org).

Randomized controlled trials dedicated to evaluate the safety and effectiveness of different modalities in brain AVM subgroups are needed, for example in patients with small brain AVMs that can be treated using any of the 3 modalities. Standardized (international) prospective registrations of conservative or interventional management of brain AVMs may provide more information for individual risk prediction.

Corresponding Author: Janneke van Beijnum, MD, Department of Neurosurgery, Leiden University Medical Center, PO Box 9600, 2300 RC, Leiden, the Netherlands (j.vanbeijnum@gmail.com).

Author Contributions: Dr van Beijnum had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: van Beijnum, van der Worp, Al-Shahi Salman, Kappelle, Rinkel, Berkelbach van der Sprenkel, Klijn.

Acquisition of data: van Beijnum, Buis, Al-Shahi Salman, Kappelle, Rinkel, Vandertop, Klijn.

Analysis and interpretation of data: van Beijnum, van der Worp, Al-Shahi Salman, Kappelle, Rinkel, Berkelbach van der Sprenkel, Algra, Klijn.

Drafting of the manuscript: van Beijnum, Klijn.

Critical revision of the manuscript for important intellectual content: van Beijnum, van der Worp, Buis, Al-Shahi Salman, Kappelle, Rinkel, Berkelbach van der Sprenkel, Vandertop, Algra, Klijn.

Statistical analysis: van Beijnum, Algra.

Obtained funding: van Beijnum, van der Worp, Al-Shahi Salman, Klijn.

Administrative, technical, or material support: van Beijnum.

Study supervision: van der Worp, Klijn.

Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest and none were reported.

Funding/Support: Dr van Beijnum was funded by the Netherlands Organization for Scientific Research (NWO); and by grant 2002B138 from the Netherlands Heart Foundation. Dr van der Worp is supported by grant 2010T075 from the Netherlands Heart Foundation. Dr Al-Shahi Salman was funded by grant G108/613 from the UK Medical Research Council (Clinician Scientist Fellowship). Dr Klijn was funded by grant 2007B048, the Netherlands Heart Association and by grant 907-00-103 from the Netherlands Organization for Health Research and Development.

Role of the Sponsor: The sponsors had no role in the study design, data collection, data analysis, data interpretation, or writing of the report.