Hemorrhage Risk of Cerebral Arteriovenous Malformations Before and During the Latency Period After Gamma Knife Radiosurgery

Influenced by the high morbidity and mortality associated with arterial aneurysms, neurosurgeons since the 1950s have been concerned that cerebral arteriovenous malformations (AVM) bear a high risk of hemorrhage.¹ Therefore, AVM were often treated after their diagnosis and this has limited the amount of research on the natural hemorrhage rate in untreated patients with AVM. The currently accepted 2% to 4% annual hemorrhage rates are based on limited numbers of long-term follow-up studies of untreated AVM^2–5 and some retrospective research evaluating the risk of hemorrhage between the diagnosis of the AVM and the time of intervention.^6–9

Since the first AVM treated with the Gamma Knife in 1970,¹⁰ hundreds of thousands of AVM had been treated with radiosurgery. Unlike microsurgery, the obliteration process occurs over time after radiosurgery and patients remain at risk for hemorrhage during the latency period of 2 to 3 years before the total obliteration of the nidus.¹¹ Compared with the natural hemorrhage rate, decreased,^12,13 unchanged,^14,15 or increased¹⁶ rates of hemorrhage after radiosurgery have been reported.

The aims of the present study are to calculate the hemorrhage rate in a large cohort of patients with AVM referred for Gamma Knife radiosurgery (GKS). We compare the hemorrhage rate after GKS with that of the natural history and the preradiosurgical hemorrhage rate in the same group of patients. Additionally, the risk factors of pre- and postradiosurgical AVM bleeding are analyzed.

Patients and Methods

Patient Population

A total number of 1204 patients with cerebral AVM were treated with Gamma Knife between 1989 and 2009 at the University of Virginia. Patients' demographics are depicted in Table 1. There were 637 males and 567 females with a mean age of 35.3 years. The most common presenting symptom leading to the diagnosis of AVM was hemorrhage. Fifty patients (8 Spetzler-Martin Grade I, 19 Grade II, 17 Grade III, and 8 Grade IV) who initially presented with symptoms other than hemorrhage experienced at least 1 episode of bleeding between the diagnosis of their AVM and GKS.

AVM Characteristics

The location and Spetzler-Martin Grading¹⁷ of the AVM are detailed in Table 1. The nidus volume ranged from 0.1 to 33 mL (mean, 3.4 mL). The nidus volume was <3 mL in 627 patients, between 3 and 5 mL in 298, between 5 and 10 mL in 249, and >10 mL in 30. The venous drainage was superficial only in 600 AVM nidi, deep only in 475, and both in 129.

Gamma Knife Technique

The details of the Gamma Knife procedure have been presented in a previous publication.¹⁸ In brief, a Leksell G-Frame was affixed to the patient's head. Stereotactic biplane angiography was then performed to delineate the nidus. Later with introduction of MRI, stereotactic MRI was also obtained for dose planning. The dose selection was based on the size and location of the nidus. The mean prescription dose to the margin of the nidus was 21.2 Gy (range, 5 to 36 Gy), and the mean maximum dose was 39.4 Gy (range, 10 to 60 Gy). The mean number of isocenters was 2.8 (range, 1 to 22).

Follow-Up

Patients underwent MRI every 6 months for 2 years and then yearly. Angiography was not performed until MRI showed an absence of flow voids. A CT or MRI was performed if patients developed new or worsened symptoms such as headaches, seizures, or neurological deficits. Hemorrhagic events in this study were defined as clinically symptomatic events along with signs of bleeding on CT or MRI as well as asymptomatic hemorrhage identified only on follow-up images as a recent bleed.

Statistic Analysis

The preradiosurgical annual hemorrhage rate was calculated as the number of hemorrhages divided by the patient-years at risk. Because AVM are generally considered as congenital anomalies, we assumed that patients were at risk for hemorrhage from their dates of birth. As an alternative metric, we also calculated the hemorrhage rate starting from the dates that AVM were diagnosed to the dates that AVM were treated. To calculate the annual rebleeding rate, the number of recurrent hemorrhages was divided by the total number of risk years from the dates of initial hemorrhage to the dates of GKS.

