Decision Analysis for Asymptomatic Lesions

Introduction

Intracranial arteriovenous malformations (iAVMs) are uncommon vascular lesions that most frequently are diagnosed following intracerebral hemorrhage or seizures. Even in the absence of hemorrhage, headache and focal neurological deficits may develop secondary to vascular steal or mass effect. Widespread use of neuroimaging and improvements in imaging techniques have resulted in a proportional increase in incidentally diagnosed iAVMs. Treatment centers on four main modalities: conservative or medical management, radiosurgery, endovascular embolization, and microsurgery. Treatment is often tailored to institutional practices, surgeon experience, and iAVM-specific features. In this chapter, we provide a summary and update of the current understanding of iAVM features and risk profiles along with the treatment options to help guide decision-making to appropriately balance the immediate risks of intervention against the lifetime risk of stroke in patients with asymptomatic AVMs.

Pathogenesis and Pathophysiology of iAVMs

Intracranial AVMs are complex congenital vascular lesions that arise from an abnormal development of the capillary network, resulting in direct connection between arterial feeders and venous drainage. The lack of resistance from a capillary bed leads to a high-flow system with attendant vascular dilation. This web of high-flow vasculature is known as the AVM nidus. The most feared complication of iAVMs remains rupture and hemorrhage. Even prior to rupture, iAVMs cause vascular reorganization of the surrounding brain parenchyma. Multiple AVM features have been associated with an increased risk of hemorrhage. These include aneurysms located within the iAVM nidus, venous drainage directly into the deep cerebral veins, venous varices, infratentorial location, and previous hemorrhages. The chronic nature of iAVMs leads to adaptation of the surrounding normal brain tissue to a state of chronic hypoperfusion. This adaptation makes excision of the AVM difficult, as abrupt removal of a high-flow circuit with subsequent shunting toward the normal hypoperfused brain parenchyma can lead to “normal perfusion pressure breakthrough,” first described by Spetzler et al. in 1978.

The etiology of iAVMs remains controversial. Multiple hereditary conditions are associated with the development of these lesions. Up to 25% of patients with hereditary hemorrhagic telangiectasia (HHT) develop iAVMs. HHT is thought to promote iAVM development via deficiency of transforming growth factor-beta as well as mutations within the RASA1 and EPHB4 genes associated with the RAS/ERK pathway. Importantly, mutations within the RASA1 gene have also been associated with familial iAVM development. These findings show that there is a clear genetic susceptibility for iAVM formation. No study has yet proven that iAVMs are present from birth, however, highlighting that genetic susceptibility may require a second hit for iAVM development. These insults may contribute to somatic mutations that can generate iAVMs. Sequencing of endothelial cells from human iAVMs showed KRAS or bRAF mutations in 10 of 16 specimens, and animal studies have shown that activating KRAS mutations within endothelial cells is sufficient for inducing AVMs in mice and embryonic zebrafish. In fact, iAVMs are characterized by increased inflammatory monocytes and microglia, with even unruptured AVMs showing histopathologic evidence of perivascular inflammation. Thus the most accepted theory on the development of iAVMs is based on a combination of genetic susceptibility and environmental insult (through stroke, trauma, radiation, seizures, etc.).

Natural History

Understanding the natural history of iAVMs is critical in deciding whether surgical intervention is warranted for asymptomatic lesions. Retrospective studies have shown that the incidence of AVM hemorrhage ranges between 0.55 and 0.86 cases per 100,000 person-years. The Olmsted study also highlighted that most patients will be symptomatic at presentation (60.4%) and that mortality following iAVM hemorrhage is high (17.6%). The New York Islands Study built off these retrospective studies with a prospective study examining iAVM incidence and rates of hemorrhage. The authors reported an annual iAVM detection rate of 1.34 per 100,000 person-years with a rate of hemorrhage of 0.51 per 100,000 person-years. The annual rate of hemorrhage was an average of 3% but was severely dependent on iAVM factors such as deep venous drainage, deep location, incidence of previous hemorrhage, and location within the brainstem; the annual hemorrhage rate for these high-risk iAVMs could be as high as 33%.

Since these landmark studies, multiple follow-up analyses have demonstrated annual rates of initial iAVM hemorrhage between 1.3% and 2.2%, with the rate greatly increasing for iAVMs that have previously hemorrhaged (4.5%–4.8%). Importantly, the meta-analysis by Kim et al. showed that the rate of hemorrhage is not predicted by iAVM size. Morgan et al. used these rates to calculate annual risk estimates for initial iAVM hemorrhage and showed that the 5-year risk ranges from 6% to 11% and the 10-year risk ranges from 12% to 20%. Hemorrhage of an iAVM after initial asymptomatic presentation has been associated with a 40% rate of morbidity and mortality. A meta-analysis performed by Gross and Du has confirmed that infratentorial iAVM location, deep venous drainage, and intranidal aneurysms are all features associated with increased risk of hemorrhage.

Risk Stratification and Grading Scales

The Spetzler-Martin grading system, first described in 1986, has been used to grade iAVMs based on the risk posed by microsurgical resection. This grading scheme is composed of five tiers based on the size of the lesion, the pattern of venous drainage, and the proximity of the lesion to eloquent brain. The Spetzler-Martin grading system has been well validated, as multiple studies have shown that patients with Spetzler-Martin grade I and II iAVMs have significantly lower morbidity and mortality rates following resection compared to patients with Spetzler-Martin grade IV and V iAVMs. The similarity in the risk profiles of Spetzler-Martin grades I and II and of grades IV and V led to the development of a simplified schema, the three-tier Spetzler-Ponce scale. Exclusive deep venous drainage has since been validated as an important predictor of outcomes following resection as well.

Pollock and Flickinger described a grading scale for predicting outcomes following radiosurgery. This grading scale identified five variables that predicted iAVM cure without the creation of new neurological deficits. Three of these variables survived multivariable regression and are outlined in the Pollock grading score equation: 0.1*iAVM volume in cm ³ + 0.02*(patient age in years) + 0.3*(lesion location: 0, frontal/temporal; 1, cerebellar, corpus callosum, intraventricular, parietal, occipital; 2, brainstem, thalamus, or basal ganglia). The Pollock equation has been validated for predicting outcomes following radiosurgery, whereas the Spetzler-Martin grade does not predict outcomes following radiosurgery. Two other accepted scales for predicting outcomes following radiosurgery in iAVM patients include the Virginia Radiosurgery AVM Scale (VRAS) and the Heidelberg scale. These scales are based on iAVM volume, eloquent location, and/or history of hemorrhage. The VRAS authors also highlighted that AVM diameter, nidus volume, and radiation dose significantly predicted AVM obliteration rates following radiosurgery. Importantly, recent comparison of these grading systems to generalized linear models for the prediction of successful treatment without new neurological deficit has shown that integer-based classification scores are outperformed by continuous grading scores, such as the radiosurgery-based AVM score (RBAS) or the proton radiosurgery brain AVM scale (PRAS).

Multiple grading scales exist for the assessment of procedural risk in the endovascular treatment of iAVMs. These include the Puerto Rico score, the Buffalo score, and the AVM embocure score (AVMES). All three incorporate the proximity to eloquent brain and the number of vascular pedicles as features of the grading system. The Buffalo score incorporates the pedicle diameter, AVMES uses overall iAVM size, and the Puerto Rico score is based on the presence of an arteriovenous fistula. While all scores predicted endovascular obliteration of iAVM, only the Buffalo score predicted complication rates following endovascular treatment. This points to the importance of iAVM pedicle diameter in guiding endovascular treatment.