Interobserver Agreement on Arteriovenous Malformation Diffuseness Using Digital Subtraction Angiography

Background: Arteriovenous malformation (AVM) diffuseness has been shown to be prognostic of treatment outcomes. We assessed interobserver agreement of AVM diffuseness among physicians of different specialty and training backgrounds using digital subtraction angiography (DSA).

Methods: All research protocols were approved by the institutional review board for this retrospective chart review. In a single-blinded setting, 2 attending neurosurgeons, 1 attending interventional neuroradiologist, and 1 senior neurosurgical resident rated 80 DSA views of 36 AVMs as either compact or diffuse. Individual interobserver agreement and subgroup agreement were analyzed using κ agreement and intraclass correlation coefficient.

Results: Disagreement regarding AVM diffuseness occurred in 43.8% of all DSA views (n = 80). Interobserver κ agreement on AVM diffuseness using DSA views among 4 physicians ranged from fair (κ = 0.40 [95% confidence interval (CI) = 0.22-0.58]) to substantial (κ = 0.65 [95% CI = 0.48-0.81]), whereas total intraclass correlation coefficient was 0.81 (95% CI = 0.73-0.87). For the 36 AVMs, κ agreement ranged from fair (κ = 0.36 [95% CI = 0.13-0.60]) to moderate (κ = 0.57 [95% CI = 0.35-0.79]), whereas intraclass correlation coefficient among all 4 physicians was 0.68 (95% CI = 0.47-0.82). Moderate agreement on AVM diffuseness (n = 80) was found between attending and resident assessments (κ = 0.57 [95% CI = 0.39-0.75]) and between neurosurgeon and interventional neuroradiologist assessments (κ = 0.55 [95% CI = 0.37-0.73]).

Conclusions: Agreement of individual physicians on AVM diffuseness varies from fair to substantial. Objective and three-dimensional measures of AVM diffuseness should be developed for consistent clinical application.