None of our patients had hemorrhage after the nidus was declared obliterated based on angiography; therefore, the postradiosurgical annual hemorrhage rate was calculated dividing the hemorrhagic events by the patient risk years from the dates of GKS to the dates that the AVM were judged to be obliterated on MRI or the dates of last follow-up if the nidi remained patent. We chose the dates that the AVM were concluded obliterated on MRI because angiography was usually performed after the MRI suggested that the nidi had obliterated. Furthermore, because the nidi should have obliterated before the final imaging studies, we assume that the AVM obliterated at the midpoint between the dates of the last images showing the AVM were still patent and the dates of the first images showing that the AVM obliterated. For hemorrhage-free survival, Kaplan-Meier analysis was used and log-rank test was applied to compare pre- and postradiosurgical hemorrhage risk.

Univariate and multivariate logistic regressions were applied to test the association of patient demographics (age, gender) and nidus characteristics (size, lobar versus deep locations, superficial only versus deep venous drainage, single versus multiple draining veins) with hemorrhagic or nonhemorrhagic presentation of the AVM. Univariate and multivariate Cox regression hazards model were used to test the risk factors of hemorrhage during the period from the diagnosis to the treatment of AVM and the follow-up period after GKS. Patients were censored from the postradiosurgery bleeding risk group when the AVM obliterated. In addition, patients were censored if they underwent further treatment, were lost to follow-up, or died. Additional factors evaluated from diagnosis to GKS of AVM included hemorrhagic or nonhemorrhagic presentation. Radiosurgical treatment parameters were added to evaluate the postradiosurgical risk of hemorrhage.

Results

Preradiosurgical Hemorrhage Rate and Rebleeding Rate

A total of 657 patients had 803 hemorrhagic events over 42 495 risk-years before GKS assuming that patients were at risk for hemorrhage since their birth. The annual hemorrhage rate was 2.0%. If we calculate the hemorrhage rate after the diagnosis of the AVM, there were 68 hemorrhagic events in 52 patients who initially presented with symptoms other than hemorrhage and 130 rebleeds in 90 patients who experienced hemorrhage before the diagnosis of AVM, yielding a hemorrhage rate of 6.6% (198 bleeds in 2984 risk-years).

The hemorrhage rate in the subgroup of 605 patients who initially presented with hemorrhage was 735 events divided by 19 712 risk years yielding a 3.7% annual hemorrhage rate assuming patients were at risk for hemorrhage since their birth. The hemorrhage rate would be 10.4% per year if we only calculated the 130 hemorrhagic events in 1243 risk-years from the diagnosis to GKS of AVM. In the subgroup of 599 patients who initially presented with symptoms other than hemorrhage, 50 patients experienced 68 episodes of hemorrhage within 1741 risk-years between the diagnosis and GKS of AVM. The hemorrhage rate was 3.9% (Table 2).

The annual rebleeding rate was 8.6% during the period from the diagnosis to GKS of AVM (146 rebleeds in 1581 risk-years). The rebleeding rate was 10.0% in the first year after the initial hemorrhage; 10.8% between 1 and 2 years; 6.5% between 2 and 5 years; 6.7% between 5 and ten 10; and 8.8% after 10 years. Of note, the mean time between the diagnosis and GKS of AVM was 2.5 years. Limited patients were available for calculating long-term rebleeding rate because they were referred for radiosurgery.

Patient Demographics and Nidus Characteristics in Hemorrhagic or Nonhemorrhagic Presentation

Patients tended to present with hemorrhage were those with young age, small nidi, deep AVM, nidi with deep venous drainage, or single draining veins (Table 3). The same risk factors were checked for the period between diagnosis of the AVM and GKS. Deep AVM and those with deep or single draining veins had a high risk of bleeding in univariate analysis. In multivariate analysis, a single draining vein was the only predicting factor for AVM rupture. After the diagnosis of AVM, small and deep AVMs were no longer the predicting factors for hemorrhage. Additionally, patients with prior hemorrhage were more likely to have recurrent hemorrhage.

