Predicting the natural history of unruptured brain arteriovenous malformations: external validation of rupture risk scores.

Abstract

Objective: The current management of brain arteriovenous malformations (bAVMs) is controversial. A group of patients will undergo conservative management assuming a higher risk from intervention than from natural history of the disease. However, patients with increased risk of future bleeding could benefit from intervention. Therefore, accurate prediction of rupture risk is needed. This study was aimed at identifying published rupture risk scoring systems and externally validating their performance in a single-center population.

Methods: This retrospective observational study included patients with bAVM who presented to the authors' institution during the period from January 2010 to December 2020. A literature search was performed to identify published rupture risk scales. Data relevant for scoring were extracted from the study population. Discrimination, classification, and calibration analyses were performed for all selected scales with ruptured presentation as outcome.

Results: A total of 269 bAVMs (in 258 patients) were included. Significant differences between ruptured and unruptured cases were observed for several variables including location, size, flow-related-associated aneurysms, single arterial feeder, and venous tortuosity. The following 4 scoring systems were selected: Nataf; race, exclusive deep location, AVM size < 3 cm, venous drainage exclusively deep and monoarterial feeding (R2eD AVM); AVM rupture index (ARI); and ventricular system involvement, venous aneurysm, deep location, and exclusively deep drainage (VALE). The performance of the Nataf and ARI scales was nondiscriminatory and that for R2eD AVM and VALE scales was poor. A logistic regression model was obtained from the component variables of selected scales and included the following variables: size, location, venous tortuosity, and ventricular involvement. Its performance was just fair.

Conclusions: Currently available rupture risk scoring systems showed nondiscriminatory to poor performance. Based on its observed performance, the R2eD AVM system could be preferentially used. Further research on rupture risk factors for derivation of better scales is needed.