Article Text

Download PDFPDF
E-039 Comparison of predictive grading systems for procedural risk in endovascular treatment of brain arteriovenous malformations – analysis of 104 consecutive patients
  1. B Pulli,
  2. C Stapleton,
  3. M Koch,
  4. A Patel
  1. Massachusetts General Hospital, Boston, MA


Introduction Endovascular embolization of brain arteriovenous malformations (AVMs) is widely utilized, often used in conjunction with micro- and/or radiosurgery. Grading systems to assess procedural risks of endovascular embolization have been proposed, but none has been independently validated. We sought to validate and compare these grading systems in 104 consecutive patients with brain AVMs who underwent endovascular embolization between 2003 and 2016 at our tertiary academic referral center.

Methods Clinical and demographic data were obtained from the medical records. Cerebral angiograms were reviewed and Buffalo, AVM Neuroendovascular (AVMN), and Spetzler-Martin (SM) grades determined. Clinical outcomes and complications were collected. Receiver operating characteristics (ROC) curve analysis was performed, and areas under the curve (AUCs) for different grading systems compared with DeLong’s test.

Results Forty-five (43.3%) patients were females, with an average age at presentation of 43.2±16.2 years. Forty-nine patients (47.1%) presented with hemorrhage. Fifty-five (52.9%) AVMs were located on the left side, 40 (38.5%) were in an eloquent brain region, and mean AVM size was 3.4±1.4 cm. There were 10 major and 17 minor complications in 25 patients (vessel rupture: 9; retained/fractured microcatheter: 5; off-target embolization: 4; extension of embolic cast into venous outflow: 2; artery dissection: 1; ischemic stroke presumed to be related to embolization: 6). Arterial pedicle size (p=0.002) and number of arterial pedicles (p=0.04) were predictors of complications, while AVM side, location in/near an eloquent brain region, patient age, AVM size, or venous drainage pattern were not. The Buffalo score was predictive of complications (p=0.004), but AVMN (p=0.23) and SM grades (p=0.35) were not (Table). ROC curve analysis revealed an area under the curve (AUC) of 0.68±0.05 for the Buffalo score, significantly better than AVMN (AUC 0.58±0.06, p=0.04) and SM grades (AUC 0.56±0.06, p=0.05, Figure).

Conclusion Our independent analysis of 104 patients with brain AVMs treated with endovascular embolization validates the predictive capacity of the Buffalo score, but not AVMN or SM grades, for endovascular embolization procedural risk. This will allow for better stratification of brain AVM patients by procedural risk, and will ultimately result in safer care of these patients.

Disclosures B. Pulli: None. C. Stapleton: None. M. Koch: None. A. Patel: 2; C; Penumbra, Medtronic.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.