The treatment of AVMs is primarily intended to eradicate the risk of potential hemorrhage. The indications and method of invasive treatment of AVMs require a thorough knowledge of AVM natural history and treatment risk. Treatment risk, particularly surgical risk, is defined by a variety of grading systems, the most popular of which is the Spetzler-Martin grading system (Table 20.2) . A predictive grading system for the endovascular management of AVMs is not yet established.
Untreated AVMs have a poor prognosis. For example, Ondra et al. prospectively followed 166 symptomatic but untreated AVM patients . The average follow-up was more than 20 years. During this time, 23% of the patients died from AVM hemorrhage and the combined annual morbidity and mortality was 2.7%. In addition, these investigators found that the life expectancy of patients harboring AVMs was significantly reduced. Among AVM patients, the mean age of death from all causes was 51 years com- pared to 73 years in the general population.
Following rupture of an AVM, the risk of early re-bleeding is significantly lower than that observed after aneurysm rupture. Consequently, AVM treatment can often be delayed for 1 or 2 months after AVM rupture, to allow the patient to recover and undergo treatment under optimal circumstances. However, some authors have observed a two to fourfold increase in the risk of re-bleeding during the first year after AVM rupture than in subsequent years . AVM rupture during pregnancy may have a more malignant natural history, having a high rate of recurrent hemorrhage during the same pregnancy . Consequently, when a ruptured AVM in a pregnant woman is surgically accessible, it should be removed as soon as possible to prevent devastating re-hemorrhage. Embolization and radiosurgery are not options during pregnancy because of potential radiation toxicity to the fetus.
The long-term risk of AVM bleeding for patients with symptomatic AVMs is estimated to be 2-3% per year . This risk may be higher among patients who have had more than one hemorrhage and also in children. Among patients with symptomatic, unruptured AVMs, the risk of hemorrhage is similar to that observed in AVMs that have previously ruptured . The presence of neurologic symptoms does not increase the risk of hemorrhage among patients who have not experienced AVM rupture. Mortality after each AVM rupture is approximately 10% . Long-term, the overall annual mortality rate per year from an AVM hemorrhage is estimated to be 1% for adults and 2% for children . The reasons for an apparently more malignant natural history of symptomatic childhood AVMs have not been defined.
Several anatomic and angiographic characteristics are thought to be associated with a greater risk of AVM hemorrhage. These factors include: (1) smaller AVM size, (2) deep hemi- spheric location, (3) intranidal aneurysms, (4) deep venous drainage and (5) draining vein stenosis. In contrast, angiomatous change, or angiographically identified vascular hyperplasia of pial vessels being recruited to feed a high-ﬂow
AVM, may be associated with a decreased risk of hemorrhage. It is postulated that high- ﬂow AVMs, identified by these angiomatous changes, have low intra-nidal pressures, thus decreasing the risk of AVM rupture.
Between 3 and 14% of AVMs are associated with aneurysms, potentially altering AVM man- agement. High blood ﬂow and subsequent hemodynamic stress may promote aneurysm formation. Aneurysms associated with AVMs may be subdivided as follows: (1) dysplastic, remote aneurysms located at some distance from the AVM that appear anatomically unre- lated to major AVM inﬂow vessels, (2) proximal aneurysms that arise from the circle of Willis or on the proximal portion of major AVM feeding vessels, (3) pedicular aneurysms located on the middle portion of a major feeding pedicle or (4) intra-nidal aneurysms within the AVM. The natural history of these combined lesions is not well understood; however, several studies suggest that the combination of an AVM and cerebral aneurysm is associated with a greater risk of intracranial hemorrhage . For example, Brown et al. observed an annual hemorrhage rate of 2% for AVMs, but 7% among AVMs associated with aneurysms . In general, when these lesions come to attention in the absence of hemorrhage, the aneurysm should be treated first because of the higher morbidity associated with aneurysm rupture. For patients who present with hemorrhage, the symptomatic lesion should be treated first.
A variety of anatomic and physiologic factors, such as AVM size and location, number and distribution of arterial feeders, pattern of venous drainage and ﬂow through the AVM nidus, inﬂuence the technical difficulty and con- sequent risk of surgical, endovascular or radio- surgical treatment of an AVM. These factors have been incorporated into a variety of grading systems that are used primarily in treatment planning to predict surgical risk. Among these various grading systems, the Spetzler-Martin grading system is now most frequently used to predict surgical risk and compare results of clin- ical series using different treatment modalities (Tables 20.2 and 20.3). The Spetzler-Martin grading system uses three radiographic vari- ables: size of the AVM nidus, pattern of venous drainage and location of the AVM in relation to eloquent cortex. A numerical score is assigned to each variable and one derives the AVM grade,
I through V, by adding the scores. Large AVMs
that are intimately associated with eloquent cortex or that are located in the hypothalamus or brainstem may be classified as inoperable Grade VI lesions. Increasing grade is signifi- cantly associated with minor and major neuro- logic deficits post-operatively. Numerous studies have confirmed the reliability of this grading system in predicting surgical outcome. Unfortunately, this grading system is not directly applicable to the endovascular treat- ment of AVMs, contributing to the difficulty in comparing treatment results of these two modalities.