Do the morphological characteristics of arteriovenous malformations affect the results of radiosurgery?

OBJECT: The authors sought to determine which morphological features of arteriovenous malformations (AVMs) are statistically predictive of preradiosurgical hemorrhage, postradiosurgical hemorrhage, and neuroimaging-defined failure of radiosurgical treatment. In addition, correlation between computerized tomography (CT) scanning and angiography for the identification of AVM structures was investigated.
METHODS: Archived CT dosimetry and available angiographic and clinical data for 268 patients in whom AVMs were treated with linear accelerator radiosurgery were retrospectively reviewed. Many of the morphological features of AVMs, including location, volume, compact or diffuse nidus, neovascularity, ease of nidus identification, number of feeding arteries, location (deep or superficial) of feeding arteries, number of draining veins, deep or superficial venous drainage, venous stenoses, venous ectasias, and the presence of intranidal aneurysms, were analyzed. In addition, a number of patient and treatment factors, including patient age, presenting symptoms, radiation dose, repeated treatment, and radiological outcome, were subjected to multivariate analyses. Two hundred twenty-seven patients were treated with radiosurgery for the first time and 41 patients underwent repeated radiosurgery. Eighty-one patients presented with a history of AVM hemorrhage and 91 patients had AVMs in a periventricular location. Twenty-six patients (10%) experienced a hemorrhage following radiosurgery. Of the 268 patients, 81 (30%) experienced angiographically defined cures, and 37 (14%) experienced MR imaging-defined cures. Eighty-six patients (32%) experienced neuroimaging-defined treatment failure, and 64 underwent insufficiently long follow up. A larger AVM volume (odds ratio [OR] 0.349; p = 0.004) was associated with a decreased rate of pretreatment hemorrhage, whereas periventricular location (OR 6.358; p = 0.000) was associated with an increased rate of pretreatment hemorrhage. None of the analyzed factors was predictive of hemorrhage following radiosurgery. A higher radiosurgical dose was strongly correlated with neuroimaging-defined success (OR 3.743; p = 0.006), whereas a diffuse nidus structure (OR 0.246; p = 0.008) and associated neovascularity (OR 0.428; p = 0.048) were each associated with a lower neuroimaging-defined cure rate. A strong correlation between CT scanning and angiography was noted for both nidus structure (p = 0.000; Fisher exact test) and neovascularity (p = 0.002; Fisher exact test).
CONCLUSIONS: Patients presenting with AVMs that are small or periventricular were at higher risk for experiencing hemorrhage. A higher radiosurgical dose correlated strongly with neuroimaging-defined success. Patients in whom the AVM had a diffuse structure or associated neovascularity were at higher risk for neuroimaging-defined failure of radiosurgery. A strong correlation between CT scanning and angiography in the assessment of AVM structure was demonstrated.