Postradiosurgical Hemorrhage Rate

After GKS, 94 patients had 1 hemorrhage and 18 patients had 2 (Figure 1). The postradiosurgical hemorrhage rate was 2.5% (130 hemorrhagic events in 5239 risk-years), which was lower than the 6.6% hemorrhage rate if we calculate the rate since the diagnosis of AVM (Figure 2). The median hemorrhage-free survival was 14.2 years from diagnosis to GKS of AVM and 16.8 years during the latency period after GKS (P<0.001).

• Download figure
• Download PowerPoint

• Download figure
• Download PowerPoint

In the subgroup of patients with preradiosurgical hemorrhage, 55 patients had 1 hemorrhage and 12 patients had 2 after GKS. The postradiosurgical hemorrhage rate in this subgroup was 2.8% (79 hemorrhagic events in 2868 risk-years), which was lower than the preradiosurgical hemorrhage rate of 10.4% calculated after diagnosis of the AVM. The median hemorrhage-free survival was 10.1 years from diagnosis to GKS of AVM and 17.9 years during the latency period after GKS (P<0.001). In the subgroup of patients who never had hemorrhage before GKS, 51 hemorrhage occurred in 2371 risk-years yielding a annual hemorrhage rate of 2.2%, which is slightly lower than the 3.9% hemorrhage rate calculated after the diagnosis of the AVM. The median hemorrhage-free survival was 18.6 years from diagnosis to GKS of AVM and 20.0 years during the latency period after GKS (P=0.001).

Risk Factors Predicting Hemorrhage After GKS

Risk factors of postradiosurgical hemorrhage were deep AVM, large nidus size, and low prescription dose in univariate analysis. In multivariate analysis, only a low prescription dose remained to be associated with a high risk of hemorrhage (Table 3).

Discussion

Natural Hemorrhage Rate of AVM

AVM had traditionally been considered a potentially fatal disease and most patients were treated with microsurgery, embolization, or radiosurgery. Ondra et al published their landmark study that prospectively followed untreated patients with AVM for 24 years and reported an annual bleeding rate of 4%.³ Additional studies from Crawford et al² (2% annual hemorrhage rate), Itoyama et al⁴ (2.1%), and Yamane et al⁵ (4.2%) led to the generally accepted 2% to 4% natural hemorrhage rate of cerebral AVM. Although the conclusion from these studies demonstrated that the risks of AVM hemorrhage are not as high as arterial aneurysms, the cumulative risk of hemorrhage remains significant because the disease usually manifests itself in adolescents.

Retrospective studies to calculate the hemorrhage risk of the AVM usually face the problem of determining since when the patients start to be at risk for hemorrhage. Although AVM are considered congenital anomalies and theoretically patients are at risk for hemorrhage since their birth, hemorrhage usually occurs in adolescence or young adulthood.¹⁹ It is believed that AVM are dynamic lesions with ongoing morphological and hemodynamic changes and some unknown factors might play a significant role causing rupture of nidi later in the progression of the vascular malformations.

Because it is not known when the patients with AVM start to be at risk for hemorrhage and most of the patients received treatment after the diagnosis, the alternative to determine the bleeding risk of AVM would be to calculate the hemorrhage rate after the diagnosis of the AVM until the dates of intervention.^8,9,13,20 This calculation can be severely biased. As most studies have shown,^8,9,20 AVM tend to rebleed for some period of time after prior hemorrhage and additionally treatment was usually prompted by hemorrhage. The high numbers of hemorrhagic events after the initial bleeding and a prematurely terminated follow-up between the diagnosis of AVM and intervention will erroneously overestimate the hemorrhage rate.^13,21

The hemorrhage rate before treatment in the present series is 2.0% if we assumed that patients were at risk for hemorrhage since their birth. If the hemorrhage rate were to be calculated after the diagnosis of the AVM, the hemorrhage rate would be 6.6%. The hemorrhage rate even increased to 10.5% in the subgroup of patients who had a hemorrhagic presentation. The last two numbers definitely need to be interpreted with caution.

