Objective: The radiosurgery-based arteriovenous malformation (AVM) grading scale was developed to predict patient outcomes after radiosurgery. The purpose of this study was to determine whether simplifying this grading system using location as a two-tiered variable detracted from the accuracy of the scale.
Methods: Regression analysis modeling on 220 patients who underwent AVM radiosurgery between 1987 and 1992 at the University of Pittsburgh Medical Center using location as a two-tiered variable resulted in the following equation: AVM score = (0.1) (volume, mL) + (0.02) (age, yr) + (0.3) (location, hemispheric/corpus callosum/cerebellar = 0; basal ganglia/thalamus/brainstem = 1). Testing of the modified grading system was performed on 247 patients who underwent AVM radiosurgery between 1990 and 2001 at the Mayo Clinic. The mean modified AVM score was 1.62. The mean duration of patient follow-up was 70 months.
Results: There was no difference between the original and modified radiosurgery-based AVM scale with regard to AVM obliteration without new neurological deficits (F = 0.92, P = 0.53) or decline in Modified Rankin Scale (F = 0.83, P = 0.56) after radiosurgery. The modified radiosurgery-based AVM scale correlated with the percentage of patients with AVM obliteration without new deficits (<or=1.00, 89%; 1.01-1.50, 70%; 1.51-2.00, 64%; >or=2.00, 46%) (r = -0.98, P < 0.01) and a decline in Modified Rankin Scale (<or=1.00, 0%; 1.01-1.50, 13%; 1.51-2.00, 20%; >2.00, 36%) (r = 0.99, P < 0.01).
Conclusion: Simplifying the radiosurgery-based AVM grading system using location as a two-tiered variable did not detract from the accuracy of the scale. This system has been validated by numerous centers performing both gamma knife- and linear accelerator-based procedures and should be used in future studies on AVM radiosurgery to stratify patients for more accurate comparative analyses.