Development and validation of a prediction model for neurological outcomes of brain arteriovenous malformations undergoing microsurgical resection: a nationwide retrospective cohort study.

OBJECTIVE: Accurate prediction of neurological outcomes after microsurgical resection of brain arteriovenous malformations (AVMs) remains challenging because existing grading systems have only modest discriminatory power. In this study, the authors aimed to develop and validate a novel prediction model for neurological outcomes after AVM microsurgery. METHODS: This prognostic cohort study utilized data from the MATCH (Multimodality Treatment for Brain Arteriovenous Malformation in Mainland China) registry from August 2011 to December 2021. The authors developed a novel prediction model, the AVM neurological outcome scale (AVM-NOS), via multivariable logistic regression analysis in a derivation cohort including 1151 patients from Beijing Tiantan Hospital and validated in a multicenter external cohort of 75 patients. The primary outcome was unfavorable neurological outcome (modified Rankin Scale score > 2) at the final clinical follow-up. The authors evaluated the AVM-NOS against the Spetzler-Martin and Lawton-Young grading systems in terms of discrimination (area under the receiver operating characteristic curve [AUC]), calibration, and decision-curve analysis. RESULTS: The AVM-NOS incorporated six independent predictors associated with unfavorable outcomes: age, AVM volume, eloquence reclassification, deep perforating artery supply, exclusively deep venous drainage, and venous aneurysm. The scale exhibited superior predictive performance, with an AUC of 0.77 (95% CI 0.71-0.83) in the derivation cohort and 0.86 (95% CI 0.77-0.96) in the validation cohort, compared to the Spetzler-Martin grading scale (0.67 [95% CI 0.61-0.73] in the derivation cohort and 0.67 [95% CI 0.50-0.84] in the validation cohort) and Lawton-Young grading scale (0.72 [95% CI 0.67-0.78] in the derivation cohort and 0.75 [95% CI 0.58-0.92] in the validation cohort). Comprehensive model validation through calibration curves and decision curve analyses further confirmed the robust predictive performance and clinical applicability of the AVM-NOS. CONCLUSIONS: By integrating a refined eloquence classification, quantitative lesion volume, and key vascular features, the AVM-NOS surpasses existing models in predicting postoperative neurological outcome.