Interobserver Agreement of R₂eD Brain Arteriovenous Malformation Scoring

The R₂eD arteriovenous malformation (AVM) score is externally validated for use in assessing hemorrhage risk in unruptured brain AVMs, supplementing surgical risk scores and clinical decision-making. This study aimed to measure interobserver agreement of R₂eD AVM scoring.

R₂eD AVM scoring was performed independently by 3 neurosurgeons with similar training on an anonymized imaging dataset. Raters scored angiographic components of the R₂eD score (location, nidus size, venous drainage, and number of arterial feeders) as well as their level of confidence in scoring. Fleiss kappa (κ) was used to assess interobserver agreement. Strength of agreement was defined as follows: poor, κ < 0; slight, κ = 0.0-0.20; fair, κ = 0.21-0.40; moderate, κ = 0.41-0.60; substantial, κ = 0.61-0.80; and almost perfect, κ = 0.81-1.00.

The 3 raters reviewed 40 brain AVMs. Total R₂eD score demonstrated only slight interobserver agreement, and this relationship did not reach statistical significance (κ = 0.036, P = .52). Of the components that constitute the total R₂eD score, deep location was associated with the greatest amount of agreement (κ = 0.689, P < .001). Scoring of venous drainage (κ = 0.400, P < .001) and arterial feeder number (κ = 0.284, P = .002) had fair agreement, whereas nidus size demonstrated poor interobserver agreement (κ = -0.222, P = .02).

Scoring of individual R₂eD AVM score components was associated with high interobserver variability. Further research efforts are needed to identify sources of variability in R₂eD AVM scoring, including familiarity with the tool, differential imaging access, and the inherent complexity of AVM angioarchitecture. Scoring standardization could assist in hemorrhagic risk stratification and help guide clinical decision-making in patients with unruptured brain AVMs.