Hemorrhage Rates of Patients With AVM With or Without Prior Bleeding

It has been reported that patients would be at a higher risk of hemorrhage after a prior one.^2,7 Forester reported that patients who had bled once had a 25% chance of rebleeding in 4 years and those who had bled twice had a 25% chance of rebleeding within 1 year.²² In the present series, the hemorrhage rate after initial bleeding was 8.6%. The rebleeding rate was approximately 10% in the first and second years and reduced slightly thereafter. Ondra et al, however, observed no difference in hemorrhage rate in patients who initially presented with hemorrhage or other symptoms.³

In contrast to patients with AVM with a history of hemorrhage, the natural history of unruptured AVM is largely unknown. Brown et al followed a group of 168 patients with unruptured AVM after a mean of 8.2 years and reported a 2.2% annual hemorrhage rate.⁶ None of their patients had a recurrent hemorrhage. Recent natural history data suggest that the annual hemorrhage rate of unruptured AVM may be as low as 1%.^23,24 In our series, the hemorrhage rate of AVM diagnosed due to symptoms other than hemorrhage was 3.9% until the time of GKS. The rebleeding rates were similar in patients with or without hemorrhage as the initial presentation.

Risk Factors of Hemorrhage in AVM

Contradictory observations have been seen in different studies regarding whether age,^2,7 size,² or pregnancy² is related to a high risk of AVM rupture. In general, most studies have agreed that AVM in deep locations² and nidi with certain angioarchitectural features (deep draining veins, single draining veins, venous outlet stenosis) are associated with a high risk of hemorrhage. In the present study, the size of AVM was smaller in patients who initially presented with hemorrhage. However, after the AVM were diagnosed, there was actually no difference in terms of risk of rebleeding in small or large AVM. Crawford et al found that small AVM more often presented with hemorrhage, but the rehemorrhage rate was comparable with 21% in small and 18% in large AVM.² It is likely that small AVM are less likely to cause neurological deficits by mass effects or steal phenomenon and would not be diagnosed until rupture. A similar observation was noted in deeply located AVM, which are less likely to present symptoms other than hemorrhage.

Does Radiosurgery Reduce the Hemorrhage Rate in Patients With AVM?

Histopathologic studies have shown that detectable radiation effects can occur as early as 3 months after radiosurgery and remain for years.²⁵ Thickening of the intima as well as proliferation and accumulation of subendothelial matrix decrease the tension of vessel wall.²⁶ Furthermore, thrombosis or occlusion of vascular channels in the nidi decreases the patent vessels; thus, less numbers of vessels are prone to rupture. Whether radiosurgery reduces the risk of AVM hemorrhage remains debated. The incidence of postradiosurgical hemorrhage has been reported between 2.1% and 5%.^14,18 Because there is no control group in all the studies, the best way to evaluate if radiosurgery has a protective effect for hemorrhage is to compare the postradiosurgical hemorrhage rates with the 2% to 4% hemorrhagic risks obtained from published natural history studies. Obviously, there does not seem to be a significant change of hemorrhage rates in patients with AVM treated with radiosurgery until total obliteration is achieved.

Based on our data, the postradiosurgery hemorrhage rate is 2.5%, which seems to be lower than the preradiosurgical rate of 6.6% calculated after the diagnosis of the AVM. We cannot exclude that the preradiosurgical hemorrhage rate is an overestimate and the decline of hemorrhage rate is not part of the natural course of the disease. However, none of the risk factors for AVM rupture remained relevant after GKS and nidi treated with a high radiosurgical dose were less likely to bleed implicates possible protective effects of GKS against AVM rupture.

Conclusions

The present study provides further evidence that patients with AVM remain at risk for hemorrhage as long as the nidus is patent after radiosurgery. However, radiosurgery seems to reduce the hemorrhage rate during the latency period. This phenomenon is more prominent in patients who present with hemorrhage before GKS.

Disclosure

None.

Footnotes